2020 Season 2 Ladder Map Statistics Retrospective

ZigguratOfUr

Iraq16955 Posts

October 04 2020 05:10 GMT

2020 Season 2 Ladder Map Statistics Retrospective

Introduction:

Two months ago I did some analysis of ladder map statistics using the data from Liquipedia and using ratings from Aligulac to filter and weigh the data. Since the ladder season is now over, and some of the maps have rotated out, it seemed like a good time to take a look at where the maps ended up balance-wise and do some analysis. Methodologically this is essentially the same thing I did two months ago, so for an explanation see: Analyzing Ladder Map Balance Statistics

I am using statistics from the entire lifespan of the maps (which is more than just the last season for some of the maps), but Liquipedia has a nice statistics portal you can check out if you want to see how balance shifted over different seasons with the same maps: (Wiki)

Statistics/2020

Statistics:

Here are the raw numbers from Liquipedia (they may differ slightly from what's 'live' on Liquipedia since I took this data a few days ago).

And here they are filtered to only include games played between top 100 players on Aligulac.

And here are the results filtered and rating adjusted:

Analysis:

These results are mostly not too surprising. In TvZ, Ice and Chrome and Pillars of Gold are the Zerg maps. Submarine is incredibly good for terran with a 60% winrate in both matchups. And so on. There are a few points worth discussing though.

When we adjust for rating, Protoss is doing well on these maps in both PvZ and PvT. The fact that Protoss is overperforming relative to their Aligulac rating means that Protoss has been doing better this season (on this patch and on these maps) than in the previous season. I'd assume it's mainly due to the patch, but regardless Protoss is doing somewhat better (at least at the top 100 pro level--this trend may or may not be reflected at the very highest levels). This is a good thing, since Protoss was and still is the weakest race statistically.

If you look at the charts you can see that in all three match-ups each race has some good maps and some bad maps. This is an improvement on 2020 Season One (seen here: + Show Spoiler +

) where literally every map was considered Terran-favoured in TvZ and Zerg-favoured in ZvP. So things have been somewhat more balanced.

Golden Wall has been an interesting map for various reasons. One thing that has been argued about is whether in TvZ it's good for Terran due to the gold base, or if the hard to harass top corner bases make the map good for Zerg. In Europe we mostly saw Zergs veto it, but in Korea it was much more of a mixed bag with Terrans frequently vetoing it (INnoVation especially). Of course there's obviously an element of playstyle/personal preference to all this, but it still makes for a bit of a conundrum. Statistics wise Golden Wall has been the second best map for Terran in TvZ by a fairly wide margin. In fact as time went on Terran seemed to win even more on it, so it makes Korean terrans' dislike for it in the match-up hard to explain. As an aside, Golden Wall's creator, superouman, posted a post mortem about it that's worth a read: Golden Wall Post-Mortem.

Pillars of Gold has consistently been a decent map for Protoss in PvZ. I'm honestly not sure why given the openness of the third, and how bad it is for Terran in TvZ. Anyone have an explanation for why?

How does this compare to previous seasons?:

As of now there have been 94 ladders maps in Legacy of the Void--four of them were just added in 2020 Season 3, and seven more only appeared in the beta and as such have a negligible number of pro games on them. This leaves 83 maps which have constituted 21 map pools to compare against. Note: In this section all the stats I'm referring to are those that are filtered to include the top 100 players on Aligulac, but not rating-adjusted. You can find the full list of results I'm referring to in this section at: SCII Maps Balance Rank/Favorability

We can define a balance score for maps as the square root of the summation of the squares of the deviation (from 50%) for each match-up. So for this season we get:

By this metric Deathaura is the 5th most balanced map in LotV which is quite impressive (of course since Deathaura's still on ladder it needs to maintain that rank), and Ever Dream, Pillars of Gold, and Eternal Empire all score relatively well. Golden Wall is exactly in the middle of the pack for balance (which is not as decent as it sounds, since LotV has had loads of horrendously busted maps), Ice and Chrome is worse, and Submarine is much much worse. Note that of this year's maps, Simulacrum and Nightshade were both just about as imbalanced as Submarine (and yet much more rarely vetoed. Being 'standard' has its perks. Simulacrum in particular was very under-vetoed in my opinion). Meanwhile Purity and Industry is the 3rd most imbalanced map in LotV.

Another comparison that can be done is figuring out how each match-up on this pool compares to previous pools. For this we need to have an idea of how things have been on average in LotV.

Averaged across all of LotV, Terran has generally been slightly favoured in TvZ. The median map is Ephemeron where Terran has a 51.2% winrate. In ZvP Zerg has been quite favoured in LotV--the median map is Orbital Shipyard where Zerg has a 52.7% winrate, and Zerg is only a sizeable underdog in less than 1/6 maps. PvT has overall been quite even in LotV--the median map is Disco Bloodbath where Protoss has a 49% winrate. As a whole this is worse for Protoss than it at first sounds, because very few maps are overall favourable to Protoss--if a map is good for Protoss in PvZ, as a rule it tends to be bad in PvT and vice versa.

So for the 2020 Season 2 map pool we have this:

TvZ-wise this pool was a bit better for Terran than the average for LotV. For ZvP this pool was ever so slightly better for Protoss than the average, but because the average is quite bad for Protoss they're still below 50%. For PvT this ladder pool was also a bit better for Terran than the LotV average.

So that's that for 2020 Season 2. It was a pretty good ladder pool overall stats wise, certainly much better than Season 1. Just based on balance it was above par (despite Submarine and Ice and Chrome). Deathaura, Pillars of Gold and Submarine are of course still in the current ladder season alongside newcomers Oxide, Jagannatha, Romanticide and Lightshade, so it'll be interesting to see how the new season's pool develops.

dbRic1203

Germany2655 Posts

October 04 2020 05:36 GMT

Very interesting
This Brings the question, what would a Protoss favoured Map need?
Open Areas and large maps -> good for Z
Large Airspace behind Main/Nat, small maps -> good for T

Is there something that is exclusivly good for P?
On Z maps, Protoss strugles vs Zerg but is alright vs T. On Terran maps its Vice versa.

ZigguratOfUr

Iraq16955 Posts

October 04 2020 05:53 GMT

On October 04 2020 14:36 dbRic1203 wrote:
Very interesting
This Brings the question, what would a Protoss favoured Map need?
Open Areas and large maps -> good for Z
Large Airspace behind Main/Nat, small maps -> good for T

Is there something that is exclusivly good for P?
On Z maps, Protoss strugles vs Zerg but is alright vs T. On Terran maps its Vice versa.

I'm really not sure if there is a way. There's a few maps where Protoss does decently well against both Zerg and Terran--Port Aleksander, Blackpink. But not that many. And in LotV there's really no Protoss equivalent to say Purity and Industry for Terran or Apotheosis for Zerg.

I don't know how you could make a map that's very good for Protoss in both matchups outside of doing some very non-standard things (e.g. Acropolis except the entire middle of the map is made uncreepable).

sneakyfox

8216 Posts

October 04 2020 08:44 GMT

Thanks for the post, it's very interesting.

Drfilip

Sweden590 Posts

October 04 2020 11:57 GMT

The post itself is interesting. The link to the Golden Wall analysis was a huge plus.

WombaT

Northern Ireland24460 Posts

October 04 2020 13:16 GMT

Cheers for the work man, interesting read! Still salty GW is gone so soon, we hardly knew ye! Indeed given how the WRs seemed to fluctuate a bit, Korean Terrans vetoing despite it on paper being good for TvZ etc, there was probably a fair bit to still flesh out and figure out on that map.

[Phantom]

Mexico2170 Posts

October 04 2020 22:56 GMT

Great post!

However could you elaborate on the third graph when you say "And here are the results filtered and rating adjusted:" what exactly did you filter and what you adjusted? It wasn't clear for me, and ZvP basically flipped completely.

ZigguratOfUr

Iraq16955 Posts

October 04 2020 23:17 GMT

On October 05 2020 07:56 [Phantom] wrote:
Great post!

However could you elaborate on the third graph when you say "And here are the results filtered and rating adjusted:" what exactly did you filter and what you adjusted? It wasn't clear for me, and ZvP basically flipped completely.

Basically the filtering based on Aligulac rating is there to exclude games between weaker players--there are some amateur events on Liquipedia, which aren't really relevant to balance at the pro level. So filtering gives you a better outlook about balance on a given map.

The rating adjustment meanwhile is there to weigh the values of results. Reynor is expected by Aligulac to win 2/3 of his games against Neeb. Therefore if he plays three games against Neeb and wins two this is exactly as expected and the rating adjusted winrate would still be at 50%. This is not useful to measure balance (since Aligulac ratings adjust), but is useful to compare maps to other maps within the same ladder pool. If on Map A you had a bunch of TvPs where the Terran is the underdog, and on Map B you had a bunch of TvPs where the Protoss is the underdog, using rating adjusted winrates allows for a more apples to apples comparison than looking at the unadjusted winrate.

In the case of ZvP, in the top 100 on Aligulac Zergs are generally higher rated than the Protosses which means that Aligulac expects the Zerg to be favored more often than not. So rating adjusting things shifts things towards the Protoss.

The filtered win rate for Deathaura is 51%, and the rating adjusted win rate for it is 47.9%--that's a shift of 3% towards Protoss. Meanwhile the filtered win rate for Golden Wall is 55.6%, but the rating adjusted win rate for it is 47.2%--a much bigger shift. This does not mean that Deathaura or Golden Wall are "good" or "bad" for Protoss in PvZ. But it does mean that Deathaura and Golden Wall are evaluated to be about the same balance wise despite the raw win rates of 51% and 55.6% at first glance being quite different. Now why would Golden Wall end up with more games where Zerg is favoured than Deathaura? My guess is that because Golden Wall is often vetoed in Bo3s unlike Deathaura you end up seeing it proportionally more in the later rounds of the tournament where you end up with the top zergs really crushing Protosses and having large rating edges.

This is explained more in depth in an earlier post I made (https://tl.net/forum/starcraft-2/561745-analyzing-ladder-map-balance-statistics).

esReveR

United States567 Posts

October 05 2020 09:29 GMT

On October 04 2020 14:36 dbRic1203 wrote:
Is there something that is exclusivly good for P?

I would guess lots of small chokes. This helps maximize the effectiveness of their aoe abilities and makes it easier to chop up armies with force fields. Submarine has this feature and was their best map in all asymmetric matchups, while ice and chrome has larger spaces and was among their worst maps.

ZigguratOfUr

Iraq16955 Posts

October 05 2020 16:08 GMT

#10

On October 05 2020 18:29 esReveR wrote:

Show nested quote +

In PvT Submarine was (and is) extremely good for Terran.

Please or register to reply.

2020 Season 2 Ladder Map Statistics Retrospective

Completed

Ongoing

Upcoming