|
On February 28 2012 01:58 ZeroTalent wrote:Show nested quote +On February 28 2012 00:17 Felnarion wrote:
I'd like to see PvZ's graph extended to include 30-35-40-45 minute games so we could see the continuationof the graph. Though I understand that the number of games in those windows is likely very small. Too lazy to post a graph, but here are the raw numbers 30-35 minutes: P 85 : 67 Z 35-40 minutes: P 21 : 28 Z 40-45 minutes: P 14 : 9 Z 45-50 minutes: P 5 : 7 Z 50-55 minutes: P 0 : 2 Z Keep in mind that the sample size even at 30 minutes is small enough that it's hard to draw super-fine conclusions.
Thanks for that. Yep, it appears as I would expect it to. I think the 30-35 minutes is the last of the brood-lord switch timings and from there it swings back and forth. As you said, the sample is small, but it swings back and forth seemingly arbitrarily after the point I'm talking about, which makes me think that all matchups are essentially balanced beyond the 30-35 minute mark, and from there, the better player seems to win more often than not.
|
On February 27 2012 20:38 Roblin wrote:
but then why would the PvZ line be red? it would make a lot more sense for it to be green if it indicates... whatever, this doesn't seem to make very perfect sense anyway, so lets just ignore the color of the line and assume that the line indicates the winrate of the first mentioned race, that would mean:
P<Z
Z<T
T>P
Sorry for the confusion; you have it right. Yes, the win rate is from the perspective of the first mentioned race. The dashed line is supposed to be a blend, but the blend of green and red is ... brown, and it's hard to make a brown that doesn't look sort of reddish :|.
Terran is favored in both TvP and TvZ; Zerg is favored in ZvP. Yet Terran is by far the least-played race, at least by total games played, and Protoss ... the weakest race by pure win rates ... is played almost as much as Zerg.
|
I'm quite surprised that TvTs aren't generally longer. I always think of them as slow, super-careful games of chess, especially when comparing them to the razorblade "one wrong move and you lose" PvP and ZvZ match-ups. I stereotyped TvT as boring 40 minute games lol.
I like reading data like this  Always useful to have. Thanks!
|
On February 28 2012 02:45 DarkPlasmaBall wrote:I'm quite surprised that TvTs aren't generally longer. I always think of them as slow, super-careful games of chess, especially when comparing them to the razorblade "one wrong move and you lose" PvP and ZvZ match-ups. I stereotyped TvT as boring 40 minute games lol. I like reading data like this Always useful to have. Thanks!
There are some very scary 1-base attacks. You can lose all your workers (and much of your army) to something like Reaper-Hellion or cloaked banshee if you're not prepared. Someone can go for a fast tank build and contain you. Etc. TvT has the 3rd highest quit rate in the 5-10 minute window (behind PvP and ZvZ), and I'm guessing that's why.
|
TvP win-rate ratio over time is brutal...
Graph clearly confirms the problem I've been experiencing / Terran players have been talking about current late-game TvP.
I was not expecting such discrepancy before late-game, though. Looks to me that Terran are trying to kill the Toss before late-game, with probably a 2 or 3 bases all-in (since we get medivacs at 10 min+). And as natural response to this trend, the Protoss goes for the kill with a 4-gate all-in (?) before the 10 min mark.
Good graphs, thanks for putting that together and sharing.
Edit : it would be very interesting if you could merge the game lenght datas per matchup with the Win-rate per game length datas (first graph with each respective matchup graph). That way we could get a more precise idea of who wins and when.
Don't know if I made myself clear... just superpose the TvP line of the first graph with the last graph, for TvP.
Edit 2 : also, I don't know how to read the "TvP overall" line : what does it mean?
|
What I don't understand is how Protoss complain about losing after doing a 2 base ALL IN
|
Great post. It will be interesting to see what happens to lategame TvZ because the Snipe nerf.
The TvP graph is interesting to look at. I was surprised to see that P had an advantage during the 0-10 minute range. I was also surprised to see lategame TvP balance out after the 35 min mark. I wonder how big those sample size segments are.
|
It would be nice if TvP were more balanced like ZvP.
As a Protoss player I would gladly take a late game nerf if Terran's early-mid game wasn't so strong. After all, you've gotta survive the early-mid game first if before the odds are in your favor.
|
On March 04 2012 05:01 Ghostface_Killa wrote:
It would be nice if TvP were more balanced like ZvP.
As a Protoss player I would gladly take a late game nerf if Terran's early-mid game wasn't so strong. After all, you've gotta survive the early-mid game first if before the odds are in your favor.
IMO it's not the late game units protoss has, it's the ability to warp in 40 zealots immediatley after an even trade of armies. Thats the problem I always seem to have.
|
On March 04 2012 05:01 Ghostface_Killa wrote:
It would be nice if TvP were more balanced like ZvP.
As a Protoss player I would gladly take a late game nerf if Terran's early-mid game wasn't so strong. After all, you've gotta survive the early-mid game first if before the odds are in your favor.
Ever held a 6gate? Ever held a 3gate robo? Ever held blinkstalker rush? Yes, nerfing terran early game is very needed. That way terran CAN'T hold those protoss allins. Even a 2 base collosus push is VERY hard to hold.
It's such a nice myth from the beta that terran's early-mid game is too strong. The only "strong" build atm is the 1-1-1, but I saw tosses roll it so easily that I wonder if it's still that good.
If you 1 rax FE and toss 1 gate FE, then I don't see how terran can be stronger then protoss @ midgame. What can he do?
- 4-5 rax marine push?
- Ghost MM push?
- Try some lucky drops?
>> I don't see how this is "so strong".
On the other hand, if we look at lategame TvP, then we can see that the protoss deathball is VERY strong.
|
On March 04 2012 05:20 Snowbear wrote:Show nested quote +On March 04 2012 05:01 Ghostface_Killa wrote:
It would be nice if TvP were more balanced like ZvP.
As a Protoss player I would gladly take a late game nerf if Terran's early-mid game wasn't so strong. After all, you've gotta survive the early-mid game first if before the odds are in your favor. Ever held a 6gate? Ever held a 3gate robo? Ever held blinkstalker rush? Yes, nerfing terran early game is very needed. That way terran CAN'T hold those protoss allins. Even a 2 base collosus push is VERY hard to hold. It's such a nice myth from the beta that terran's early-mid game is too strong. The only "strong" build atm is the 1-1-1, but I saw tosses roll it so easily that I wonder if it's still that good. If you 1 rax FE and toss 1 gate FE, then I don't see how terran can be stronger then protoss @ midgame. What can he do? - 4-5 rax marine push? - Ghost MM push? - Try some lucky drops? >> I don't see how this is "so strong". On the other hand, if we look at lategame TvP, then we can see that the protoss deathball is VERY strong.
I don't think anyone's arguing terran is "too strong" pre-stim and pre-medivac. It's timings where medivacs and stim are out, but not storm/colossi where terran is very strong. Personally I think that the shortness of terran's tech tree is at fault. The 1-1-1 literally unlocks 85% of terran units by about the 8 minute mark. As a result they can't be as powerful as their protoss counterparts, resulting in being very strong when terran has their tech lead, but weak once the other races have caught up. IMO the fusion core should become more like a science facility, being needed for ravens (with an associated buff), battlecruisers, and potential HotS units.
On an unrelated note, it's also kind of interesting that the game length looks like a maxwell-boltzmann distribution.
|
On March 05 2012 16:48 dainbramage wrote:
\On an unrelated note, it's also kind of interesting that the game length looks like a maxwell-boltzmann distribution.
It also looks sort of like a Gamma distribution (EDIT: haha, it turns out the Maxwell-boltzmann distribution is a special form of the Gamma distribution). This is perhaps unsurprising, as lots of things humans do end up in Gamma distributions. But we don't have enough data points or observations to make really robust conclusions about the distribution function. Even if I bucketed in 1-minute increments, we have only 4500 or so ZvZ games, so it's hard to draw much of a conclusion.
|
|
|
So Protoss is stronger at the early and late game while Terran is only good at the mid game?And people still wonder why terran representation lessens every passing season.
|
Racial win rate by game length charts almost make the game look balanced.
|
What is really interesting is that in each matchup a different race has an advance in the early game. That makes it hard for blizzard to just fix something.
|
The paradox is that North America Playhem players are not good enough to show the truth about balance, but the GSL sample size is too small...
Only thing which is canonic about all balance graphs is the early winrate in TvP and ZvP where T and Z both have some nice all-ins.
|
Very nice , but just cause a game ended at "x" time doesn't mean it was lost there
In the end overall win rate with subjective strategic analysis must remain the first criteria
For instance, the same medivac timing contributing to a win rate bump in one period may be responsible for losses in the next five minute period
|
United Kingdom1381 Posts
Wow, that graph really illustrates how volatile PvP is.
|
Is it difficult to secure a collection of pro korean replays to analyze? I, preconceivably, think that pure NA data would differ significantly from EU and Kor data.
Would be nice, but I understand how difficulty in securing thsoe replays.
|
|
|
|
|
|
EPT
