|
Modern 4-player maps tend to be quite similar, with differences often found in the design of the center and the location of the 3rd base (and/or 4th base, if there is one).
It’s arguable that these differences often have a minor influence on the outcome of the TvZ matchup. This is largely because, in simple terms, the Zerg game plan revolves around taking another main base and its natural expansion, while the Terran strategy generally involves preventing this.
However, win rates for the matchup can differ significantly depending on the map.
I believe it all comes down mostly to a single map feature. There will be other factors as well but none are as clear and simply identifiable as this one. Depending on this one feature, the map will either be:
- slightly Terran-favored (~52% TvZ win rate)
- clearly Terran-favored (~57% TvZ win rate)
For example, let's look at Retro + Show Spoiler + and FS + Show Spoiler +.
The overall resemblance is obvious since Retro is a remake of FS.
It all comes down to this: + Show Spoiler +.
That's it. It's true (almost) every single time. Just pick whatever 4p map, take a look at the ZvT win rate, and then go look at a picture of the map.
It seems that Zerg needs a full arc of highground behind the natural expansion minerals (to apply sufficient pressure with mutalisks) to have a fair chance in the game.
Last, I'd like to point out that there are almost no maps that favor Zerg in the TvZ matchup. Overall, 9 out of 10 (!) maps are slightly or heavily Terran-favored. If we can conclude that the argument is true or mostly true, it could help us design more balanced maps in the future.
|
Interesting but i thought FS was clearly T favoured in "recent" times?
|
On August 28 2024 22:22 WGT-Baal wrote:
Interesting but i thought FS was clearly T favoured in "recent" times?
Well, if true, this would make my point weaker. Unfortunately I don't know any way to filter recent korean progamer games by date, so not sure how to check for this. Suggestions very welcome.
There is this thread from 2015, which examines the fact that FS seemed to be pretty balanced during the kespa era but then turned to be very terran favoured post-kespa. I think that one should be careful when using old data, because we saw several changes to the meta game and, since then, acquired a lot of knowledge regarding what is the optimal way to play.
To draw a parallel to chess, you wouldn't use games from, say, pre-2000 to evaluate whether some opening is viable at the highest level of play. + Show Spoiler +actually you wouldn't use any games from before ~2018 at all, because back then we got better chess engines due to the use of neural networks which gained over 300 elo points and invalidated a lot of previous "knowledge"...
At eloboard, which samples daily proleague, sponsored games, etc. since like ~2019 (not sure about this though, maybe someone knows?), FS stands at
- ZvT: 385 wins, 414 losses (48.2%)
so you'd need to argue that, for example, it was at a perfect ZvT: 300 wins, 300 losses (50%), and then went to a ZvT: 85 wins, 114 losses (42%), which to me seems unlikely and I don't remember FS ever being considered as perfectly balanced at any point.
|
United States10253 Posts
On August 28 2024 23:31 Kraekkling wrote:Show nested quote +On August 28 2024 22:22 WGT-Baal wrote:
Interesting but i thought FS was clearly T favoured in "recent" times? Well, if true, this would make my point weaker. Unfortunately I don't know any way to filter recent korean progamer games by date, so not sure how to check for this. Suggestions very welcome. There is this thread from 2015, which examines the fact that FS seemed to be pretty balanced during the kespa era but then turned to be very terran favoured post-kespa. I think that one should be careful when using old data, because we saw several changes to the meta game and, since then, acquired a lot of knowledge regarding what is the optimal way to play. To draw a parallel to chess, you wouldn't use games from, say, pre-2000 to evaluate whether some opening is viable at the highest level of play. + Show Spoiler +actually you wouldn't use any games from before ~2018 at all, because back then we got better chess engines due to the use of neural networks which gained over 300 elo points and invalidated a lot of previous "knowledge"... At eloboard, which samples daily proleague, sponsored games, etc. since like ~2019 (not sure about this though, maybe someone knows?), FS stands at - ZvT: 385 wins, 414 losses (48.2%)
so you'd need to argue that, for example, it was at a perfect ZvT: 300 wins, 300 losses (50%), and then went to a ZvT: 85 wins, 114 losses (42%), which to me seems unlikely and I don't remember FS ever being considered as perfectly balanced at any point.
Hey that's my old article!
To address your concerns Kraekkling, I would look at all the other popular 4p maps and examine their features. Conceptually, I think your sentiment regarding map features is what a lot of people would assume to be the same: more room behind the nat minerals to make muta harass more effective. Another one I'd add would be other third base locations that only have 1 ramp leading to the base to minimize the amount of lurkers needed to hold the base. Examples are: Radeon (12/6), Allegro (12/3/6/9 but slightly wider ramp), Vermeer (12/3/6/9). Obviously, we also have other maps like Apocalypse and Sylphid that also have exposed natural mineral lines and alternative 1-choke thirds, but those are are 3p maps so maybe unfair to compare them to 4p maps. That being said, here are some results:
Vermeer: ZT: 1124 wins 1588 losses (41.4%) - Exposed natural and 1-choke 3rd
Radeon: ZT: 356 wins, 359 losses (49.8%) - Exposed natural and 1-choke 3rd
Allegro: ZT: 374 wins, 396 losses (48.6%) - No exposed natural and 1-choke 3rd
Citadel: ZT: 220 wins, 342 losses (39.1%) - No exposed natural and 1-choke 3rd
Revolver: ZT: 224 wins, 265 losses (45.8%) - Exposed natural and 2-choke 3rds
Tempest: ZT: 129 wins, 144 losses (47.3%) - No exposed natural and 2-choke 3rds (actually one of the worst naturals since only the gas is exposed).
I don't exactly see a trend here between the features and which encourages a better ZvT winrate. I would expect Vermeer to do reasonably well, but both it and Citadel are dreadful maps against Terran. One has the more exposed mineral line, one doesn't, and both have 1-choke 3rds for alternative base placements.
On top of that, I even expected that with more unpathable terrain and high ground areas on both Vermeer and Citadel that picking off marines as they push across the map would be beneficial to their winrate, compared to Radeon with a very wide open center that provides no cover for mutas to harass the bio ball. But that isn't the case either. Allegro doesn't have as many ridges/cover for mutas and again, that map does fine. Same with Revolver. Polypoid also has a solid amount of ridges but doesn't do as well. Retro's middle is quite open but the side paths are very constricted making it much easier for muta control against bio balls trying to take those side paths (tho pros rarely do take those paths in the mid-game).
Is it perhaps that what conventionally may be considered good features for Zerg might actually lead to Zergs getting over aggressive and taking too much damage trying to abuse the map features? Hence, on a weaker map like Allegro with no exposed natural mineral line, Zergs play more disciplined rather than trying to harass Terran worker lines and end up losing mutas during that period. Idk the answer obviously, but it's something to consider.
The alternative is obviously, just make more 2p and 3p maps which seems to make the matchup much better for Zerg while still having good winrates in the other matchups (Apoc, NDO, Eclipse, Butter, even Monopoly has a solid spread for the matchups). But 3p maps are difficult to map, and 2p maps encourage more variance with cheese and Terrans will cry about gas steals.
|
I completely agree that other features can affect the matchup. Citadel, with its non-standard ramp, is a great example.
Vermeer: ZT: 1124 wins, 1588 losses (41.4%) - Exposed natural and 1-choke 3rd
Radeon: ZT: 356 wins, 359 losses (49.8%) - Exposed natural and 1-choke 3rd
Allegro: ZT: 374 wins, 396 losses (48.6%) - No exposed natural and 1-choke 3rd
Citadel: ZT: 220 wins, 342 losses (39.1%) - No exposed natural and 1-choke 3rd
Revolver: ZT: 224 wins, 265 losses (45.8%) - Exposed natural and 2-choke 3rds
Tempest: ZT: 129 wins, 144 losses (47.3%) - No exposed natural and 2-choke 3rds
This is a useful list where the trend doesn’t always seem to hold.
Hopefully, this isn't too hand-wavy, but we need to consider statistical error and the significance of these numbers. For instance, even if something has a 10% probability, in 10 cases, you'd expect the unlikely outcome to occur once. This issue is more severe with fewer games and higher statistical uncertainty. Thus, we might see both under-performance and over-performance, even if the underlying "true" distribution is correct.
This is how I’d view the numbers for Allegro, Revolver, and Tempest. I think they are overall within what we'd expect when looking at several maps.
The big question for me is Vermeer. Despite the large number of games and the highly singificant result, Zerg seems to underperform severely.
|
United States10253 Posts
Yes, by all accounts, Vermeer seems to me to have a lot of features that would favor Zerg but performs extremely poorly. Perhaps the closer/easier third for Terran (3rd placed in front of their natural) is a very good base for them to secure and thus makes their mid-late game very good? But Radeon has a similar idea with the standard 3rd base position as well...
|
Tempest is Zerg favored because sunkens are on the high ground! Its defensive nature is why I would expect it to be Zerg favored. It's not true that 3rd are two choke because people take a natural of another main as a third on Tempest. It's an uphill one choke third, and you can Nydus to connect it and then take a back 4th
|
Radeon probably just has a longer rush distance, I noticed stuff coming later by a few seconds to my base. Also someone should test whether the closer air distance spawns (horizontal spawns) are favoring Zerg or not.
|
This seems right about both Tempest and Radeon. I think Tempest is already sufficiently non-standard, so the thesis might not be fully applicable.
Anyways, to repeat the initial claim: it seems to me like for maps which are composed of standard elements the difference between an exposed natural and no exposed natural results in roughly 5% win rate difference. Lets say, we take Radeon and only make changes to the natural so it is no longer exposed - will this result in a 45% ZvT win rate, compared to the current 50%?
|
Vermeer still needs an explanation though. Whatever it is, it also favours Terran in TvP (56% win rate).
Is it the general tightness? Too many easily accessable expansions for Terran to take?
Zerg brothers, why did you struggle on Vermeer?
|
United States10253 Posts
On August 30 2024 19:13 Kraekkling wrote:
Vermeer still needs an explanation though. Whatever it is, it also favours Terran in TvP (56% win rate).
Is it the general tightness? Too many easily accessable expansions for Terran to take?
Zerg brothers, why did you struggle on Vermeer?
So I asked Hawk on discord about this. He says it comes down to a few other things:
1) No good overlord spots to see which way the bio ball is moving out. Compared to Radeon, which has 4 really good spots to have overlords in position to see drops and the bio ball's direction in the middle of the map, Vermeer doesn't have a good central overlord spot just outside the natural to see the bio ball direction. There's only really the spot over the natural, and the spot over the water at the pocket base.
2) Pocket bases favor Terran, since they don't need to spread their units as much to defend, they simply move their rally forward a bit. Compared to Radeon where the 3rd is a bit farther away. Plus, the pocket base is also a gas base. He said if it were a mineral only, it would not help quite as much (Allegro as an example).
3) Even if a side base (12/3/6/9) are one choke bases, them being on low ground instead of high ground makes a huge difference when defending. High ground with small ramps would always be better than any other choke point. So my original idea that those bases should help Zerg maybe isn't as fruitful as I once thought.
So ideas for map makers should be: adding more overlord vision spots to help them spot the Terran army. These spots don't really impact the other matchups, since PvZ corsairs can kill those overlords and they'll be forced to retreat, and those spots aren't really used for observers in PvT since, well, they're cloaked anyways. That's probably the one big change that wouldn't impact the other matchups.
Pocket bases being mineral only, however, would definitely impact the other matchups. Zergs want to get that 3rd gas, so they likely will skip the pocket base in favor of a farther away 3rd gas base instead. This might make reinforcing against early zealot attacks a bit harder for Zerg in ZvP, but I don't think that matchup needs to be favoring Zerg anyways. In PvT, Terran obviously wants as much gas as possible, so it's a toss up between playing riskier and taking the 3rd gas or playing more solid and taking a mineral only, going for more vulture heavy army compositions. But that obviously causes issues when dealing with carriers for example where you may not have enough gas.
Finally, high ground 1 ramp 3rd alternative bases (Radeon 12/6, Sylphid 3/7/11) would be beneficial, and I don't think it would hurt the other matchups that much. It's already difficult in ZvP to get zealots into the base regardless, and TvP Terran would mine/wall as well. Just some things to consider for the future. Maybe we'll see this trend in the future to help ZT become more balanced.
|
Northern Ireland26225 Posts
Fascinating reading to an interested and frequent observer but absolute BW scrub
It’s certainly something BW has over SC2 anyway, I really do like that the maps diverge quite a lot from each other, even if in only subtle ways
|
Maybe I'm bringing water to the sea here, but in BW the matchups have always been balanced by the map makers, whereas in SC2 they modified the units. The latter it is definitely easier if the maps are similar.
|
United States10253 Posts
On August 31 2024 22:15 Navane wrote:
Maybe I'm bringing water to the sea here, but in BW the matchups have always been balanced by the map makers, whereas in SC2 they modified the units. The latter it is definitely easier if the maps are similar.
This is true. I also think it's an observable historical artifact that we can look back to and see how the game has evolved over time. Whereas its much harder to conceptualize balance patches, you can put maps over the years in a line and see how the game has evolved.
|
|
|
|
|
|
EPT![[image loading]](https://liquipedia.net/commons/images/b/bf/Retro_1.2.jpg)
![[image loading]](https://liquipedia.net/commons/images/b/bc/Fighting_spirit_1_4.jpg)
![[image loading]](https://i.ibb.co/r6qXcg3/asdffffasd.png)
