Hello, this is the developer. Something that has been bothering me for a while is how further ahead Taihou/Fun on the Baize has been from the other skins. Finally, I stopped being lazy and decided to actually look at the vote logs to see what's going on.
Below you can see the statistics for the matches that particular Taihou skin has been in:
Target: 307076 - Taihou / Fun on the Baize
Total votes: 190
Wins: 154 Losses: 36 Win rate: 81.05%
Unique IPs (overall): 135
IP uniqueness (overall): 70.90%
Head-to-head (target vs opponent)
---------------------------------------------------------------------------------------------------------
Opponent Games W L Win% UniqIP Uniq%
---------------------------------------------------------------------------------------------------------
319011 - Kashino / Hot Springs Relaxation 31 17 14 54.84% 28 90.00%
307161 - Hakuhou / Floral Ninja Roleplay 15 13 2 86.67% 14 92.86%
107250 - Cowpens / Cowpens 11 7 4 63.64% 10 90.00%
307075 - Taihou / Seaside Daydreams 10 10 0 100.00% 9 88.89%
319013 - Kashino / Robust Floral Strings 7 5 2 71.43% 6 83.33%
107251 - Cowpens / Milky Ministrations 6 5 1 83.33% 6 100.00%
319014 - Kashino / Fresh and Sweet 5 0 5 0.00% 4 75.00%
499091 - Hindenburg / Delirious Duel 4 4 0 100.00% 4 100.00%
718012 - Kronshtadt / Animal Agent 3 3 0 100.00% 3 100.00%
...
As I've already explained in the FAQ section of the About page, 85% of the pairups against a skin will be similarly rated skins, so it makes a lot of sense that the most faced oponent is Kashino/Hot Springs Relaxation, since that's the skin in second place and there's not many skins close to this rating. Taihou and Kashino in this particular matchup are considered almost the same size, with Taihou having a slight advantage, winning 54% of the time. The elo rating actually seems to be pretty accurate here, predicting that Taihou should beat Kashino 59.4% percent of the time, not really that far off of reality.
A very weird outlier is Kashino/Fresh and Sweet, beating Taihou 5 out of 5 times. Not sure why that would be the case, but 5 matches isn't a lot (and 2 of those were from the same person), so maybe just a fluke.
Now let's take a look at Kashino/Hot Springs Relaxation:
Target: 319011 - Kashino / Hot Springs Relaxation
Total votes: 194
Wins: 146 Losses: 48 Win rate: 75.26%
Unique IPs (overall): 149
IP uniqueness (overall): 76.68%
Head-to-head (target vs opponent)
---------------------------------------------------------------------------------------------------------
Opponent Games W L Win% UniqIP Uniq%
---------------------------------------------------------------------------------------------------------
307076 - Taihou / Fun on the Baize 31 14 17 45.16% 28 90.00%
319013 - Kashino / Robust Floral Strings 23 11 12 47.83% 23 100.00%
319014 - Kashino / Fresh and Sweet 14 12 2 85.71% 14 100.00%
307161 - Hakuhou / Floral Ninja Roleplay 11 10 1 90.91% 11 100.00%
107250 - Cowpens / Cowpens 10 7 3 70.00% 10 100.00%
307075 - Taihou / Seaside Daydreams 8 7 1 87.50% 8 100.00%
107251 - Cowpens / Milky Ministrations 7 5 2 71.43% 7 100.00%
1199011 - Gouden Leeuw / An Intimate Cleaning 5 5 0 100.00% 5 100.00%
...
Weirdly enough, Hot Springs Relaxation usually loses to Robust Floral Strings. Also, most people do agree that Kashino looks bigger than Cowpens.
Now let's pick a skin that isn't that highly rated. Bremerton/Day-Off Date is currently rated at #387 and here are its stats:
Target: 103243 - Bremerton / Day-Off Date
Total votes: 190
Wins: 110 Losses: 80 Win rate: 57.89%
Unique IPs (overall): 140
IP uniqueness (overall): 73.54%
Head-to-head (target vs opponent)
---------------------------------------------------------------------------------------------------------
Opponent Games W L Win% UniqIP Uniq%
---------------------------------------------------------------------------------------------------------
107120 - Hornet II / Hornet II 3 2 1 66.67% 3 100.00%
107141 - Ticonderoga / Show Stopper! 3 2 1 66.67% 3 100.00%
904020 - Strasbourg / Strasbourg 3 1 2 33.33% 3 100.00%
202280 - Dido µ / Dido µ 2 2 0 100.00% 2 100.00%
207051 - Formidable / The Lady of the Beach 2 2 0 100.00% 2 100.00%
399050 - Hakuryuu / Hakuryuu 2 2 0 100.00% 2 100.00%
406011 - Weser / Obsidian Elegance 2 2 0 100.00% 2 100.00%
804010 - Masséna / Masséna 2 2 0 100.00% 2 100.00%
102341 - Santa Fe / The Rumor Mill Never Sleeps 2 1 1 50.00% 2 100.00%
203072 - York / Summoner from the Azure Depths 2 1 1 50.00% 2 100.00%
206071 - Albion / Silvermoon Faerie Princess 2 1 1 50.00% 2 100.00%
207140 - Formidable μ / Formidable μ 2 1 1 50.00% 2 100.00%
299050 - Plymouth / Plymouth 2 1 1 50.00% 2 100.00%
...
As you can see, matches are way more equaly distributed. This makes a lot of sense, since there are lots of similarly-rated skins.
So, Taihou is rated that high because she mostly is matched up with Kashino and she wins around 54% of the time. Her rating predicts she should win 59.4% of the time in that case, so she is slightly overrated but not really by that much. I do think that it's a good idea to make her be matched up with more skins, it is totally possible that she'd lose to some currently lower rated skins. e.g. it's not obvious to me that Taihou is bigger than Suzuya for example.
To try to make this better I made it so there has to be a pool of at least 10 skins a skin can be matched up against. So previously it'd be almost impossible for Taihou to be matched up against Suzuya, but now it can totally happen. This could make things worse (e.g. Taihou wins even more and climbs even higher) or it could make things better (e.g. we find out that she actually loses to someone lower down consistenly). Guess we'll just have to wait and see.
Yours truly,
Dekapai Developer