|
Now with MMR data.
I recently delved into a database containing around 8 million replays. A huge thanks to Dakota_Fanning for granting me access to the data from repmastered.app.
Introduction
We're all aware of the slight imbalances in various matchups. For instance, there's a ~53% win rate for Terran in TvZ, with similar minor imbalances in other matchups. While these numbers can vary based on factors like maps, spawn locations, player skill, and the current meta, they are generally recognized and accepted by the community.
With the data at hand, I tried to examine how some of these factors might influence the result of a game. Specifically, I examined how win rates in non-mirror matchups change as the game progresses over time.
In essence, we'll be looking at the win rate segmented into 1-minute intervals.
For example, for the 7-8 minute interval, we only consider games that concluded between 7:00 and 8:00 in-game time. This gives us an idea of the likelihood of a particular race winning at that specific time in the game. By doing this for every 1-minute interval, we can observe the win rate over time.
Before diving into the results, let's understand the data used. You can skip this part if you're only interested in the results.
Data
+ Show Spoiler +The dataset comprises roughly 8 million 1v1 games played since the start of 2018. It doesn't provide complete information from a replay but rather some extracted data. Build order or income details were not available.However, the dataset did include: - Player races
- Game winner
- Game duration
- Spawn locations
- Player APM & effective APM
To refine the dataset, a few filters were applied: - Games played after 01.01.2018
- Game duration > 2 minutes
- Exclude draws
- Exclude games with afk players
- Exclude games on fastest maps and similar
- Exclude games on maps with fewer than 20000 plays
Below is a histogram displaying the frequency of each map after applying these filters. ("13" represents Fighting Spirit 1.3) + Show Spoiler +
Win rates vs. game time
As mentioned earlier, the y-axis displays the win rate (colored and linked data points, label on the plot's left side) against game time on the x-axis. A dashed, green line represents the 50% win rate, helping us identify when matchup dynamics shift.
The grey background data, with axis labels on the right, shows how many games conclude in each interval. This helps gauge the frequency of games ending at specific times and how often they progress to certain stages.
PvZ + Show Spoiler + ![[image loading]](https://i.ibb.co/gV5KS5p/racial-win-rate-vs-time-pvz.png) This particular plot initially sparked my interest, so let's start here. Key observations include: - Zerg dominates the game's early stages. - The significant dip around 7-8 minutes can likely be attributed to hydra bust builds. This is the most striking example I've seen in the data so far of a single strategy or build having such a profound impact. - Protoss takes the lead during the mid-game, peaking around 13-16 minutes. This surge might partly result from the hydra bust builds: games that didn't end earlier are likely ones where Protoss successfully fended off the bust, often securing a favorable position. - From 16-20 minutes, Protoss's win rate drops, with Zerg taking the upper hand. Several factors might contribute, such as Protoss typically mining out their main and natural bases around 16 minutes, while Zerg continues mining at (at least) three bases until roughly 20 minutes. The shift also reflects Zerg's hive tech coming into play. - Games extending beyond 30 minutes show a slight uptick for Protoss, but these instances are rare and may not be very significant. Overall, while there weren't any major surprises, it was nice to quantify these trends. Comparing close spawn vs. cross spawn win rates will be interesting. A key takeaway is that Protoss might benefit from trying to wrap up games while on two bases, as Zerg seems favored in later stages.
PvT + Show Spoiler +![[image loading]](https://i.ibb.co/qBv8LW4/racial-win-rate-vs-time-pvt.png) Breaking this down chronologically: - Protoss has the edge in the game's very early phases (up to 6 minutes). - A notable number of games conclude between 6-8 minutes, probably due to Protoss timings. - Terran seems to have a strong timing around 10-12 minutes, which appears to be an optimal window for them to end the game. - The small bump in Protoss's win rate around 14 minutes is not completely clear to me; it seems to wane a few minutes later. Maybe it represents games where Protoss successfully defends against Terran's 10-12 minute push? - After 12 minutes, Terran's win rate starts to decline, with games lasting over 20 minutes favoring Protoss. PvT is the matchup where I have the least knowledge overall, so maybe someone else will have better insights on this?
TvZ + Show Spoiler +![[image loading]](https://i.ibb.co/xGVyrmT/racial-win-rate-vs-time-tvz.png) Again, let's look at this chronologically: - The first few minutes are dominated by wins due to zerglings - Terran wins quite a bit at 5-7 minutes, probably mostly sunken busts? - Zerg has a strong timing at around 7 minutes when lair tech kicks in - If Terran doesn't lose to zerg lair tech timing, they have a really great time until about 12-13 minutes, when zerg finally gets defilers. This pre-defiler timing is probably the most dominant point in the game out of any matchup - both the win rate for Terran is very high and the number of games which end there is large. - If Zergs can stabilize, they slowly but surely flip the matchup dynamics at the 12-20 minute period. - And thus we enter a period of advantage for Zerg after 18 minutes. We should note here, that fewer games come to this point where zerg can reap their advantage. - However, we also see a clear trend, where, if Zerg is not able to finish out the game once they unlock the full hive tech (defiler + upgraded ultra) at ~20 minutes, as the games drag on, the matchup becomes balanced again. Here it was cool to see how the matchup flips multiple times. Also, if you just visually compare the shape of the win rate vs time of TvZ and PvZ, you'll see striking similarities. In both matchups, Zerg seems to have strong early timings, then gets dominated in the mid-game, but comes back at a later point.
Influence of spawn locations: close vs cross-spawn
Here we use 4 player maps only and compare the winrate on close spawning positions to the diagonal case. This is of particular interest from a map making point of view, since the rush distances have a big influence on how strong certain builds are.
PvT+ Show Spoiler +![[image loading]](https://i.ibb.co/QQVhGYZ/racial-win-rate-vs-time-pvt-selectioncomparison-4p-Close-Cross.png) Zealot rushes are strong on close positions, as expected. Looks like Protoss overall prefers diagonal spawns. The longer distance means that it takes longer for the Terran to reach him.
PvZ+ Show Spoiler +![[image loading]](https://i.ibb.co/CVpQvN5/racial-win-rate-vs-time-pvz-selectioncomparison-4p-Close-Cross.png) Protoss feels a lot safer in the early game on diagonal spawns. Later in the game Protoss also seems to profit from diagonal spawns.
TvZ+ Show Spoiler +![[image loading]](https://i.ibb.co/yQ7059M/racial-win-rate-vs-time-tvz-selectioncomparison-4p-Close-Cross.png) Early Terran wins due to sunken busts are much more likely on close spawns. Then, for the most part of the game the distances don't seem to matter at all.
Influence of player skill effective apm
No MMR data is available to us as of right now, so we'll be using the effective apm instead to estimate the players skill.
Results in brackets of effective apm. + Show Spoiler +.
Here we can see a few categories of players and the balance in their games. A lower-level group is defined by taking games where both players are slower than roughly half of all the players, that is effective apm below 150. Those are the red data points.
Then we have two groups where both players are required to be above some certain level: both above 150 in purp, and both above 180 in blue. These are the higher-level groups. Note that one is inclusive of the other here, so we expect a strong correlation.
PvT+ Show Spoiler +![[image loading]](https://i.ibb.co/583C3RC/racial-win-rate-vs-time-pvt-selectioncomparison-eapm-Groups.png) Man wtf is happening at the 5 minute mark in the lower-level group. Is this the delayed Zealot rush hitting? Overall, slow Terrans get punished terribly by slow Protosses. For the better groups, it seems like if Terran can survive to the 10th minute, they will have a comfortable 10 minutes of advantage to win the game. If they do not manage that, the game tilts towards Protoss again,
PvZ+ Show Spoiler +![[image loading]](https://i.ibb.co/Zd4zF5X/racial-win-rate-vs-time-pvz-selectioncomparison-eapm-Groups.png) Zealot rushes only work in the lower group. Hydras are deadly at all levels. It looks like the lower group Zergs are worse in converting their late-game advantage into a win compared to the higher groups.
TvZ+ Show Spoiler +![[image loading]](https://i.ibb.co/2gpXD2B/racial-win-rate-vs-time-tvz-selectioncomparison-eapm-Groups.png) Again, Terran profits the most from just "getting better". Looking at the 5-6 minute mark we nicely see that the higher groups gain their wins from sunken busting at an earlier time than the lower group. Terran dominance before defiler looks really scary in the higher groups.
Influence by level of play based on MMR.
Here, MMR data is used to examine the balance. A match-mmr is determined by taking the average MMR of the two players (initial MMR values before the match was played). Several brackets of games are defined and compared. Additionally, a requirement is imposed on the max difference of player MMR, to ensure only games of a meaningful skill difference are taken into account. For example, in red are games with
- match-mmr below 1800, and difference of MMR of both player is below 200
The requirements are somewhat loosened as games of higher level are examined, as can be seen in the labels. Take note of the big error bars in some of the brackets in the later stages of the game; that is don't draw too many conclusions from those data points.
PvT+ Show Spoiler +![[image loading]](https://i.ibb.co/nmZ06cJ/racial-win-rate-vs-time-pvt-selectioncomparison-MMR-Groups.png) The game balance is even at an MMR above 1800 and tils slowly towards Terran the higher we go. At the highest level the matchup is roughly 52% in Terrans favour.
PvZ+ Show Spoiler +![[image loading]](https://i.ibb.co/Qb40CXm/racial-win-rate-vs-time-pvz-selectioncomparison-MMR-Groups.png) Protoss is behind in all brackets and gets dominated at the highest levels where they achieve an overall win rate of 46%.
TvZ+ Show Spoiler +![[image loading]](https://i.ibb.co/RCDHCv2/racial-win-rate-vs-time-tvz-selectioncomparison-MMR-Groups.png) Again we see a clear trend where the matchup becomes more favoured towards Terran as player skill increases. It looks like high-level Zerg vs Terran is the hardest matchup in broodwar.
|
This is fascinating, great work
|
Very interesting. Thanks for the info!
|
Thanks a lot for for this. I've always wanted to see this kind of "win probability vs time" analysis.
Regarding PvT
- The small bump in Protoss's win rate around 14 minutes is not completely clear to me; it seems to wane a few minutes later. Maybe it represents games where Protoss successfully defends against Terran's 10-12 minute push?
imo this is probably the result of the 1 Arbiter stasis with max army timing.
After 20 minutes PvT favors Protoss probably because for 90% of the player base, Terrans can't manage to play split map and manage more than 4 bases at a time.
I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank
|
|
|
Amazing post.
My big takeaway is that Artosis was right about PvT  .
I would love to see a ZvZ plot about game length and frequency. What are the odds of lategame ZvZ? It might be interesting to see for other mirrors as well.
|
On October 11 2023 08:32 TMNT wrote:Thanks a lot for for this. I've always wanted to see this kind of "win probability vs time" analysis. Regarding PvT Show nested quote +- The small bump in Protoss's win rate around 14 minutes is not completely clear to me; it seems to wane a few minutes later. Maybe it represents games where Protoss successfully defends against Terran's 10-12 minute push? imo this is probably the result of the 1 Arbiter stasis with max army timing. After 20 minutes PvT favors Protoss probably because for 90% of the player base, Terrans can't manage to play split map and manage more than 4 bases at a time. I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank
The bump around 14 minutes is probably due to there not being a timing then. The spike for T wins from 9-13 minutes is probably from 2 base timings. 3 base 2-1 timings hit around 14 1/2 minute mark and P usually leaves a little bit after that (if it works) which is why the T spikes again between 16-21 minutes (this is probably just 2-1 timings working well and P scrambling and leaving at different points depending on how the game is going). Then P spikes in winning because at the 22 minute mark, P should know if they're going to win or not based on how bad they got smashed or how bad they held the Terran timing attacks and if they're still in the game beyond 22 minutes, they clearly held and are in a decent enough position to continue.
Overall, pretty impressive research and quite interesting. I feel like I can pin point certain builds as to why most of those are the way that they are but it's interesting for it to not be theory and for it to actually have data to point to. It makes theorycrafting a lot clearer.
|
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work.
|
On October 11 2023 08:32 TMNT wrote:
I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank
On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work.
Just to add on to this, and I'm sure someone else has already done this, but could you show TvZ and TvP win rates with and without Flash?
|
Great work. Confirms my theory that PvZ is massively P favored in the midgame :D
|
On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work.
Based on OP's description of the existing data, its not. You would need to get new data and I don't know how easy that is to do. The way I would do it is to filter out the lowest effective APM players, maybe the bottom 60% or so, since those are disproportionately going to be the worst players. It's far from a perfect proxy, but it would at least keep a lot of D and C rank games from polluting the data.
|
|
|
Thanks this is amazing!
This sort of solidifies the fact TvP is the hardest matchup. There are so many ways Terran can die to Protoss in the early-mid game but not the other way around. Once the game goes past 20 mins it gets favoured for Protoss again due to the number of bases Protoss has and the option to switch to carriers.
If you run it at the top level I would think the early-mid game win rates for Protoss lowers abit but overall it wouldn’t change too much.
|
On October 11 2023 16:41 Dante08 wrote:
Thanks this is amazing!
This sort of solidifies the fact TvP is the hardest matchup. There are so many ways Terran can die to Protoss in the early-mid game but not the other way around. Once the game goes past 20 mins it gets favoured for Protoss again due to the number of bases Protoss has and the option to switch to carriers.
If you run it at the top level I would think the early-mid game win rates for Protoss lowers abit but overall it wouldn’t change too much.
i dont disagree that TvP is the hardest, but this data doesn't really prove that. All of your arguments equally apply to PvZ (easy to die early, and then zerg pulls ahead again in the late game).
|
Dakota_Fanning
Hungary2341 Posts
Amazing job! Nice stats!
Don't throw away your tools you generated the stats with, as we might repeat this with a bigger dataset, which might also include matchmaking stats I'm gathering recently (e.g. MMR).
|
On October 11 2023 13:37 imBLIND wrote:Show nested quote +On October 11 2023 08:32 TMNT wrote:
I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank Show nested quote +On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work. Just to add on to this, and I'm sure someone else has already done this, but could you show TvZ and TvP win rates with and without Flash?
Assuming equal distribution between zvp/tvz/pvt, we're talking about 2.3Million games in each matchup. Excluding any one player, no matter the name, shouldn't make a dent. Out of the sheer volume of games analyzed.
Great work OP! It's very interesting to see the winrate evolution over game time.
|
moktira
Ireland1542 Posts
This is really amazing, well done on doing this, so cool to see how it evolves in time rather than just %winrate.
As a few others have expressed it would be very interesting to see for the higher APM users though to get an idea of what it's like among the top players.
|
I've noticed post 40 PvZ protoss becomes extremely resource efficient compared to zerg as the map mines out and toss gets a chance to use reavers, archons, and high templars to defend their last couple bases at a very very low cost. The zerg on the other hand has to spam large volumes of units trying to break into a fortified position with units that die in the blink of an eye.
|
On October 11 2023 13:37 imBLIND wrote:Show nested quote +On October 11 2023 08:32 TMNT wrote:
I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank Show nested quote +On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work. Just to add on to this, and I'm sure someone else has already done this, but could you show TvZ and TvP win rates with and without Flash?
You already have it. T win rate without Flash is literally the current win rate on eloboard, since the site started around the time he left for military. Then compare it with whatever you have before that point. I looked up a while ago and they're basically unchanged.
|
On October 11 2023 18:11 angry_maia wrote:Show nested quote +On October 11 2023 16:41 Dante08 wrote:
Thanks this is amazing!
This sort of solidifies the fact TvP is the hardest matchup. There are so many ways Terran can die to Protoss in the early-mid game but not the other way around. Once the game goes past 20 mins it gets favoured for Protoss again due to the number of bases Protoss has and the option to switch to carriers.
If you run it at the top level I would think the early-mid game win rates for Protoss lowers abit but overall it wouldn’t change too much. i dont disagree that TvP is the hardest, but this data doesn't really prove that. All of your arguments equally apply to PvZ (easy to die early, and then zerg pulls ahead again in the late game).
The PvZ win area is much bigger than TvP no?
|
On October 11 2023 18:11 angry_maia wrote:Show nested quote +On October 11 2023 16:41 Dante08 wrote:
Thanks this is amazing!
This sort of solidifies the fact TvP is the hardest matchup. There are so many ways Terran can die to Protoss in the early-mid game but not the other way around. Once the game goes past 20 mins it gets favoured for Protoss again due to the number of bases Protoss has and the option to switch to carriers.
If you run it at the top level I would think the early-mid game win rates for Protoss lowers abit but overall it wouldn’t change too much. i dont disagree that TvP is the hardest, but this data doesn't really prove that. All of your arguments equally apply to PvZ (easy to die early, and then zerg pulls ahead again in the late game).
not equally. the winrate bump in the 10-20th ish min for protoss in PvZ is clearly way more pronounced than the terran one in PvT. terran's "compensation" through their 2/3 base timing push in TvP is seemingly not as effective as protoss's 2 base pushes in PvZ.
its actually surprising to me how for all the crying about hydra busts protoss have their own hydra bust esque win spike in the early stage of PvT. even with toss early game shenanigans i never thought it would be just as bad
|
On October 11 2023 19:14 RJBTV wrote:
I've noticed post 40 PvZ protoss becomes extremely resource efficient compared to zerg as the map mines out and toss gets a chance to use reavers, archons, and high templars to defend their last couple bases at a very very low cost. The zerg on the other hand has to spam large volumes of units trying to break into a fortified position with units that die in the blink of an eye.
yeah ultra lategame PvZ is known to be heavily toss favoured for the general reasons u mention but i guess theres not enough sample size
|
On October 11 2023 16:41 Dante08 wrote:
This sort of solidifies the fact TvP is the hardest matchup..
It does nothing of the sort, lol
|
You guys have to reserve all the balance discussion or which matchup is harderst sort of things until the analysis with regard to player skill comes out.
For example, even without this data we already know Protoss dominates at almost every level until S rank. Even BSL is dominated by Protoss.
its actually surprising to me how for all the crying about hydra busts protoss have their own hydra bust esque win spike in the early stage of PvT. even with toss early game shenanigans i never thought it would be just as bad
Another example, the near 70% win rate around 3-4 minutes in PvT could very well be about Terran noobs not being able to handle the first Zealots lol. I don't think there are a lot of pro games where the games end that soon.
|
Also another thing to consider: I don't think the data from repmastered.app contains much of the progamers/Korean portion. The spon games between progamers are custom games which are held between themselves that even cwal.gg can't have access. Only recently do we have access to progamers' replays when they clash with each other on the ladder.
|
On October 11 2023 20:14 TMNT wrote:
Another example, the near 70% win rate around 3-4 minutes in PvT could very well be about Terran noobs not being able to handle the first Zealots lol. I don't think there are a lot of pro games where the games end that soon.
yes i know i mean thats the point, if u are talking about ANY game balance u have to care about the lower levels not just pro levels. how to get people to buy your games or have a community if u don't care that the effort reward ratio is disproportionately skewed in early game TvP just like in ZvP? or i guess thats a moot point for old dead BW? if u lot are just boomers not wanting to make concessions then fine done.
|
On October 11 2023 20:24 ggsimida wrote:Show nested quote +On October 11 2023 20:14 TMNT wrote:
Another example, the near 70% win rate around 3-4 minutes in PvT could very well be about Terran noobs not being able to handle the first Zealots lol. I don't think there are a lot of pro games where the games end that soon. yes i know i mean thats the point, if u are talking about ANY game balance u have to care about the lower levels not just pro levels. how to get people to buy your games or have a community if u don't care that the effort reward ratio is disproportionately skewed in early game TvP just like in ZvP? or i guess thats a moot point for old dead BW? if u lot are just boomers not wanting to make concessions then fine done.
I mean, you're kind of confusing between points here.
If you want to care about lower levels and how to get people to buy the game, then you're talking about DESIGNING the game. It was already done 25 years ago and through several patches later. It can't be changed. It is what it is now.
Then if we are TALKING about balance alone, as in we're just discussing the game, you can either talk about it as a whole or talk about it at a specific level. And most of the times we're implicitly talking about it at high levels (like when we're discussing ASL or something).
Then the point you made about hydra bust in PvZ being comparable to the P win spike in early game PvT, well it's not comparable at all, since the former exists at high levels (maybe even at all levels?) but the latter only exists at noob levels - and it's not where the discussions are at all.
|
On October 11 2023 09:37 Monochromatic wrote:
[...]
I would love to see a ZvZ plot about game length and frequency. What are the odds of lategame ZvZ? It might be interesting to see for other mirrors as well.
Great idea, I'll add those later.
On October 11 2023 18:15 Dakota_Fanning wrote:
Amazing job! Nice stats!
Don't throw away your tools you generated the stats with, as we might repeat this with a bigger dataset, which might also include matchmaking stats I'm gathering recently (e.g. MMR).
No worries! I was hoping for you to say something like this, so in the future we maybe could re-run the replay parser to get more information from the replays and also include MMR ratings.
@ggsimida and TMNT: please wait for the upcoming data which will examine the behaviour in higher-level and lower-level games, it is going to be rather enlightening.
As already mentioned in the thread, currently no MMR information is available, thus we can't just get stats for S-rank. However, I think that the effective APM is sufficiently correlated with skill and rank, so we can use it as a proxy to create a sample of high-level and low-level games.
See here for normalized distributions of effective APM, really beautiful gaussian.
+ Show Spoiler +
|
On October 11 2023 21:30 Kraekkling wrote:
As already mentioned in the thread, currently no MMR information is available, thus we can't just get stats for S-rank. However, I think that the effective APM is sufficiently correlated with skill and rank, so we can use it as a proxy to create a sample of high-level and low-level games.
I am a bit skeptical about that. If you go through cwal.gg you'll probably see B C ranks players having the apm of progamers.
But the guy who runs cwal.gg does have stats for each specific mmr range, it was posted on reddit a while ago:
https://www.reddit.com/r/broodwar/comments/15nvvrk/matchups_by_winrate_including_2300_s_rank/
I suppose maybe you can grab the data for S rank only and extract the win rate vs time from it? Obviously need to contact him to ask for those data though.
|
Dakota_Fanning
Hungary2341 Posts
On October 11 2023 21:45 TMNT wrote:Show nested quote +On October 11 2023 21:30 Kraekkling wrote:
As already mentioned in the thread, currently no MMR information is available, thus we can't just get stats for S-rank. However, I think that the effective APM is sufficiently correlated with skill and rank, so we can use it as a proxy to create a sample of high-level and low-level games. I am a bit skeptical about that. If you go through cwal.gg you'll probably see B C ranks players having the apm of progamers. But the guy who runs cwal.gg does have stats for each specific mmr range, it was posted on reddit a while ago: https://www.reddit.com/r/broodwar/comments/15nvvrk/matchups_by_winrate_including_2300_s_rank/I suppose maybe you can grab the data for S rank only and extract the win rate vs time from it? Obviously need to contact him to ask for those data though.
I'm also skeptical about the rank-skill correlation. I see many S rank players that have a series of games where opponent leaves early (to artificially push the player up), and I also see many top players leave so they get to low ranks to stomp unskilled players.
While rank (MMR) should be (and probably is) a better indication of skill than APM, it often isn't.
|
Great work. I'd love to see the same data with only S rank and above filtered.
|
On October 11 2023 21:45 TMNT wrote:Show nested quote +On October 11 2023 21:30 Kraekkling wrote:
As already mentioned in the thread, currently no MMR information is available, thus we can't just get stats for S-rank. However, I think that the effective APM is sufficiently correlated with skill and rank, so we can use it as a proxy to create a sample of high-level and low-level games. I am a bit skeptical about that. If you go through cwal.gg you'll probably see B C ranks players having the apm of progamers. But the guy who runs cwal.gg does have stats for each specific mmr range, it was posted on reddit a while ago: https://www.reddit.com/r/broodwar/comments/15nvvrk/matchups_by_winrate_including_2300_s_rank/I suppose maybe you can grab the data for S rank only and extract the win rate vs time from it? Obviously need to contact him to ask for those data though.
You're right that there are players and games at B and C rank with the apm of programers.
I also looked at the apm distributions - those had a much more prominent tail at the high end of the distribution with entries at the 500-600 apm range, which is obviously due to spam. I don't have the plot at hand right now but I might post it later. In short: it didn't look pretty and you could see that there is a lot of spam going on. The effective apm distributions are much cleaner though, as can be seen in the post above.
I don't think this is the perfect way to do it - but it seems reasonable enough to try and see what comes out. It's also probably the best that can be done with the data at hand.
I'd also suggest you might take a look at the games on repmastered.app . I was mostly eyeballing the stats, but to me it seemed reasonable to assume that if we require something like
- both player have an effective apm above 180
we end up with almost exclusively A/S-rank games. Please feel free to suggest any improvements in this regard.
Regarding data from cwal.gg: we might get MMR data added to the current data anyway, once it's there it would be used.
|
On October 11 2023 20:14 TMNT wrote:You guys have to reserve all the balance discussion or which matchup is harderst sort of things until the analysis with regard to player skill comes out. For example, even without this data we already know Protoss dominates at almost every level until S rank. Even BSL is dominated by Protoss. Show nested quote +its actually surprising to me how for all the crying about hydra busts protoss have their own hydra bust esque win spike in the early stage of PvT. even with toss early game shenanigans i never thought it would be just as bad Another example, the near 70% win rate around 3-4 minutes in PvT could very well be about Terran noobs not being able to handle the first Zealots lol. I don't think there are a lot of pro games where the games end that soon.
Yes once you get to the pro level Terrans don’t die as easily but it still happens, I’ve seen it happen in ASL, ultimate battle and various other Korean pro games. Also when you flip it the other way around it’s true at pro level Terran almost never beats Protoss early game.
Now when you sample the absolute top level Terrans like Light Rush etc then things become different but we are not just looking at the absolute top level.
|
On October 11 2023 19:54 vOdToasT wrote:Show nested quote +On October 11 2023 16:41 Dante08 wrote:
This sort of solidifies the fact TvP is the hardest matchup.. It does nothing of the sort, lol
Care to explain, the TvP win area is by far the smallest in all matchups
Edit for typo
|
On October 11 2023 22:09 Dakota_Fanning wrote:Show nested quote +On October 11 2023 21:45 TMNT wrote:On October 11 2023 21:30 Kraekkling wrote:
As already mentioned in the thread, currently no MMR information is available, thus we can't just get stats for S-rank. However, I think that the effective APM is sufficiently correlated with skill and rank, so we can use it as a proxy to create a sample of high-level and low-level games. I am a bit skeptical about that. If you go through cwal.gg you'll probably see B C ranks players having the apm of progamers. But the guy who runs cwal.gg does have stats for each specific mmr range, it was posted on reddit a while ago: https://www.reddit.com/r/broodwar/comments/15nvvrk/matchups_by_winrate_including_2300_s_rank/I suppose maybe you can grab the data for S rank only and extract the win rate vs time from it? Obviously need to contact him to ask for those data though. I'm also skeptical about the rank-skill correlation. I see many S rank players that have a series of games where opponent leaves early (to artificially push the player up), and I also see many top players leave so they get to low ranks to stomp unskilled players. While rank (MMR) should be (and probably is) a better indication of skill than APM, it often isn't.
It doesn't matter for S rank and above though. S rank players may tank their mmr to stomp noobs, but the accounts in S rank still belong to legitimate S rank players.
A rank players who get accidentally bumped to S by leavers should not be too many. And even if there are a lot of them (I don't think there are though), it's A being mixed up with S which is still fine as far as "high skill" goes (think Artosis hitting 2400 a while ago). But more importantly, it occurs for all 3 races so it auto evens out.
You can't say the same for eapm. It's very normal for C and B ranks players (even D) to have ~ 200 eapm. So using eapm you get B and C mixed up with S.
|
It's very normal for C and B ranks players (even D) to have ~ 200 eapm.
This is not true. Can you in any way show this claim to be true or quantify what you mean by "very normal"?
|
On October 11 2023 22:44 Kraekkling wrote: This is not true. Can you in any way show this claim to be true or quantify what you mean by "very normal"?
Okay maybe "normal" is not the right word (as in most of them have 200 eapm) but what I mean is the proportion of lower rank players with high eapm is still large enough to affect the results. Of course I can't verify it with stats but if you just randomly browse the B-C rank sections in cwal you'll see a lot of (Korean) players around 180-200 eapm.
On the opposite side, you may have S rank players with less than 200 eapm too. For example just look at Stork and Soulkey here:
https://cwal.gg/players/gateway/30/player/Stork
https://cwal.gg/players/gateway/30/player/LC_needmoney
these dudes have sub 200 eapm all the time, some games even 160-170 eapm lol.
I'm quite sure that if filtered by eapm, you will leave the Storks and the Soulkeys out of the equation while adding in some random B rank players. At what proportion I dont know but what I'm saying is the data filtered by eapm is not as "clean" as you think.
|
I'm quite sure that if filtered by eapm, you will leave the Storks and the Soulkeys out of the equation while adding in some random B rank players.
This is obviously true and I can't think of any way to fix this without using MMR data.
Actually, we introduce another bias by applying a requirement on eapm. Since the distributions have different mean values but we impose the same selection for both races (e.g. above 180 eapm), we bias our selection samples for the races.
That is, if the mean is 160eapm for Terran but 150eapm for Zerg, you could say that we impose a higher requirement on the Zerg players and therefore bias our sample towards better Zerg players, compared to the Terran players sample.
There are even more problems to this, depending on how rigorous you'd like to be, but let's ignore that for now.
At what proportion I dont know but what I'm saying is the data filtered by eapm is not as "clean" as you think.
I will refrain from further replying to your posts until I posted the rest of the data, but as I previously mentioned, it is clear that this is not the best way to do it. But it's probably the best way to do it with the available data.
|
On October 11 2023 23:42 Kraekkling wrote:
I will refrain from further replying to your posts until I posted the rest of the data, but as I previously mentioned, it is clear that this is not the best way to do it. But it's probably the best way to do it with the available data.
Yes of course. If you can't have mmr data this is obviously the best way, if not the only way, to filter player skill. It would also be interesting if, once you can get access to the mmr data, to redo it for mmr and compare both sets of results.
And may also be interesting to do a mmr vs eapm comparison as well, although it should be more or less a positve correlation anyway, except from A/S rank it would probably flatten out.
|
yes well the problem is eAPM is also a flawed measure in BW which can easily be spammed, it has to do with unit movement/attack commands (protoss has less of this so their eAPM is usually less than t/z), so a zerg player will normally have more eAPM just because they're constantly microing mutas etc., basically the more time you spend on your units the higher your eAPM will be, at high lvls thats fine because progamers will also macro but at lower lvls that data is flawed because many players will sacrifice their macro for that... but i guess it's better than nothing
to me a sample size of like ~1k hand picked games of high lvl players is worth way more than that type of data
the thing about BW is you can't really unlock the "true form" of the game until you hit a certain lvl of execution, any data before that is pretty meaningless (that said it's still interesting to see the trends among all skill lvls) because it snowballs into the later stages of the game from a flawed state
to put it in perspective, imo even 2500 KR ladder players are playing a different game than the progamers are, 2500 (this number changes depending on the ladder season, in general it's around this range tho) is sort of the starting point of competitive BW play
|
On October 12 2023 00:27 TT1 wrote:
yes well the problem is eapm is also a flawed measure in BW which can easily be spammed, it has to do with unit movement/attack commands (protoss has less of this so their eapm is usually less than t/z), so a zerg player will normally have more eapm just because they're constatantly microing mutas etc... but i guess it's better than nothing
Actually in the data it can be seen that Protoss have higher eapm than Zergs. + Show Spoiler +
to me a sample size of like ~1k hand picked games of high lvl players is worth way more than that type of data
The problem with that is that with such a small sample, if you'd repeat this type of win rate vs game time analysis, you'll get completely dominated by statistical fluctuations so your data starts to look (almost) random.
Just for the win rates (no time dependence), you can refer to eloboard, where you have a sample of pro games.
|
Interesting. I have to agree with TT1, though; as a P main, I always had higher EAMP when off-racing as Z, mostly because mutalisks.
|
On October 12 2023 00:46 Kraekkling wrote:Show nested quote +On October 12 2023 00:27 TT1 wrote:
yes well the problem is eapm is also a flawed measure in BW which can easily be spammed, it has to do with unit movement/attack commands (protoss has less of this so their eapm is usually less than t/z), so a zerg player will normally have more eapm just because they're constatantly microing mutas etc... but i guess it's better than nothing
Actually in the data it can be seen that Protoss have higher eapm than Zergs. + Show Spoiler +Show nested quote +
to me a sample size of like ~1k hand picked games of high lvl players is worth way more than that type of data
The problem with that is that with such a small sample, if you'd repeat this type of win rate vs game time analysis, you'll get completely dominated by statistical fluctuations so your data starts to look (almost) random. Just for the win rates (no time dependence), you can refer to eloboard, where you have a sample of pro games.
yes Z won't have mutas most of the time and the type of muta micro is much different in ZvP <> ZvT
P has sairs/zeals which are heavy movement/attack units, it depends on the playstyle and the stage of the game (map control is key here), the early to middgame will be P eAPM intensive (relative to say PvT but that also depends on the playstyle, i.e: forge opener will be less eAPM intensive than 1g expo in PvZ) and most games will end around that stage as well (depending on whether or not P manages to slow Z down enough/bust)
like i said eAPM means something at high/pro lvl because players aren't sacrificing their macro for good unit movement, at lower levels they're sacrificing their macro the majority of the time tho
basically at lower lvls it comes down to how that player enjoys playing the game, some are more macro heavy others more micro heavy and that'll have a big impact on their eAPM, but ultimately in both cases their play is extremely flawed
|
Btw, from my gut feeling I also expected Protoss players to have lower eapm than Zergs.
|
Eapm has a massive anti Zerg bias. It counts pressing select larva multiple time as spam, even though that is an effective way to grab larva right as it hatches. Other races can just queue. Mutalisks and lings which are used in every match up require a lot of move commands. If you can give Zerg and exception for move commanding mutalisk, lings, and pressing select larva then it will be more accurate
|
This is great stuff, thank you for putting this together. Really looking forward to the spawn/skill plots.
Question: how do you detect who won the match? I thought this information was not included within the replay.
Also, more technical question here, just curious, I noticed you included error bars, are these the sample std deviation of your data?
|
Lol, tons of work man! Good job.
|
Thank you very much for the work!
|
I've added plots to the cross-spawn part and the apm group comparisons.
On October 12 2023 01:38 Volka wrote:
This is great stuff, thank you for putting this together. Really looking forward to the spawn/skill plots.
Question: how do you detect who won the match? I thought this information was not included within the replay.
Also, more technical question here, just curious, I noticed you included error bars, are these the sample std deviation of your data?
Yes its the standard error of the bin content.
|
With regards to mmr/eapm and such, couldn't you get a lot of leverage a simplistic ELO implementation? Is there enough info about player handles/accounts to do that?
Perhaps it's just easier to wait for mmr data, seems that will be forthcoming.
|
After looking at the apm vs winrate figure, I feel like there's something wrong with the data or treatment of data.
For example:
+ Show Spoiler +
In the overall figure, in the 14-15th minute interval, the win rate is exactly at 50%
+ Show Spoiler +
But in the apm figure, the win rates in that interval for the <150 and >150 apm groups are both over 50%. That can't be right?
Also you can see here that the curves for the >150 and >180 apm groups are almost the same. It shows you that apm progression doesn't reflect skill progression. Meanwhile in the stats posted by cwal, you can see the win rate shifts quite strongly with the progression of mmr:
+ Show Spoiler +
(image taken from reddit)
|
On October 12 2023 05:52 TMNT wrote:After looking at the apm vs winrate figure, I feel like there's something wrong with the data or treatment of data. For example: + Show Spoiler +In the overall figure, in the 14-15th minute interval, the win rate is exactly at 50% + Show Spoiler +But in the apm figure, the win rates in that interval for the <150 and >150 apm groups are both over 50%. That can't be right?
I can't give you a certain answer to this now but we are throwing away games where players do not fit any of the brackets, for example games where one player has an effective apm of 120 and the other of 160. This might account for it, I'm not sure why it should though. I might take a deeper look into whats happening here.
Also you can see here that the curves for the >150 and >180 apm groups are almost the same. It shows you that apm progression doesn't reflect skill progression. Meanwhile in the stats posted by cwal, you can see the win rate shifts quite strongly with the progression of mmr: + Show Spoiler +(image taken from reddit)
Feel free to draw whatever conclusion you find most interesting from the data - I'm just presenting the data.
On October 12 2023 05:01 sophisticated wrote:
With regards to mmr/eapm and such, couldn't you get a lot of leverage a simplistic ELO implementation? Is there enough info about player handles/accounts to do that?
Perhaps it's just easier to wait for mmr data, seems that will be forthcoming.
Yeah I'd rather use the mmr data if possible.
|
On October 11 2023 19:05 Branch.AUT wrote:Show nested quote +On October 11 2023 13:37 imBLIND wrote:On October 11 2023 08:32 TMNT wrote:
I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work. Just to add on to this, and I'm sure someone else has already done this, but could you show TvZ and TvP win rates with and without Flash? Assuming equal distribution between zvp/tvz/pvt, we're talking about 2.3Million games in each matchup. Excluding any one player, no matter the name, shouldn't make a dent. Out of the sheer volume of games analyzed. Great work OP! It's very interesting to see the winrate evolution over game time.
On October 11 2023 19:18 TMNT wrote:Show nested quote +On October 11 2023 13:37 imBLIND wrote:On October 11 2023 08:32 TMNT wrote:
I'm very much waiting for the analysis on player skill influence, as I expect we will some notable changes in the shape of those curves around S rank On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work. Just to add on to this, and I'm sure someone else has already done this, but could you show TvZ and TvP win rates with and without Flash? You already have it. T win rate without Flash is literally the current win rate on eloboard, since the site started around the time he left for military. Then compare it with whatever you have before that point. I looked up a while ago and they're basically unchanged.
You guys are absolutely right, Flash's games won't make a dent in 8.3 million total games. I should've been more specific in saying I'd like to see Flash's T v P/Z rates versus the other S-rank Terran's T v P/Z rates. Unfortunately, it appears that Kraekkling is unable to get just S-rank games, so my request is rather useless anyways T-T...
|
On October 12 2023 05:52 TMNT wrote:After looking at the apm vs winrate figure, I feel like there's something wrong with the data or treatment of data. For example: + Show Spoiler +In the overall figure, in the 14-15th minute interval, the win rate is exactly at 50% + Show Spoiler +But in the apm figure, the win rates in that interval for the <150 and >150 apm groups are both over 50%. That can't be right? Also you can see here that the curves for the >150 and >180 apm groups are almost the same. It shows you that apm progression doesn't reflect skill progression. Meanwhile in the stats posted by cwal, you can see the win rate shifts quite strongly with the progression of mmr: + Show Spoiler +(image taken from reddit)
I think it's caused by different samples. First graph counted all games; second one counted only those on 4p maps.
Speaking of which, I'd be very interested in winrate-by-length graphs for 2p, 3p and 4p maps respectively.
|
|
|
On October 12 2023 05:52 TMNT wrote:After looking at the apm vs winrate figure, I feel like there's something wrong with the data or treatment of data. Also you can see here that the curves for the >150 and >180 apm groups are almost the same. It shows you that apm progression doesn't reflect skill progression. Meanwhile in the stats posted by cwal, you can see the win rate shifts quite strongly with the progression of mmr: + Show Spoiler +(image taken from reddit)
PvT's taken into equation with eAPM are +/- 10% of PvTs from the 1st graph, and I think S rank is for 1% only if i'm not mistaken. eAPM is not super effective tool for measuring MMR as well. So here is your answer probably.
|
Until you break down the balance at an S rank these analyses are not useful at all. If you're C rank and struggle to beat a certain race, just get better at the game. It really is not out of anyones reach to get to S rank with any race. Then the problems start, because Terran becomes so overpowered that it dominates every high level ladder/high level tournament and that is an issue of balance.
|
Awesome analysis. Quite unfortunate that build orders are not available as it would be even cooler if we could break these stats down by build order matchups.
|
This is fantastic, looks great and is a great read thanks for the cool analysis!
I can understand if you are not accepting requests but if you have a todo list I would suggest two points.
As a visually regressed person it would be great to have the mirrored graphs for the matchups
If you are still going to look at the data in the future would you check some derivations of these plots?
- P(vZ) has around 10 minutes when the race is above 50%? From the looks of it out of 38 mins or so 1-8 -> Z, 9-19 -> P, 20-40 -> Z ~28 minutes go to zerg, and ~10 to protoss. But by the win percentage volume of those 10 minutes for P I'm not even sure it cancels out the first 8 minutes (and the hydra bust specifically) of Z slaughter
- Accumulating the win rates over time would also be an interesting comparison to those graphs
- Average rate of change for winrates and their direction could perhaps iindicate which parts of the matchup need to be revolutionised in the next 8 million games :D
Sorry, I got carried away by looking at your graphs. Really well done thanks for your good work!!
|
Certainly, those are a lot of games. Good statistics, man.
|
Dakota_Fanning
Hungary2341 Posts
I've already handed over a new dataset including 8.5 million games with MMR info included. Be patient until Kraekkling can analyze the new dataset and extract meaningful stats.
|
Based on this data, 6 pool i op zvp.
|
On October 13 2023 15:09 Dakota_Fanning wrote:
I've already handed over a new dataset including 8.5 million games with MMR info included. Be patient until Kraekkling can analyze the new dataset and extract meaningful stats.
How are you gonna know that data about 8.5m games? People change nicknames all the time and the MMR varies over time... Hard to imagine the MMR data to be any accurate.
|
Dakota_Fanning
Hungary2341 Posts
On October 13 2023 17:13 sexyMIStrZZZ wrote:Show nested quote +On October 13 2023 15:09 Dakota_Fanning wrote:
I've already handed over a new dataset including 8.5 million games with MMR info included. Be patient until Kraekkling can analyze the new dataset and extract meaningful stats. How are you gonna know that data about 8.5m games? People change nicknames all the time and the MMR varies over time... Hard to imagine the MMR data to be any accurate.
The data is from Blizzard ladder, and the MMRs are from the time when the game was played.
|
You don't care about the players. I guess you just treat the mmr of the winner of each game as a sample, regardless of the mmr of the loser:
1st sample: Terran - 2300 - win
2nd sample : Protoss - 2100 - win
....
8000000th sample: Zerg 2000 - win
Or you can only take the games where both players are in a same range of mmr, say 2100-2300, then 2300-2500,...and leave all the games that have a mmr discrepancy out.
I think you could do both. Interesting to see if there is a big difference between the two methods.
|
I was thinking about something like
- mmr difference between players < 200 (can be adjusted to be looser or tighter)
- determine "match-mmr" by taking the mean of the mmr of both players
- present results in "match-mmr"-brackets, e.g. 1800-2000, 1900-2100, 2000-2200,..., 2400++, etc
I'll likely try out a few things to see where we run into problems due to low statistics.
I'm looking for input on of how to group the games/players, feel free to suggest ideas.
I've already handed over a new dataset including 8.5 million games with MMR info included. Be patient until Kraekkling can analyze the new dataset and extract meaningful stats.
I'll probably get to it on the weekend.
|
On October 13 2023 16:25 iloveav wrote:
Based on this data, 6 pool i op zvp.
That's not how I would read it. What we see in the data just means that at this point in the game, it is much more likely for Zerg to win, because those are the games which end because the 6pool was successful. But games where the Protoss manages to defend will be longer and you can't pinpoint a single point in time where we expect Protoss to convert their advantage into a win.
You're likely to instantly win games with a successful rush - but you don't instantly win games where you successfully defended a rush.
We see this in all matchups - the race which has the "quickest rush" will win more games in the very early period of the game.
|
Come to think of it. There are many interesting ways to work with this data.
Kraekkling, can you do something like "Race vulnerability vs mmr"? That is
- you filter out all the games where the mmr difference is 300 or less, meaning you'll only work with games where there is a 300+ mmr diffrence (A)
- you count the games where the loser is the one with higher mmr (B)
- then (B)/(A) is your "vulnerability rate" of the race, which reflects how a supposedly better player can be beaten by a supposedly weaker player by cheese/luck/build order...
Of course you'd do that for each mmr range to see when a race becomes more solid. From instinct we'd probably say Terran is the most vulnerable while Zerg is probably the most resistant from dying by cheese. And I guess at higher ranks when skill improves we'd see all 3 races become less vulnerable. But it'd be nice to see it represented with numbers.
Also I'd expect to see B rank Protosses win the most vs players 300 mmr higher than them lol.
|
Northern Ireland24179 Posts
Solid work there! Will give it a proper read when I’m free but very interesting skim so far
|
|
|
On October 13 2023 17:23 Dakota_Fanning wrote:Show nested quote +On October 13 2023 17:13 sexyMIStrZZZ wrote:On October 13 2023 15:09 Dakota_Fanning wrote:
I've already handed over a new dataset including 8.5 million games with MMR info included. Be patient until Kraekkling can analyze the new dataset and extract meaningful stats. How are you gonna know that data about 8.5m games? People change nicknames all the time and the MMR varies over time... Hard to imagine the MMR data to be any accurate. The data is from Blizzard ladder, and the MMRs are from the time when the game was played.
I hope the data will be reliable and if it will then I'm pretty confident that Protoss at a high level will be performing the worst out of all 3 races.
|
On October 13 2023 03:56 sexyMIStrZZZ wrote:
because Terran becomes so overpowered that it dominates every high level ladder/high level tournament and that is an issue of balance. Terran is so OP that in last 4 ASL tourneys top-8 were
ASL 13 - 3 T, 3 Z, 2 P
ASL 14 - 3 T, 3 Z, 2 P
ASL 15 - 2 T, 5 Z, 1 P
ASL 16 - 3 T, 4 Z, 1 P
Total - 11 T, 15 Z, 6 P
Top-16 were
ASL 13 - 4 T, 7 Z, 5 P
ASL 14 - 6 T, 6 Z, 4 P
ASL 15 - 6 T, 7 Z, 3 P
ASL 16 - 6 T, 6 Z, 3 P
Total - 22 T, 26 Z, 15 P
Protoss suck the most at the very top level, yeah.
But if anything - it's Zerg who's OP according to the ASL top-8 / top-16 statistics. 
Of course, if you're looking only at top-1 or top-2...
|
The highest level is most definitely the worst for protoss, but I'm talking about high level, eg. 2000+ mmr on battlenet ladder, or basically just getting to the ASL is good enough. Protoss has been underperforming there for so many years.
|
|
|
wow, even more data to prove tesagi
|
is awesome32269 Posts
This is super cool, thanks for sharing
|
|
|
One question: how do you get the error bars for each point of data? As I see, it's not like you have several win rates at each minute of the game and take an average of them. You have certain number of wins over a total number of games, and out of that you have 1 win rate at each minute. What do the error bars represent really?
|
On October 16 2023 03:00 Kraekkling wrote:Added MMR data. Also added overall win rate numbers. + Show Spoiler +
Completely expected. I am a bit surprised to see TvZ slightly more favoured to T than ZvP to Z though. One thing is beyond doubt though - P has it by far the hardest at pro level. If the game were updated, this would have been patched pronto.
|
Arty if you're reading this you can always use any of the "valid" counterarguments below:
a) Terrans train harder, Snow and Dewalt don't. Don't ask, they just don't. Just check the ladder, terrans have the most games, it's not like you can just delete an account and create a new one within 20 seconds.
b) It's about the maps, the maps are the reason Terrans win
c) Have you seen how many games do Protosses win at 5 minute mark? That's absolutely disgusting
d) But when Jyj won ASL he played against Protoss only once and he lost that one game. Check-mate Protossists!
e) Have you seen any Terran beat Protoss that is 400 MMR higher than them?
f) Protoss has invisible units? That can't be fair, it's fucking bullshit
g) I believed that P>T for 20 years, so there's nothing that can convince me, you're just wrong
h) Someone messed up the data and its actually the other way around
i) Did you see Snow beat Flash? Well, you just gotta train harder, if he could do that then you can do it too (completely forgetting Snow's performance in the last ASL, that didn't happen at all guys)
Just remember - never talk about facts, at least not when they undoubtebly prove that Terran is overpowered at a high enough level (which here is already above 2000 MMR and probably even slightly below 2k if you check the stats closely).
|
You know what, one thing that has been bugging me is the 60-70% win rate around 2-3 minutes in PvT at all mmr range. Because it makes little sense.
I can understand that win rate in all the Zerg matchups because of 4 pool. But for PvT, if you do proxy gate, the first zealot arrives at Terran's base around 2:30-2:40, so the Terrans just leave the game immediately without a fight? But if that is the case, the win rate must be close to 100% for Protoss. So I guess the 30-40% Terran wins at that minute mark is due to Protoss leaving the game immediately after BBS. So no one bothers to pull probes and SCVs to micro?
|
either scv kills/double scv kills with absurd probe micro and/or rage quits and/or proxy gates that terrans failed to defend properly
u dont have many tools to kill protoss as terran and u do have some tools as protoss, so occasionally protoss will win quick games more often than terran. Despite that Terran has overall >50% wr at 2000+MMR which is pretty crazy to think of if we include those lucky few min wins by Protoss. Basically you need to gamble and risk as much as possible early game in order to have a win from time to time in early games just to compensate for Terran being too strong later on.
|
On October 16 2023 07:07 TMNT wrote:
You know what, one thing that has been bugging me is the 60-70% win rate around 2-3 minutes in PvT at all mmr range. Because it makes little sense.
I can understand that win rate in all the Zerg matchups because of 4 pool. But for PvT, if you do proxy gate, the first zealot arrives at Terran's base around 2:30-2:40, so the Terrans just leave the game immediately without a fight? But if that is the case, the win rate must be close to 100% for Protoss. So I guess the 30-40% Terran wins at that minute mark is due to Protoss leaving the game immediately after BBS. So no one bothers to pull probes and SCVs to micro?
If you went rax expand and Protoss went zealots then yeah it makes sense to leave immediately lol.
|
On October 16 2023 06:48 sexyMIStrZZZ wrote:
Arty if you're reading this you can always use any of the "valid" counterarguments below:
a) Terrans train harder, Snow and Dewalt don't. Don't ask, they just don't. Just check the ladder, terrans have the most games, it's not like you can just delete an account and create a new one within 20 seconds.
b) It's about the maps, the maps are the reason Terrans win
c) Have you seen how many games do Protosses win at 5 minute mark? That's absolutely disgusting
d) But when Jyj won ASL he played against Protoss only once and he lost that one game. Check-mate Protossists!
e) Have you seen any Terran beat Protoss that is 400 MMR higher than them?
f) Protoss has invisible units? That can't be fair, it's fucking bullshit
g) I believed that P>T for 20 years, so there's nothing that can convince me, you're just wrong
h) Someone messed up the data and its actually the other way around
i) Did you see Snow beat Flash? Well, you just gotta train harder, if he could do that then you can do it too (completely forgetting Snow's performance in the last ASL, that didn't happen at all guys)
Just remember - never talk about facts, at least not when they undoubtebly prove that Terran is overpowered at a high enough level (which here is already above 2000 MMR and probably even slightly below 2k if you check the stats closely).
literally noone cares
|
On October 16 2023 07:07 TMNT wrote:
You know what, one thing that has been bugging me is the 60-70% win rate around 2-3 minutes in PvT at all mmr range. Because it makes little sense.
I can understand that win rate in all the Zerg matchups because of 4 pool. But for PvT, if you do proxy gate, the first zealot arrives at Terran's base around 2:30-2:40, so the Terrans just leave the game immediately without a fight? But if that is the case, the win rate must be close to 100% for Protoss. So I guess the 30-40% Terran wins at that minute mark is due to Protoss leaving the game immediately after BBS. So no one bothers to pull probes and SCVs to micro?
You could try and go to repmastered.app and examine this type of games, maybe you'll be able to find some pattern?
Other than that, we're looking at a small numer of games overall.
There also is some baseline probabilty of players just randomly quitting games or disconnecting, which results in a number of games within this period with a win rate of 50% for both races.
This effect could be quantified by looking at the win rate at the 0-1 minute period, where we expect basically no player interaction at all.
This information is not available to us, because games shorter than 2 minute were filtered out in the pre-selection of data. The only useful estimate which I can provide here is that the overall rate of disconnects is somewhere around 2% of all games.
The win rate in PvT around 2-3 minutes is 58% for all players combined. In the MMR-bracket plot all data points except the one with the biggest error bar are close to 60%.
Some speculative thoughts: + Show Spoiler +Scenarios that might make a Terran quit around 2-3 minutes:
- scouting probe enters before depot finishes
- barracks delayed due to scouting probe
- scouting probe kills scv
- gas stolen
- manner pylon
- Terran scouts proxy gate(s)
- zealot enters
Scenarios that might make a Protoss quit around 2-3 minutes:
- something involving 12nex?
- ???
|
On October 16 2023 21:51 Comedy wrote:Show nested quote +On October 16 2023 06:48 sexyMIStrZZZ wrote:
Arty if you're reading this you can always use any of the "valid" counterarguments below:
a) Terrans train harder, Snow and Dewalt don't. Don't ask, they just don't. Just check the ladder, terrans have the most games, it's not like you can just delete an account and create a new one within 20 seconds.
b) It's about the maps, the maps are the reason Terrans win
c) Have you seen how many games do Protosses win at 5 minute mark? That's absolutely disgusting
d) But when Jyj won ASL he played against Protoss only once and he lost that one game. Check-mate Protossists!
e) Have you seen any Terran beat Protoss that is 400 MMR higher than them?
f) Protoss has invisible units? That can't be fair, it's fucking bullshit
g) I believed that P>T for 20 years, so there's nothing that can convince me, you're just wrong
h) Someone messed up the data and its actually the other way around
i) Did you see Snow beat Flash? Well, you just gotta train harder, if he could do that then you can do it too (completely forgetting Snow's performance in the last ASL, that didn't happen at all guys)
Just remember - never talk about facts, at least not when they undoubtebly prove that Terran is overpowered at a high enough level (which here is already above 2000 MMR and probably even slightly below 2k if you check the stats closely). literally noone cares
Wrong, and you should care too.
So now we know for sure:
While some religiously claim for years
T>Z>P>T
We actually have
T>>Z>>P<T - enshrined racial imbalance at pro level. And this is oh so visibly reflected in titles won in the last 20+ years.
|
On October 17 2023 02:17 Rainalcar wrote:
So now we know for sure:
While some religiously claim for years
T>Z>P>T
We actually have
T>>Z>>P<T - enshrined racial imbalance at pro level. And this is oh so visibly reflected in titles won in the last 20+ years.
The T>Z>P>T is still true at most levels below 2000 mmr according to the analysis though. That's where 90% of the players are.
Keep in mind that the T>>Z>>P<T you just mentioned is from analysis of ladder games above 2000 mmr. That's not "pro level" stats, although pros account for a part of high rank ladder surely.
Here's the actual pro level stats from eloboard (since 2021):
ZvP 8448 - 7454 (53.1%)
PvT 8221 - 8518 (49.1%)
TvZ 9114 - 7426 (55.1%)
Sponbbang had the stats from 2016-2021 but sadly the site is dead now. But iirc it's similar to this. The stats probably still lies somewhere in a tl thread though.
So as much as Snow and Best carrying the PvT matchup for Protoss, i.e. no matter how many Bo9s they won against Royal and Rush, they can't skew the stats from 16k games for Protoss to have a positive win rate against Terran.
|
Very nice analysis. Could you maybe also add the cumulative distribution functions of your histograms? This would paint a better picture of how much a time frame with a very one-sided win rate actually matters for the overall win rate.
|
On October 17 2023 02:53 TMNT wrote:Show nested quote +On October 17 2023 02:17 Rainalcar wrote:
So now we know for sure:
While some religiously claim for years
T>Z>P>T
We actually have
T>>Z>>P<T - enshrined racial imbalance at pro level. And this is oh so visibly reflected in titles won in the last 20+ years. The T>Z>P>T is still true at most levels below 2000 mmr according to the analysis though. That's where 90% of the players are. Keep in mind that the T>>Z>>P<T you just mentioned is from analysis of ladder games above 2000 mmr. That's not "pro level" stats, although pros account for a part of high rank ladder surely. Here's the actual pro level stats from eloboard (since 2021): ZvP 8448 - 7454 (53.1%) PvT 8221 - 8518 (49.1%) TvZ 9114 - 7426 (55.1%) Sponbbang had the stats from 2016-2021 but sadly the site is dead now. But iirc it's similar to this. The stats probably still lies somewhere in a tl thread though. So as much as Snow and Best carrying the PvT matchup for Protoss, i.e. no matter how many Bo9s they won against Royal and Rush, they can't skew the stats from 16k games for Protoss to have a positive win rate against Terran.
You are right, it's not all pros. Still, eloboard results are not that far off. P is still visibly weakest, while T has an inherent advantage.
|
damn good stuff, ty for the MMR stats Kraekkling
it gives a good idea of the general skill lvl among the higher MMRs but like i mentioned before, progamers play a different game, and even among them there's huge discrepancies at the highest lvl (Flash in all MUs, Snow PvT etc)
|
On October 17 2023 07:33 TT1 wrote:
damn good stuff, ty for the MMR stats Kraekkling
it gives a good idea of the general skill lvl among the higher MMRs but like i mentioned before, progamers play a different game, and even among them there's huge discrepancies at the highest lvl (Flash in all MUs, Snow PvT etc)
They may play a different game, but the balance stats of their games suggest the exact same thing as 2000+ mmr results in this database, which gives even more evidence that Protoss is the weakest race in the game.
|
On October 17 2023 07:43 sexyMIStrZZZ wrote:Show nested quote +On October 17 2023 07:33 TT1 wrote:
damn good stuff, ty for the MMR stats Kraekkling
it gives a good idea of the general skill lvl among the higher MMRs but like i mentioned before, progamers play a different game, and even among them there's huge discrepancies at the highest lvl (Flash in all MUs, Snow PvT etc) They may play a different game, but the balance stats of their games suggest the exact same thing as 2000+ mmr results in this database, which gives even more evidence that Protoss is the weakest race in the game.
well the trends u see in winrates get amplified even more the higher up u go in skill, so ZvP and TvP keeps getting worse :D
but there's still some outliers where pure skill flips everything on its head, like Snow PvT, it would be nice to see Flash at his best vs Snow
personally if we were to talk about actual balance (where the players are playing the game at the highest lvl humanly possible) we would have to run games between the top 1-2 players of each race and check those stats, everything else is pretty meaningless (even midd/low lvl pro games, even A tier progamers to a certain extent)
anything below that is just a whine fest of ppl just crying about their own experiences, the truth about BW is balance only really matters to the top .1% of players in this game, the skill cap of this game is so big that u can overcome it with skill, the higher up u go that cap becomes tighter and tighter tho
|
On October 17 2023 08:01 TT1 wrote:Show nested quote +On October 17 2023 07:43 sexyMIStrZZZ wrote:On October 17 2023 07:33 TT1 wrote:
damn good stuff, ty for the MMR stats Kraekkling
it gives a good idea of the general skill lvl among the higher MMRs but like i mentioned before, progamers play a different game, and even among them there's huge discrepancies at the highest lvl (Flash in all MUs, Snow PvT etc) They may play a different game, but the balance stats of their games suggest the exact same thing as 2000+ mmr results in this database, which gives even more evidence that Protoss is the weakest race in the game. well the trends u see in winrates get amplified even more the higher up u go in skill, so ZvP and TvP keeps getting worse :D but there's still some outliers where pure skill flips everything on its head, like Snow PvT, it would be nice to see Flash at his best vs Snow personally if we were to talk about actual balance (where the players are playing the game at the highest lvl humanly possible) we would have to run games between the top 1-2 players of each race and check those stats, everything else is pretty meaningless (even midd/low lvl pro games, even A tier progamers to a certain extent) anything below that is just a whine fest of ppl just crying about their own experiences, the truth about BW is balance only really matters to the top .1% of players in this game, the skill cap of this game is so big that u can overcome it with skill, the higher up u go that cap becomes tighter and tighter tho
There's some truth to that, but watching how the Z>P and T>P advantage is highly correlated with mmr one could assume that if we had the absolute best P ever play infinite games against the absolute best T ever, T would win vast majority of the time. Same with T>Z and Z>P.
I think that the correlation of how T's wr goes up with mmr against P means more than the actual winrate itself. And the wr/mmr correlation is positive for T>Z Z>P and T>P.
TBH I think Probes should have 41 hp so that Vultures wouldnt kill them with 2 shots which is absolutely ridiculous and then Reaver should be fixed so that it would always hit the target. Maybe increase damage to bio with storm so that PvZ would get fixed and then weaken the stimpack to fix TvZ.
I also think that low mmr balance stats should be disregarded given that low mmr players could easily improve to become high mmr players while high mmr players do not have that opportunity and are much closer to their skill ceiling and any improvement of their game is minor and doesn't affect their winrates significantly. Low mmr stats should be used solely in order to show the trend of how the wr develops with increase of mmr.
|
On October 17 2023 08:01 TT1 wrote:Show nested quote +On October 17 2023 07:43 sexyMIStrZZZ wrote:On October 17 2023 07:33 TT1 wrote:
damn good stuff, ty for the MMR stats Kraekkling
it gives a good idea of the general skill lvl among the higher MMRs but like i mentioned before, progamers play a different game, and even among them there's huge discrepancies at the highest lvl (Flash in all MUs, Snow PvT etc) They may play a different game, but the balance stats of their games suggest the exact same thing as 2000+ mmr results in this database, which gives even more evidence that Protoss is the weakest race in the game. well the trends u see in winrates get amplified even more the higher up u go in skill, so ZvP and TvP keeps getting worse :D but there's still some outliers where pure skill flips everything on its head, like Snow PvT, it would be nice to see Flash at his best vs Snow personally if we were to talk about actual balance (where the players are playing the game at the highest lvl humanly possible) we would have to run games between the top 1-2 players of each race and check those stats, everything else is pretty meaningless (even midd/low lvl pro games, even A tier progamers to a certain extent) anything below that is just a whine fest of ppl just crying about their own experiences, the truth about BW is balance only really matters to the top .1% of players in this game, the skill cap of this game is so big that u can overcome it with skill, the higher up u go that cap becomes tighter and tighter tho
Flash at his best vs Snow resulted in a 4-0 whoop.
|
Thank you for this awesome contribution. Love to see big data sets used to investigate.
A strategy guide written from the perspective of "when you will die" and what from sounds funny.
"At this point, if you are wrong, 14 zerglings with speed will destroy you - micro very hard before resigning"
"... and if your scan reveals there to be five completed sunken colonies in front of your marine and medic army, press f2 and watch the two turrets in your mineral line fall to 8 mutalisks"
"Congratulations you reached the late-game, you will die in 2 minutes and 15 seconds from guardian harass that will misplace your army allowing a defiler-push on your third or forth expansion, you played a good game"
|
United States12231 Posts
First of all, this is great stuff. Finally a quality post on TL after so many years. The way it's presented is fantastic. Second, I hate that yellow line, it's so incredibly hard to see.
As for the actual data, I see a lot of conclusions being drawn about balance over time at "pro/semipro-level" MMR and frankly I find the sample size too low to say much of anything for that cursed yellow line. The variance as depicted by the I-bars is extremely high, particularly beyond minute 15. Even the 2200+ data I think has too much variance to really be useful in concluding much of anything. I think when you get into the 2000-2200 MMR range that's where you'll find pretty good sample sizes, but then by that point you're having a different discussion about a different cohort. Still, I'd say that's our best lead.
I also see a lot of people trying to explain various parts of the graph (makes sense, people love interpreting data) but I think you can only maybe do that for openers. After say minute 10, the game state is so highly dependent upon the effectiveness of the openers of both players that virtually anything could happen. There are so many potential outcomes resulting from openers that it is impossible to make any meaningful attribution - all we know for sure is X number of games tend to end at Y time with Q race being the victor.
Very cool post.
|
First of all, congratulations, it must have been a lot of work and time. Second i want to say about the apm. i checked my id (StarCast_Sziky ) on the remastered ap and i saw the next : avg 233 eapm. what is not possible. or this is the a difffernt program what i know the last 10 years. + only 165 games? i am not remember correctly but it was much more than that everywhere. ( so I don't understand which games are included in the data and which are not) Third talk about the balance
with the ladder statistic just a joking, and makes no sense. Listen the guys who is 100x better as we us and know more we us. ( Flash,Mini, Jaedong and etc. ).
|
Dakota_Fanning
Hungary2341 Posts
On October 17 2023 17:03 sas.Sziky wrote:
First of all, congratulations, it must have been a lot of work and time. Second i want to say about the apm. i checked my id (StarCast_Sziky ) on the remastered ap and i saw the next : avg 233 eapm. what is not possible. or this is the a difffernt program what i know the last 10 years. + only 165 games? i am not remember correctly but it was much more than that everywhere. ( so I don't understand which games are included in the data and which are not) Third talk about the balance
with the ladder statistic just a joking, and makes no sense. Listen the guys who is 100x better as we us and know more we us. ( Flash,Mini, Jaedong and etc. ).
Obviously repmastered.app doesn't have all ladder replays, "just" about 8.5 million (that are longer than 1:50).
You say it's not possible that you have 233 EAPM, why not? Is it too low or too high? Note this is EAPM (effective APM), not APM.
Also you obviously had multiple accounts, many which we probably don't know about (not public). There are 35 players on repmastered.app that have "sziky" in its name, here's the combined player profile of all those 35:
35 combined players containing "sziky"
These combined players have 4,494 games total, having 327 APM and 211 EAPM. Only 320 games are from Blizzard ladder.
|
Too high, yes. i mean my max eapm around 240. avarage 200++. so the second data it's more real + 320 ladder games very few minimum X10? or something like that:D I just mentioned that on this because there could be problems here again from a statistic point of view, but its okay, and a good job!!!
|
Dakota_Fanning
Hungary2341 Posts
On October 17 2023 20:26 sas.Sziky wrote:
Too high, yes. i mean my max eapm around 240. avarage 200++. so the second data it's more real + 320 ladder games very few minimum X10? or something like that:D I just mentioned that on this because there could be problems here again from a statistic point of view, but its okay, and a good job!!!
Unfortunately I can only get the latest 25 matches from ladder at any given time (although I'm collecting replays for some time now). Older matches are only available if someone uploads them.
|
Thanks Kraekkling for your work! I really appreciate it.
|
On October 16 2023 07:07 TMNT wrote:
You know what, one thing that has been bugging me is the 60-70% win rate around 2-3 minutes in PvT at all mmr range. Because it makes little sense.
I can understand that win rate in all the Zerg matchups because of 4 pool. But for PvT, if you do proxy gate, the first zealot arrives at Terran's base around 2:30-2:40, so the Terrans just leave the game immediately without a fight? But if that is the case, the win rate must be close to 100% for Protoss. So I guess the 30-40% Terran wins at that minute mark is due to Protoss leaving the game immediately after BBS. So no one bothers to pull probes and SCVs to micro?
It only shows the win-rates for the given duration. Longer games probably mostly end in a normal way. When filtering for games that are shorter than what is reasonable, the data is likely dominated by abnormal stuff. It's hard to guess what fractions of the super short games are actual decisive results and what is basically ragequits, from losing a worker, from not wanting to play vs gas steal or proxy or from accidentally cancelling a supply depot.
Also, note that analysing the win-by-duration graphs does not actually show what you might expect it to show. Matchmaking sort-of guarantees that the win-rate over all possible game lengths will be close to 50%, so a strong early timing cannot change the overall balance of the matchup (because matchmaking controls for it), but must instead "move" or "borrow" win percentage from other game durations to the timing window.
An easy way of understanding this is imagining a player who learns a great PvT DT build an suddenly wins a lot more games around 8 minutes. If he still plays exactly the same as before in all his non-DT games, his win rate at other durations will still drop because his MMR adjusts and his opponents are better.
|
On October 17 2023 01:20 Kraekkling wrote:Show nested quote +On October 16 2023 07:07 TMNT wrote:
You know what, one thing that has been bugging me is the 60-70% win rate around 2-3 minutes in PvT at all mmr range. Because it makes little sense.
I can understand that win rate in all the Zerg matchups because of 4 pool. But for PvT, if you do proxy gate, the first zealot arrives at Terran's base around 2:30-2:40, so the Terrans just leave the game immediately without a fight? But if that is the case, the win rate must be close to 100% for Protoss. So I guess the 30-40% Terran wins at that minute mark is due to Protoss leaving the game immediately after BBS. So no one bothers to pull probes and SCVs to micro? You could try and go to repmastered.app and examine this type of games, maybe you'll be able to find some pattern? Other than that, we're looking at a small numer of games overall. There also is some baseline probabilty of players just randomly quitting games or disconnecting, which results in a number of games within this period with a win rate of 50% for both races. This effect could be quantified by looking at the win rate at the 0-1 minute period, where we expect basically no player interaction at all. This information is not available to us, because games shorter than 2 minute were filtered out in the pre-selection of data. The only useful estimate which I can provide here is that the overall rate of disconnects is somewhere around 2% of all games. The win rate in PvT around 2-3 minutes is 58% for all players combined. In the MMR-bracket plot all data points except the one with the biggest error bar are close to 60%. Some speculative thoughts: + Show Spoiler +Scenarios that might make a Terran quit around 2-3 minutes:
- scouting probe enters before depot finishes
- barracks delayed due to scouting probe
- scouting probe kills scv
- gas stolen
- manner pylon
- Terran scouts proxy gate(s)
- zealot enters
Scenarios that might make a Protoss quit around 2-3 minutes:
- something involving 12nex?
- ???
I'm not even joking when I say this, I think there is an effect from protoss players who leave when they don't scout the Terran first. I have definitely went back into replay to see why a protoss left and it was when they reach first empty base with scouting probe. They are sick people LOL
|
Physician
United States4146 Posts
physician aka entropy very humbly bows down in silent respect and then discreetly withdraws again into fog..
|
As this work looks good and meticulous I would like to point at just one small detail. There is no such a thing as a 7:00 - 8:00 1 minute interval. It can be 7:00 - 7:59 or 7:01 8:00. I believe that does not change the data in any way, but can make you count twice all the games a good amount of games, precisely 1 every 30.
|
Dakota_Fanning
Hungary2341 Posts
On October 20 2023 06:50 ajmbek wrote:
As this work looks good and meticulous I would like to point at just one small detail. There is no such a thing as a 7:00 - 8:00 1 minute interval. It can be 7:00 - 7:59 or 7:01 8:00. I believe that does not change the data in any way, but can make you count twice all the games a good amount of games, precisely 1 every 30.
I'm sure the 7-8 minute phrase is just for easy comprehension. I'm sure the upper bound is exclusive and no game is counted twice in multiple periods. It's much easier to read and write "7-8 minute" than "7:00-7:59".
|
moktira
Ireland1542 Posts
Have you thought about writing this up for an academic publication?
|
thank you! great!
I see that the majority of the games are new, but what was wrong with reps before 2018? do you consider the dataset too small before that?
|
wow this is amazing and fascinating. Thank you for your work!
|
First, thanks to everyone who found this whole thing interesting.
If you really enjoyed the insights, consider donating a small amount to repmastered, which made this possible. Many more working hours were needed for that project and there are recurring costs to host all those replays...
On October 17 2023 03:08 Cryoc wrote:
Very nice analysis. Could you maybe also add the cumulative distribution functions of your histograms? This would paint a better picture of how much a time frame with a very one-sided win rate actually matters for the overall win rate.
I thought about this but couldn't come up with a non-cluttered way to visualize this. In particular for the plots with multiple selections, you'd need cumulative distributions for each of them. Also when looking at mutlipe ones on the same plot they'd need to be normalized. And, one might need to adjust some of the selection criteria to deal with low statistics in the samples. That's where I mostly stopped thinking about it. If you have something specific in mind, I might look into it.
On October 22 2023 22:43 zimp wrote:
I see that the majority of the games are new, but what was wrong with reps before 2018? do you consider the dataset too small before that?
No particular reason for that, the cutoff is arbitrary. But yes, we're losing a negligible amount of data.
|
|
|
On October 22 2023 22:43 zimp wrote:thank you! great! I see that the majority of the games are new, but what was wrong with reps before 2018? do you consider the dataset too small before that?
I wouldn't say it's a small data set even before 2018 (just comparatively small).
I think the newer the data the better. As with time more things have been figured out.
|
somewhat related question
is it possible to search the repmastered database for matching hotkeys between different playernames?
e.g. analyze hotkey usage from old FlaSh replays and then cross-check recent/new replays to see if he is active again...
|
This is amazing work
I would love to see a comparison of race win percentages vs mmr which could then be correlated to the timings to see what players potentially need to work on in mmr brackets for a particular race. Ex) +1 timings seem effective in tvp up until this MMR range at which point terran need to incorporate new strategies
As an aside, is it possible to replace the yellow with another color? It is challenging to track visually.
|
On October 11 2023 12:21 TT1 wrote:
Is it possible to do this with progamer replays or 2500+ ladder games from cwal (the ladder games would prob be way better cus of sample size)? I know the sample size would be way less but the quality of games is way more important for this type of analysis. Thanks for this tho, great work.
There's already one with all the ASL/KSL games featured in a recent Tasteless video
|
|
|
|
|
|
EPT![[image loading]](https://i.ibb.co/fFk9F7T/mapshist.png)
![[image loading]](https://i.ibb.co/4p1MNH4/effapm-allraces.png)
![[image loading]](https://i.imgur.com/ZloyhHr.png)
![[image loading]](https://i.imgur.com/sBZKpJB.png)
