EPTArticle is here: https://medium.com/@evangball1990/starcraft-brood-war-a-bgh-analysis-part-1-9ff9e1ac296c
Enjoy!
BGH - An Analysis. With Data!
| Forum Index > Brood War Strategy |
|
sgKingdom
4 Posts
Article is here: https://medium.com/@evangball1990/starcraft-brood-war-a-bgh-analysis-part-1-9ff9e1ac296c Enjoy! | ||
|
WGT-Baal
France3516 Posts
 Amazing first TL post. Congrats! | ||
|
fgt4w
24 Posts
| ||
|
CruiseR
Poland4014 Posts
Small drawback: Some of the data [each matchup split] had only 20-40 games, which is a low statistical significance. Interesting fragments: win% over time- P stable, and Z still very high (50%+) even in longest games. (I thought Z is a little weaker lategame, as it's hardest race to carry, even if you become big). Nice post nevertheless, read it all. | ||
|
needsmoarsalt
2 Posts
On January 25 2025 02:39 CruiseR wrote: Yeah, nice post, but for someone who plays team games most of it was expected. Similar is with ZX team in 2v2, always has an edge over no Z team. Small drawback: Some of the data [each matchup split] had only 20-40 games, which is a low statistical significance. Interesting fragments: win% over time- P stable, and Z still very high (50%+) even in longest games. (I thought Z is a little weaker lategame, as it's hardest race to carry, even if you become big). Nice post nevertheless, read it all. Thanks for the kind words! Yeah a lot of the insights won't be too surprising to anyone whose played a lot of BGH. Personally I was a bit surprised that Terran never exceeds a 50% win rate even in longer games, given how strong their mid/late game is. The low sample size is definitely an issue for some matchups. Most lobbies will have something like 2 players picking P and one Z, and the rest Random, which makes certain matchups a lot less likely to be played than others. Regarding the high late game Z win rate, i think a lot of those are games where the Zerg team got a game winning advantage early on, but weren't able to actually end the game until much later. For example they eliminated one enemy 5 minutes into the game, but the other 2 were able to turtle shared at 11-12 for like 20 minutes before finally getting overwhelmed. | ||
|
Kraekkling
734 Posts
Btw, you might want to include one or two of your plots in the OP here on TL to make the people more likely to click on the blog link | ||
|
iopq
United States1148 Posts
| ||
|
mutantmagnet
United States3807 Posts
I hope players who like this mode see this article. | ||
|
sgKingdom
4 Posts
On January 30 2025 19:45 iopq wrote: Fan of the article, not a fan of medium, though This sent me down quite a rabbit hole lol. I mostly picked Medium because I couldn't make TL's image service work and Medium's is very easy. Didn't know anything about Medium's business model. Glad you enjoyed the article! | ||
Chosi
Germany1309 Posts
| ||
|
Dakota_Fanning
Hungary2368 Posts
For a larger dataset, you may consider all BGH MMR EU games: https://repmastered.icza.net/stats?ladder=bghmmreu So far more than 28k BGH MMR EU games. Interestingly, win ratio of races there is quite balanced: P: 51%, Z: 50% and T: 48% (note that this only contains 3v3 games). For the "absolute" dataset, consider all BGH games: https://repmastered.icza.net/map/Big Game Hunters So far almost half a million games, but this includes all formats. Win ratio is again quite balanced: P: 50%, Z: 49% and T: 47%. | ||
2Pacalypse-
Croatia9547 Posts
As someone who has been playing BGH for over 20 years now, during the "golden" age of BGH (mid to late 00s) there was only one matchup that was viable at the highest levels: PPP. Which is why pretty much all BGH tournaments that I've organized or played in had PPP v PPP finals. Now don't get me wrong, PPP can still lose to certain other matchups (PPZ, ZZT, PTZ), but those matchups have to rely on luck much more (spawn positions, or killing one player early). What I suspect went wrong with your analysis is twofold: - The dataset you're using is probably lacking most of the top BGH players and games of the old - Most of the top BGH games used to be played in private lobbies, and as such they're probably not uploaded to repmastered (I've uploaded my share tho :d) Additionally, one factor that you couldn't account for by just looking at the replays is how many of those races were chosen in the lobby and how many were the result of choosing Random. This can have a huge effect in the outcome of games. I think the data you analyzed paints a pretty accurate picture of the current situation in the BGH scene and especially in the "public" games. But (un)fortunately, there's a clear difference between most of the BGH games played today (even at the highest levels in the BGH MMR ladder) and privately organized games between top BGH players that were played ~15 years ago. I actually prefer the games these days somewhat haha, as they're more varied; PPP v PPP gets pretty boring pretty fast :d | ||
|
Alexander34
8 Posts
On February 11 2025 00:04 2Pacalypse- wrote: This is a pretty good, data-driven article, but unfortunately it comes to the wrong conclusion :p As someone who has been playing BGH for over 20 years now, during the "golden" age of BGH (mid to late 00s) there was only one matchup that was viable at the highest levels: PPP. Which is why pretty much all BGH tournaments that I've organized or played in had PPP v PPP finals. Now don't get me wrong, PPP can still lose to certain other matchups (PPZ, ZZT, PTZ), but those matchups have to rely on luck much more (spawn positions, or killing one player early). What I suspect went wrong with your analysis is twofold: - The dataset you're using is probably lacking most of the top BGH players and games of the old - Most of the top BGH games used to be played in private lobbies, and as such they're probably not uploaded to repmastered (I've uploaded my share tho :d) Additionally, one factor that you couldn't account for by just looking at the replays is how many of those races were chosen in the lobby and how many were the result of choosing Random. This can have a huge effect in the outcome of games. I think the data you analyzed paints a pretty accurate picture of the current situation in the BGH scene and especially in the "public" games. But (un)fortunately, there's a clear difference between most of the BGH games played today (even at the highest levels in the BGH MMR ladder) and privately organized games between top BGH players that were played ~15 years ago. I actually prefer the games these days somewhat haha, as they're more varied; PPP v PPP gets pretty boring pretty fast :d I think best is PPP then PPT , PTZ , PZZ, , TZZ, ZZZ, TTT | ||
|
sgKingdom
4 Posts
On February 11 2025 00:04 2Pacalypse- wrote: Additionally, one factor that you couldn't account for by just looking at the replays is how many of those races were chosen in the lobby and how many were the result of choosing Random. This can have a huge effect in the outcome of games. Most definitely. In particular a big advantage teams with multiple zergs often have is randoming into them and only getting scouted after its too late for the other team to respond optimally. Unfortunately random picks aren't available in this dataset, so I can't analyze that. On February 11 2025 00:04 2Pacalypse- wrote: As someone who has been playing BGH for over 20 years now, during the "golden" age of BGH (mid to late 00s) there was only one matchup that was viable at the highest levels: PPP. Which is why pretty much all BGH tournaments that I've organized or played in had PPP v PPP finals. Now don't get me wrong, PPP can still lose to certain other matchups (PPZ, ZZT, PTZ), but those matchups have to rely on luck much more (spawn positions, or killing one player early). Don't think I can agree with this one. PPP might have crushed everything 20 years ago, but the game has definitely evolved since then. The data (and my own personal experience) strongly suggests that PPP is very strong in specific matchups (vs PPZ for example), but has some very serious weaknesses (PTZ and PPT in particular). It's definitely true though that the PPP data is a bit thin, and these numbers could change with a larger sample size. In general, i think PPP will struggle against any Terran comp where killing/walling the terran early is difficult for most spawns, which i think explains the low win rates vs PTZ and PPT. | ||
|
WombaT
Northern Ireland27266 Posts
| ||
|
ted.
United States85 Posts
Games used to feel much more aggressive than they do now a days, and top protoss players just dominated it. Would love to see the top guys like koolam and merf etc from back then playing protoss vs the top players today. Honestly no idea how that would play out. | ||
|
ted.
United States85 Posts
I think today it feels like players have adapted to slowing down just a tad and using other strategies. I’m still uncertain whether this is a stronger and updated meta, or if the “all in units at all times” is still better if played by highly skilled players. Probably the later given it feels top hunters players have always been incredibly aggressive. | ||
Peeano
Netherlands5701 Posts
I also think PPP is the most OP with PPZ being the next best thing. If your small sample size data is correct on Zerg winning the most, then I will assume that is partially due to a large number of people not knowing how to hold early pool builds, nor knowing how to punish Zerg pumping drones, and that P only "seems" to have the most win percentage because winning within the first 5 minutes (majority of Zerg wins) are just not as memorable in general. Also playing uphill match ups with bad spawns in BGH is the worst, especially if you're the best player in your team. And in my experience especially when being Zerg. Zerg just can't trade well in small corridors and setting up defense in small corridors as Zerg both takes times and kills your economy. Whenever I'm Zerg on BGH I wanna win asap or otherwise gain big advantage early on, which is either stomping one player or setting up effective containment that requires a P or T ally (with a brain), that then allows you to economy boom or get immediate game winning tech, i.e. lurker or muta. If your data is right, I would also expect Zerg mains to be high up on the BGH MMR ranking list. Or either Zerg mains to actually play only or a lot of Zerg. Neither is the case as far as I can see. Also the trend there (without checking a whole lot of profiles) seems to be Protoss getting the best winrate among random or race picking players. | ||
|
Peeano
Netherlands5701 Posts
On February 26 2025 13:04 ted. wrote: I feel like the bgh games back then was just all in units all the time, which might explain the total Protoss dominance. I think today it feels like players have adapted to slowing down just a tad and using other strategies. I’m still uncertain whether this is a stronger and updated meta, or if the “all in units at all times” is still better if played by highly skilled players. Probably the later given it feels top hunters players have always been incredibly aggressive. Aggro on really short rush distances creates opportunity to effectively kill one early or either set up containment to get a lead that way, so it makes sense. On the other hand turtling is an investment that only pays off if you get to utilize your timing window. This option is much more demanding as it requires way more multitasking as soon as you break out, IF you or both your allies didn't already die before you even get to that stage. | ||
|
sgKingdom
4 Posts
On February 26 2025 23:28 Peeano wrote: If your data is right, I would also expect Zerg mains to be high up on the BGH MMR ranking list. Or either Zerg mains to actually play only or a lot of Zerg. Neither is the case as far as I can see. Also the trend there (without checking a whole lot of profiles) seems to be Protoss getting the best winrate among random or race picking players. I would actually expect the opposite. Zerg's strength is its early game mobility, which allows it to reliable force 1v2 or 1v3 fights assuming their teammates coordinate well with them. The data presented in the article is filtered to include only games with 6 decent players, so that coordination will usually happen. Most MMR bot games have at most 4 decent players, and usually at least 1 very weak player. Once you get high up in mmr bot, you're always going to get teamed with the weakest player in the lobby, which means you need to be able to hard carry 1 or 2 noobs. Zerg is not good for that, since their economy is so weak early on. This is why you see mostly Protoss/Terran at the top of the mmr bot rankings. | ||
| ||
