Win Probability Added

Dural is in the model, since she is somebody's opponent, but she is left off the rows and out of the baseline, as on the tier list.

How WPA is worked out

A win rate mixes three things: who the character's players were paired against, how good those players are, and the character. WPA predicts every match from the first two and gives the character whatever is left over.

1. Ranks. Each match starts from the win chance the two ranks give, read off the ladder's own rank-against-rank record (the Rank Impact report). A 35 facing a 31 is not an even match, whatever either of them picked.
2. Player skill. Each player gets one number: how much better or worse than their rank they play, fitted over every match they played with every character. A player who beats most people at their rank gets a positive number, whichever character they were on.
3. The character. Each character gets one number, fitted at the same time: how much more it wins than the ranks and the players predict. In points, that is its WPA: in a match that was even on ranks and players, picking it makes the chance 50% plus the WPA. Zero is the average of the nineteen characters, each counted once.
4. Opponents' characters. What is left of the bar is the characters it met. The baseline counts every character once and the most-picked characters sit below it, so everyone faces a slightly soft field; this piece is that.

The character's number is read most directly off people who play more than one character, the same person on two different characters, and off opponents, since every match has two characters in it. The table counts how many of each character's players also played another. A character whose players almost all play only it is harder to separate from them, and its interval is wider for it.

What it cannot say. Characters are chosen, not dealt, so this is an adjusted comparison and not proof of cause: a character that players pick up while they are improving, or bring out only in matchups they like, carries that into its number. Skill is one number per player for the whole period. And only matches with both players in the band count, as on every banded report.

The interval: players are dealt into 20 groups by a hash of their id, and the model is refitted 20 times, each time leaving out one group's players and all their matches. The spread of those refits is the error (a grouped jackknife). With 20 groups the p-value is read on a t distribution with 19 degrees of freedom (the 95% line is at t = 2.09), and q corrects the p-values for the characters tested together (Benjamini–Hochberg, 5%).

Survive correction0 of 19

Matches counted70,432

Players463

Rank bandRank 40–46

Period2026/03/26 – 2026/10/18

Every character, in pieces

points against 50%; the interval is 95%

Character	Win rate	Edge	Ranks	Player skill	Opponents' characters	WPA	95% interval	Players	Also play others	p	q (corrected)
Eileen	53.0%	+3.0	−0.4	−0.1	+0.2	+3.3	+0.7 … +6.0	161	118	0.018	0.055
El Blaze	51.4%	+1.4	+0.1	−1.0	+0.4	+1.9	+0.7 … +3.2	250	198	0.005	0.019
Lei Fei	49.4%	−0.6	−0.2	−2.5	+0.5	+1.7	−0.7 … +4.0	199	134	0.152	0.321
Lion	51.4%	+1.4	+0.6	−1.1	+0.5	+1.4	+0.2 … +2.6	315	232	0.020	0.055
Taka	53.6%	+3.6	+1.1	+0.8	+0.4	+1.2	−1.4 … +3.8	184	155	0.332	0.573
Vanessa	50.1%	+0.1	+0.2	−1.6	+0.5	+1.1	−0.6 … +2.9	202	143	0.190	0.361
Aoi	50.0%	+0.0	+0.4	−1.6	+0.5	+0.7	−1.3 … +2.7	201	132	0.474	0.609
Sarah	49.9%	−0.1	−0.0	−1.2	+0.5	+0.6	−1.1 … +2.4	246	165	0.478	0.609
Brad	50.4%	+0.4	+0.3	−0.8	+0.5	+0.5	−0.9 … +1.9	273	199	0.481	0.609
Shun	49.4%	−0.6	+0.7	−2.1	+0.5	+0.3	−1.0 … +1.6	249	167	0.614	0.648
Kage	51.9%	+1.9	+0.9	+0.4	+0.5	+0.1	−1.1 … +1.3	390	310	0.847	0.847
Jeffry	50.9%	+0.9	−0.3	+1.0	+0.5	−0.3	−1.7 … +1.0	379	325	0.598	0.648
Lau	50.8%	+0.8	+0.3	+0.5	+0.5	−0.4	−1.4 … +0.6	334	281	0.400	0.609
Pai	49.8%	−0.2	+0.2	−0.5	+0.5	−0.4	−2.1 … +1.3	284	216	0.595	0.648
Jean	49.8%	−0.2	−0.3	+1.0	+0.6	−1.4	−2.7 … −0.1	295	245	0.032	0.075
Akira	48.8%	−1.2	−0.3	+0.8	+0.5	−2.1	−3.2 … −1.1	453	286	<0 .001< td>	0.004
Wolf	48.1%	−1.9	−0.9	+0.8	+0.5	−2.4	−3.7 … −1.0	325	242	0.001	0.007
Jacky	49.0%	−1.0	−0.5	+1.9	+0.5	−2.9	−3.9 … −1.9	496	369	<0 .001< td>	<0 .001< td>
Goh	48.8%	−1.2	−0.3	+1.5	+0.6	−2.9	−4.4 … −1.4	288	230	<0 .001< td>	0.004

Character	Win rate	Edge	Ranks	Player skill	Opponents' characters	WPA	95% interval	Players	Also play others	p	q (corrected)
Eileen	51.0%	+1.1	−1.6	−9.8	−0.3	+12.7	−26.4 … +40.1	37	18	0.530	0.804
Lau	54.0%	+4.0	+1.3	−1.4	+0.1	+4.0	+0.4 … +7.5	63	44	0.031	0.294
Aoi	48.4%	−1.6	−0.9	−3.9	−0.4	+3.5	−12.7 … +19.0	24	11	0.657	0.804
Pai	48.4%	−1.6	−1.2	−3.1	−0.5	+3.2	−2.0 … +8.3	38	22	0.218	0.804
Jeffry	47.1%	−2.9	−3.0	−1.9	+0.1	+1.9	−13.0 … +16.5	48	39	0.797	0.804
Shun	48.1%	−1.9	+0.1	−3.7	+0.2	+1.5	−7.0 … +9.9	34	16	0.716	0.804
Kage	49.7%	−0.3	−0.0	−1.5	+0.3	+1.0	−3.9 … +5.9	71	48	0.677	0.804
Taka	53.5%	+3.5	+0.9	+1.6	+0.1	+0.9	−6.6 … +8.3	29	18	0.804	0.804
Brad	49.2%	−0.8	+0.2	−1.9	+0.1	+0.7	−4.2 … +5.5	47	33	0.775	0.804
Akira	51.8%	+1.8	+0.6	+1.8	+0.2	−0.8	−5.0 … +3.4	71	41	0.709	0.804
Lion	49.0%	−1.0	−0.0	+0.3	+0.1	−1.4	−5.3 … +2.5	62	38	0.448	0.804
El Blaze	53.1%	+3.1	+0.4	+4.2	+0.4	−1.9	−8.4 … +4.7	42	27	0.563	0.804
Lei Fei	54.5%	+4.5	+3.1	+3.3	+0.3	−2.1	−7.5 … +3.3	32	15	0.425	0.804
Jean	46.3%	−3.7	−1.8	+0.3	+0.1	−2.3	−6.8 … +2.3	45	41	0.307	0.804
Vanessa	45.4%	−4.7	−1.6	−1.3	+0.6	−2.3	−9.1 … +4.6	31	13	0.491	0.804
Goh	47.6%	−2.4	−0.2	+0.2	+0.3	−2.6	−17.6 … +12.8	21	14	0.729	0.804
Jacky	51.2%	+1.2	+1.3	+3.5	+0.2	−3.8	−6.7 … −0.8	84	52	0.015	0.283
Wolf	48.1%	−1.9	−0.8	+3.1	+0.4	−4.6	−15.7 … +7.0	35	17	0.421	0.804
Sarah	46.4%	−3.6	−2.1	+5.8	+0.4	−7.8	−30.4 … +18.8	24	14	0.559	0.804

Dural is in the model, since she is somebody's opponent, but she is left off the rows and out of the baseline, as on the tier list.

A win rate mixes three things: who the character's players were paired against, how good those players are, and the character. WPA predicts every match from the first two and gives the character whatever is left over.

1. Ranks. Each match starts from the win chance the two ranks give, read off the ladder's own rank-against-rank record (the Rank Impact report). A 35 facing a 31 is not an even match, whatever either of them picked.
2. Player skill. Each player gets one number: how much better or worse than their rank they play, fitted over every match they played with every character. A player who beats most people at their rank gets a positive number, whichever character they were on.
3. The character. Each character gets one number, fitted at the same time: how much more it wins than the ranks and the players predict. In points, that is its WPA: in a match that was even on ranks and players, picking it makes the chance 50% plus the WPA. Zero is the average of the nineteen characters, each counted once.
4. Opponents' characters. What is left of the bar is the characters it met. The baseline counts every character once and the most-picked characters sit below it, so everyone faces a slightly soft field; this piece is that.

The character's number is read most directly off people who play more than one character, the same person on two different characters, and off opponents, since every match has two characters in it. The table counts how many of each character's players also played another. A character whose players almost all play only it is harder to separate from them, and its interval is wider for it.

What it cannot say. Characters are chosen, not dealt, so this is an adjusted comparison and not proof of cause: a character that players pick up while they are improving, or bring out only in matchups they like, carries that into its number. Skill is one number per player for the whole period. And only matches with both players in the band count, as on every banded report.

The interval: players are dealt into 20 groups by a hash of their id, and the model is refitted 20 times, each time leaving out one group's players and all their matches. The spread of those refits is the error (a grouped jackknife). With 20 groups the p-value is read on a t distribution with 19 degrees of freedom (the 95% line is at t = 2.09), and q corrects the p-values for the characters tested together (Benjamini–Hochberg, 5%).