Meta Over Time

Character	First month	Last month	Change	Slope / month	Players	p	q (corrected)
Aoi	47.0%	48.6%	+1.7	+0.48	679	0.004	0.076
Lei Fei	48.7%	51.5%	+2.8	+0.68	532	0.009	0.085
Jeffry	53.4%	50.1%	−3.2	−0.52	980	0.029	0.184
Vanessa	49.2%	51.1%	+1.9	+0.32	569	0.076	0.361
Lion	51.9%	49.7%	−2.1	−0.21	841	0.164	0.573
Sarah	50.7%	48.5%	−2.2	−0.24	781	0.181	0.573
Jean	50.4%	52.0%	+1.6	+0.20	767	0.234	0.635
El Blaze	52.9%	51.9%	−0.9	−0.19	598	0.352	0.727
Brad	51.2%	50.0%	−1.2	−0.17	683	0.381	0.727
Pai	50.2%	49.2%	−1.0	−0.10	845	0.404	0.727
Kage	50.6%	51.6%	+1.0	−0.12	898	0.421	0.727
Goh	49.1%	49.0%	−0.1	−0.10	871	0.531	0.841
Jacky	48.6%	49.3%	+0.7	+0.05	1,240	0.659	0.897
Akira	47.6%	48.7%	+1.2	+0.06	1,201	0.661	0.897
Wolf	49.2%	49.3%	+0.1	+0.05	960	0.723	0.916
Lau	52.4%	51.8%	−0.6	+0.03	845	0.884	0.961
Shun	48.7%	49.9%	+1.2	+0.03	654	0.915	0.961
Eileen	50.0%	50.0%	−0.0	+0.03	434	0.918	0.961
Taka	51.8%	53.8%	+1.9	+0.01	519	0.961	0.961

None of the 19 trends survives the test. The closest is Aoi: 47.0% → 48.6% (+0.48 points a month; p 0.004, q 0.076).

A character's rate can move because players figured it out, or because different people now play it. This report can't tell the two apart.

Aoi: the 296 players who played it before and after July went from 46.8% to 48.0%; the 180 who stopped before July had been at 47.3%; the 203 who started from July on are at 51.6%.

Each trend is a straight line through every match's result against its month, and its error is counted in players, so one person's run of matches is not read as many people agreeing. A win-rate trend's p is the share of 1,000 shuffles — calendar days dealt to months at random, each month keeping its number of days — whose trend is at least as far from zero; q corrects it across the 19 characters (Benjamini–Hochberg). Under the same shuffles the usual normal p called 5.5% of win-rate trends significant where 5% is right, so it is not used; for pick rates it called 5.9% and is used as it stands.

The per-match pull holds no matches at all on 18 days inside the window (1 Apr, 28 Apr, 1 May, 1 Jun, 1 Jul, 1 Aug, 7 Aug–16 Aug, 1 Sep, 1 Oct); the months around them are counted from the days that do.

Patch sources

• Official update notes (fetched 25 Sep)
• EventHubs: v1.10 balance notes
• Final Weapon: v1.10 patch notes

Character	First month	Last month	Change	Slope / month	Players	p	q (corrected)
Aoi	45.8%	47.0%	+1.2	+0.51	545	0.011	0.209
Lei Fei	49.0%	50.8%	+1.8	+0.54	392	0.103	0.730
Lion	52.8%	48.0%	−4.8	−0.38	623	0.165	0.730
Kage	50.1%	50.6%	+0.5	−0.38	624	0.185	0.730
Jacky	49.0%	50.5%	+1.5	+0.25	916	0.226	0.730
Eileen	50.8%	47.1%	−3.7	−0.59	321	0.247	0.730
Brad	51.8%	52.7%	+0.9	−0.39	500	0.311	0.730
Akira	46.1%	48.2%	+2.1	+0.20	932	0.345	0.730
El Blaze	53.1%	51.7%	−1.4	−0.30	424	0.346	0.730
Jeffry	53.1%	51.2%	−1.9	−0.31	759	0.442	0.840
Pai	50.8%	50.0%	−0.8	−0.12	656	0.559	0.887
Wolf	50.0%	50.4%	+0.4	+0.12	754	0.560	0.887
Goh	48.4%	50.5%	+2.0	+0.13	689	0.677	0.908
Shun	49.4%	48.9%	−0.6	−0.12	499	0.754	0.908
Lau	52.6%	51.2%	−1.4	−0.08	612	0.792	0.908
Sarah	50.7%	49.8%	−0.9	−0.05	617	0.834	0.908
Jean	52.9%	53.6%	+0.8	−0.05	582	0.875	0.908
Vanessa	51.9%	50.5%	−1.4	+0.05	434	0.887	0.908
Taka	56.1%	53.9%	−2.2	+0.04	375	0.908	0.908

None of the 19 trends survives the test. The closest is Aoi: 45.8% → 47.0% (+0.51 points a month; p 0.011, q 0.209).

A character's rate can move because players figured it out, or because different people now play it. This report can't tell the two apart.

Aoi: the 197 players who played it before and after July went from 45.0% to 46.4%; the 167 who stopped before July had been at 49.3%; the 181 who started from July on are at 51.2%.

Each trend is a straight line through every match's result against its month, and its error is counted in players, so one person's run of matches is not read as many people agreeing. A win-rate trend's p is the share of 1,000 shuffles — calendar days dealt to months at random, each month keeping its number of days — whose trend is at least as far from zero; q corrects it across the 19 characters (Benjamini–Hochberg). Under the same shuffles the usual normal p called 5.2% of win-rate trends significant where 5% is right, so it is not used; for pick rates it called 5.7% and is used as it stands.

The per-match pull holds no matches at all on 25 days inside the window (1 Apr, 5 Apr, 7 Apr, 28 Apr–29 Apr, 1 May–3 May, 1 Jun, 1 Jul, 1 Aug, 7 Aug–16 Aug, 1 Sep, 20 Sep–21 Sep, 1 Oct); the months around them are counted from the days that do.

Patch sources

• Official update notes (fetched 25 Sep)
• EventHubs: v1.10 balance notes
• Final Weapon: v1.10 patch notes

Character	First month	Last month	Change	Slope / month	Players	p	q (corrected)
Jeffry	54.2%	49.8%	−4.3	−0.60	379	0.008	0.095
Aoi	48.5%	51.3%	+2.8	+0.49	201	0.010	0.095
Vanessa	47.5%	51.7%	+4.2	+0.75	202	0.028	0.138
Jean	49.4%	51.4%	+2.1	+0.47	295	0.029	0.138
Lei Fei	48.0%	50.6%	+2.6	+0.49	199	0.054	0.205
Kage	49.5%	53.0%	+3.5	+0.29	390	0.115	0.364
Goh	50.2%	48.6%	−1.7	−0.28	288	0.171	0.461
Sarah	50.9%	47.7%	−3.2	−0.35	246	0.194	0.461
Shun	48.6%	51.2%	+2.7	+0.21	249	0.335	0.669
Akira	48.9%	48.0%	−0.9	−0.20	453	0.352	0.669
El Blaze	53.6%	52.8%	−0.8	−0.21	250	0.457	0.761
Taka	47.8%	55.4%	+7.6	+0.24	184	0.552	0.761
Eileen	50.9%	51.0%	+0.1	+0.29	161	0.573	0.761
Lau	50.9%	50.6%	−0.3	−0.11	334	0.594	0.761
Jacky	48.5%	48.4%	−0.1	−0.08	496	0.601	0.761
Lion	51.4%	51.1%	−0.2	−0.05	315	0.820	0.898
Pai	48.8%	48.7%	−0.1	−0.04	284	0.838	0.898
Brad	51.0%	49.5%	−1.5	−0.03	273	0.851	0.898
Wolf	48.2%	48.3%	+0.1	+0.02	325	0.911	0.911

None of the 19 trends survives the test. The closest is Jeffry: 54.2% → 49.8% (−0.60 points a month; p 0.008, q 0.095).

A character's rate can move because players figured it out, or because different people now play it. This report can't tell the two apart.

Jeffry: the 159 players who played it before and after July went from 50.6% to 49.0%; the 120 who stopped before July had been at 57.6%; the 100 who started from July on are at 52.6%.

Each trend is a straight line through every match's result against its month, and its error is counted in players, so one person's run of matches is not read as many people agreeing. A win-rate trend's p is the share of 1,000 shuffles — calendar days dealt to months at random, each month keeping its number of days — whose trend is at least as far from zero; q corrects it across the 19 characters (Benjamini–Hochberg). Under the same shuffles the usual normal p called 6.6% of win-rate trends significant where 5% is right, so it is not used; for pick rates it called 4.9% and is used as it stands.

The per-match pull holds no matches at all on 21 days inside the window (1 Apr, 28 Apr–29 Apr, 1 May–3 May, 1 Jun, 1 Jul, 1 Aug, 7 Aug–16 Aug, 1 Sep, 1 Oct); the months around them are counted from the days that do.

Patch sources

• Official update notes (fetched 25 Sep)
• EventHubs: v1.10 balance notes
• Final Weapon: v1.10 patch notes

Character	First month	Last month	Change	Slope / month	Players	p	q (corrected)
Kage	57.1%	48.2%	−8.9	−1.47	71	0.017	0.285
Lei Fei	52.1%	56.5%	+4.4	+1.64	32	0.030	0.285
Eileen	42.9%	55.5%	+12.6	+1.49	37	0.052	0.329
Jeffry	48.4%	43.5%	−4.9	−1.32	48	0.113	0.537
Taka	50.8%	39.7%	−11.1	−1.29	29	0.161	0.612
Aoi	50.6%	44.3%	−6.2	−1.47	24	0.199	0.630
Lau	57.4%	56.3%	−1.1	+0.42	63	0.418	0.975
Wolf	46.2%	47.5%	+1.3	+0.32	35	0.532	0.975
Akira	51.6%	55.2%	+3.6	+0.27	71	0.588	0.975
Jacky	46.6%	49.9%	+3.2	+0.24	84	0.614	0.975
Vanessa	45.3%	50.9%	+5.6	+0.27	31	0.621	0.975
Sarah	48.8%	45.5%	−3.4	−0.57	24	0.639	0.975
El Blaze	50.2%	47.3%	−2.9	+0.26	42	0.735	0.975
Lion	50.4%	50.0%	−0.4	−0.12	62	0.804	0.975
Brad	49.1%	46.6%	−2.5	+0.08	47	0.908	0.975
Pai	51.9%	47.4%	−4.5	−0.06	38	0.912	0.975
Shun	45.4%	47.6%	+2.2	−0.04	34	0.925	0.975
Goh	46.6%	43.9%	−2.7	−0.04	21	0.944	0.975
Jean	44.1%	47.6%	+3.5	−0.03	45	0.975	0.975

None of the 19 trends survives the test. The closest is Kage: 57.1% → 48.2% (−1.47 points a month; p 0.017, q 0.285).

A character's rate can move because players figured it out, or because different people now play it. This report can't tell the two apart.

Kage: the 38 players who played it before and after July went from 52.6% to 49.5%; the 11 who stopped before July had been at 49.0%; the 22 who started from July on are at 43.6%.

Each trend is a straight line through every match's result against its month, and its error is counted in players, so one person's run of matches is not read as many people agreeing. A win-rate trend's p is the share of 1,000 shuffles — calendar days dealt to months at random, each month keeping its number of days — whose trend is at least as far from zero; q corrects it across the 19 characters (Benjamini–Hochberg). Under the same shuffles the usual normal p called 8.0% of win-rate trends significant where 5% is right, so it is not used; for pick rates it called 4.0% and is used as it stands.

The per-match pull holds no matches at all on 20 days inside the window (1 Apr, 28 Apr, 1 May, 1 Jun, 1 Jul, 1 Aug, 7 Aug–16 Aug, 1 Sep, 20 Sep–21 Sep, 1 Oct); the months around them are counted from the days that do.

Patch sources

• Official update notes (fetched 25 Sep)
• EventHubs: v1.10 balance notes
• Final Weapon: v1.10 patch notes