The Burrow, Report 2: figures

Figures in palette v1, built from the census outputs by build-report2-figures.py. Each figure names the real-world method it borrows. Hover any mark for its value; every figure has a table view.

A seventh of us walked away

Call every game the four nerfed champions showed up in before 26.20 a hundred. Two weeks after the patch, players put them into about 86. Of every hundred games the champions still played, about one and a half more ended in a loss.

Games with the nerfed four14 of 100 went to other championsWins lost per 100 games playedabout 1.6, somewhere from 0 to 3.4
Games with the nerfed four14 of 100 went to other championsWins lost per 100 games playedabout 1.6, somewhere from 0 to 3.4

From risk communication: icon arrays and expected frequencies (Spiegelhalter, The Art of Statistics, 2019, pp. 31–34). There: a hundred people, with the extra cancers from eating bacon every day set apart. Here: a hundred games with the four nerfed champions, with the 14 that went to other champions set apart.

Method: pick and win change for Ambessa, Ashe, K’Sante and Yunara, two weeks after 26.20 against the two weeks before, minus the same change for champions Riot left alone in the same role and rank group; 95% ranges from 2,000 player reruns. Picks −14% (−17 to −10); win −1.6 points (−3.4 to +0.3). Source: Rift Census Report 2, 202,725 NA ranked games (claim trace rows 5 and 6).
Table view
MeasureChange against unchanged champions95% range
Pick rate−14%−17% to −10%
Win rate−1.6 points−3.4 to +0.3 points

Players moved further than the champions did

Each dot is a champion in the two weeks after 26.19 or 26.20, placed by how many games it had. The funnel lines show how far a champion Riot left alone drifts in two weeks (solid 95%, dashed 99.8%). One by one, most changed champions are inside that drift: on picks, 2 of 8 nerfed and 2 of 13 reskinned champions are outside the inner funnel, on wins 1 and 0. Taken together, the nerfed are low on picks and near the middle on wins, which is the pooled result in the article.

How often players picked itchange against unchanged champions−40%−20%0+20%+40%+60%3,00010,000games before and after the patch (log scale)JannaJannaK'SanteK'SanteYunaraYunaraHow often it wonpoints against unchanged champions−8−40+4+83,00010,000games before and after the patch (log scale)JannaJannaK'SanteK'SanteYunaraYunaranerfedbuffednew skin onlyunchanged
How often players picked itchange against unchanged champions−40%−20%0+20%+40%+60%3,00010,000games before and after the patch (log scale)JannaJannaK'SanteK'SanteYunaraYunaraHow often it wonpoints against unchanged champions−8−40+4+83,00010,000games before and after the patch (log scale)JannaJannaK'SanteK'SanteYunaraYunaranerfedbuffednew skin onlyunchanged

From healthcare regulation: funnel plots (Spiegelhalter et al., JRSS A, 2012; Dover and Schopflocher, Population Health Metrics, 2011). There: hospitals’ death rates against their size, with limits for what chance allows. Here: each champion’s change against its games, with limits for how far an untouched champion drifts.

Method: each champion’s change against champions Riot left alone in the same role and rank group, two weeks after the patch against the two weeks before (champions with 1,000+ games each side). Funnel width from the standard errors of the 2,000-rerun ranges, scaled by the drift of untouched champions (overdispersion). Source: Rift Census Report 2 (part1_champions.csv).
Table view
ChampionPatchChangeGamesPicksWin (points)
Yunara26.20nerfed14,089−35% (−38 to −32)−1.3 (−3.5 to +1.0)
Nocturne26.19nerfed7,068−22% (−28 to −14)−0.5 (−3.5 to +2.5)
Ryze26.19nerfed2,869−12% (−21 to −3)−3.9 (−8.5 to +0.5)
Vi26.19nerfed4,552−12% (−18 to −5)−1.1 (−4.7 to +2.6)
Ashe26.20nerfed11,935−7% (−11 to −2)−0.5 (−2.7 to +1.7)
Ambessa26.20nerfed4,334−6% (−14 to +1)+0.8 (−3.4 to +4.7)
Poppy26.19nerfed3,365−6% (−14 to +1)−2.4 (−6.5 to +1.7)
K'Sante26.20nerfed3,020−1% (−11 to +10)−5.3 (−10.0 to −0.2)
Zed26.20new skin only9,3530% (−5 to +6)+0.3 (−2.0 to +2.9)
Kennen26.20buffed2,504+1% (−14 to +16)−1.8 (−7.0 to +3.6)
Aphelios26.20new skin only7,485+1% (−7 to +10)+0.8 (−2.4 to +4.1)
Vayne26.20buffed6,210+1% (−5 to +8)+2.4 (−0.6 to +5.2)
Zyra26.20new skin only6,326+4% (−3 to +10)+0.1 (−2.7 to +3.1)
Anivia26.19new skin only3,894+5% (−4 to +15)0.0 (−4.3 to +4.2)
Yone26.20new skin only22,139+6% (+2 to +10)+0.9 (−0.7 to +2.6)
Mel26.19new skin only6,810+6% (+0 to +13)+0.2 (−2.4 to +2.9)
Sett26.20new skin only9,342+8% (+3 to +14)−0.5 (−2.9 to +1.9)
Tahm Kench26.20buffed7,502+9% (+1 to +16)+0.2 (−2.5 to +2.9)
Jax26.20new skin only8,532+9% (+3 to +16)+1.9 (−0.8 to +4.5)
Neeko26.20buffed2,946+9% (−3 to +23)+0.3 (−3.8 to +4.8)
Urgot26.19new skin only5,483+10% (+1 to +19)+0.7 (−2.5 to +3.9)
Aatrox26.19buffed6,278+10% (+3 to +18)−4.7 (−8.0 to −1.4)
Ezreal26.20new skin only10,722+10% (+5 to +16)+1.1 (−1.3 to +3.5)
Lucian26.20buffed8,091+11% (+4 to +18)−0.3 (−3.1 to +2.6)
Lee Sin26.20new skin only7,511+11% (+3 to +19)−0.5 (−3.7 to +2.6)
Kayn26.20new skin only10,345+13% (+7 to +19)+0.4 (−1.9 to +2.8)
Swain26.20buffed8,472+15% (+9 to +21)+0.1 (−2.3 to +2.5)
Aurora26.19buffed3,648+16% (+4 to +27)−2.2 (−6.7 to +2.1)
Mordekaiser26.20buffed10,063+16% (+11 to +22)+1.4 (−1.0 to +3.7)
Diana26.20buffed7,436+18% (+10 to +27)+3.2 (+0.5 to +6.0)
Fiora26.19buffed5,032+20% (+10 to +30)+1.3 (−2.2 to +4.8)
Smolder26.20buffed6,864+20% (+12 to +29)−1.5 (−4.3 to +1.5)
Lillia26.19buffed5,685+20% (+13 to +29)+2.4 (−0.9 to +5.8)
Draven26.19buffed5,763+21% (+12 to +31)+1.4 (−2.0 to +4.6)
Lillia26.20buffed7,072+22% (+15 to +31)+0.5 (−2.4 to +3.4)
Kindred26.20buffed3,752+25% (+15 to +35)+3.1 (−1.0 to +7.0)
Volibear26.19buffed7,313+30% (+22 to +38)+2.5 (−0.3 to +5.4)
Kha'Zix26.19buffed5,932+49% (+38 to +61)−3.6 (−7.3 to −0.1)
Aphelios26.19buffed6,234+50% (+37 to +62)−1.9 (−5.1 to +1.5)
Janna26.19new skin only6,409+64% (+53 to +77)+0.7 (−2.7 to +4.3)
Elise26.19buffed2,946+66% (+47 to +85)−3.4 (−8.9 to +1.8)

Nobody left on patch day, except Yunara’s players

Appearances per 100 games each day, before and after 26.20. The dashed line is where each would have gone if it had moved like the champions Riot left alone. Yunara’s line breaks on patch day; the four together drift away over two weeks.

Yunara0481216−14−7patch+7+13if it had moved like the restif it had moved like the restwhat players didwhat players didThe four 26.20 nerfs, together09182736−14−7patch+7+13if it had moved like the restif it had moved like the restwhat players didwhat players didappearances per 100 games, by day from 26.20 (7 October)
Yunara0481216−14−7patch+7+13if it had moved like the restif it had moved like the restwhat players didwhat players didThe four 26.20 nerfs, together09182736−14−7patch+7+13if it had moved like the restif it had moved like the restwhat players didwhat players didappearances per 100 games, by day from 26.20 (7 October)

From health-policy evaluation: interrupted time series (Lu et al., BMJ, 2014). There: antidepressant use per quarter before and after an FDA warning aimed at young people. Here: appearances per 100 games each day before and after a patch note aimed at pros.

Method: daily weighted appearances per 100 games (one copy per match); the expected line is each one’s pre-patch average, moved day by day with the untouched champions’ level against their own pre-patch average. The tested estimates and ranges are in the claim trace (rows 19 to 22). Source: Rift Census Report 2 (fig_data_r2.py).
Table view
DayYunaraexpectedFour nerfsexpected
-1412.228.8
-1312.529.4
-1212.427.8
-1113.129.9
-1012.729.1
-912.329.2
-814.031.2
-712.428.3
-613.230.7
-513.630.0
-412.228.0
-312.927.9
-212.527.5
-112.027.9
09.812.425.728.4
19.612.525.628.5
29.512.524.828.5
39.512.625.228.7
48.512.624.828.8
58.112.623.528.8
67.612.723.128.9
77.112.722.028.8
87.612.723.028.9
98.912.724.528.9
108.012.623.028.8
117.312.621.328.8
128.612.723.128.9
137.912.722.628.9

The stage plays by its own notes

Pick-and-ban presence at each Worlds, 2019 to 2026. Champions nerfed in the run-up stay the most drafted; buffed ones match the unchanged. Azir drops to zero and back whatever his notes say.

Average presence by change before Worlds0%25%50%75%100%'19'20'21'22'23'24'25'26nerfednerfedunchangedunchangedbuffedbuffedThree champions, asleep and awake0%25%50%75%100%'19'20'21'22'23'24'25'26AzirAzirYunaraYunaraK'SanteK'Sante
Average presence by change before Worlds0%25%50%75%100%'19'20'21'22'23'24'25'26nerfednerfedunchangedunchangedbuffedbuffedThree champions, asleep and awake0%25%50%75%100%'19'20'21'22'23'24'25'26AzirAzirYunaraYunaraK'SanteK'Sante

From statistical graphics: small multiples (Tufte, The Visual Display of Quantitative Information, 2001, pp. 170–75). There: the same small chart repeated across places or times. Here: the same presence chart repeated for each group and each champion.

Method: presence = share of games in which a champion was picked or banned (Play-In and Main Event); champions grouped by Riot’s stated intent for changes in the Worlds patch and the two before it (414 changes coded). 2026 through 26 October (51 games). Source: Leaguepedia Picks and Bans pages, read in the browser; part0_results.json.
Table view
WorldsGamesNerfedUnchangedBuffedAzirYunaraK'Sante
201912040%11%13%4%0%0%
202011432%12%6%25%0%0%
202112138%8%13%23%0%0%
202212728%9%18%72%0%0%
202311029%11%10%65%0%55%
202410630%7%19%0%0%38%
20258443%9%7%82%88%38%
20265140%10%3%0%20%51%

K’Sante, the riskier first pick

Each thread is one lane opponent with 30 or more games: the raw win rate on top, the estimate after removing small-sample noise below. Small samples are pulled hardest toward his average. Before the nerf the bottom line bunches tight; after it, it spreads, and Teemo is furthest out.

K’Sante top before the nerfraw win rateafter removing noise20%30%40%50%60%70%Kayle: 30% (33 games)Kayle: 30% (33 games)47.5%47.5%After the nerfraw win rateafter removing noise20%30%40%50%60%70%Teemo: 26% (35 games)Teemo: 26% (35 games)36.0%36.0%
K’Sante top before the nerfraw win rateafter removing noise20%30%40%50%60%70%Kayle: 30% (33 games)Kayle: 30% (33 games)47.5%47.5%After the nerfraw win rateafter removing noise20%30%40%50%60%70%Teemo: 26% (35 games)Teemo: 26% (35 games)36.0%36.0%

From statistics, via baseball: Stein’s paradox, the two-line plot (Efron and Morris, Scientific American, 1977, p. 119). There: eighteen batting averages after 45 at-bats, each pulled toward the group average. Here: K’Sante’s win rate against each lane opponent, pulled toward his average by how few games it rests on.

Method: DerSimonian-Laird random-effects fit across lane opponents with 30+ games; Q-profile 95% range for the spread. 26.19: 48.9%, spread 2.3 points (0.0 to 9.3), worst 47.5% vs Kayle. 26.20: 44.3%, spread 6.6 points (0.0 to 12.9), worst 36.0% vs Teemo. Under 2,000 games a patch, so a description, not the planned safe-pick test. Source: Rift Census Report 2 (claim trace rows 37 and 38).
Table view
Opponent26.19 raw (games)26.19 estimate26.20 raw (games)26.20 estimate
Aatrox50% (40)49.0%45% (33)44.7%
Camille61% (33)50.4%
Darius52% (54)49.2%48% (67)46.1%
Fiora35% (37)47.8%
Gangplank52% (42)49.2%52% (33)46.9%
Garen46% (87)48.5%33% (57)38.5%
Gnar53% (36)49.2%48% (33)45.8%
Gwen35% (31)48.0%40% (30)42.8%
Illaoi54% (37)49.3%26% (34)36.6%
Jax54% (41)49.3%
Kayle30% (33)47.5%50% (32)46.3%
Malphite58% (38)49.8%
Mordekaiser57% (46)49.6%45% (65)44.5%
Nasus43% (35)48.5%
Ornn42% (33)43.6%
Renekton49% (41)48.9%55% (33)48.0%
Sett49% (49)48.9%53% (59)48.5%
Shen50% (36)49.0%
Teemo50% (46)49.0%26% (35)36.0%
Tryndamere51% (41)47.2%
Urgot45% (33)48.7%38% (32)41.7%
Volibear66% (35)50.2%
Yone55% (62)49.6%36% (70)39.4%

Solo queue found mid Kog’Maw first

Share of Kog’Maw’s solo queue games played mid each day from 2 October, against limits for an ordinary day set from the two weeks before MVK’s first stage game. The shaded days, 16 and 17 October, are when MVK played him mid at Worlds; both are inside the limits. The one day outside is 7 October, the day patch 26.20 went live. By patch: 9% on 26.18, 16% on 26.19, 14% on 26.20.

0%10%20%30%40%baseline 15%95% limit99.8% limit2 Oct5 Oct8 Oct11 Oct14 Oct17 Oct20 Oct23 Oct7 Oct, patch 26.207 Oct, patch 26.20MVK plays him mid at WorldsMVK plays him mid at Worlds
0%10%20%30%40%2 Oct8 Oct14 Oct20 Oct7 Oct, patch 26.207 Oct, patch 26.20MVK on stageMVK on stage

From statistical process control: the control chart (Dover and Schopflocher, Population Health Metrics, 2011, pp. 2–3, 6). There: a health rate watched against limits set by a baseline period. Here: Kog’Maw’s daily share of mid games watched against limits set by the two weeks before MVK played him.

Method: weighted daily share (one copy per match); limits are binomial around the baseline share at 95% and 99.8%, using each day’s effective games (scaled by the weighting’s design effect) and widened by the baseline’s overdispersion (φ = 1.66). Source: Rift Census Report 2 (fig_data_r2.py); MVK games from Leaguepedia (claim trace rows 44 to 46, 60).
Table view
DayKog’Maw gamesPlayed mid95% limits
2026-10-0212010%5% to 25%
2026-10-0314715%6% to 24%
2026-10-0416518%7% to 23%
2026-10-0515221%6% to 24%
2026-10-0614717%6% to 24%
2026-10-0713026%5% to 25%
2026-10-0813017%5% to 25%
2026-10-0915016%6% to 24%
2026-10-1014613%6% to 23%
2026-10-1113811%6% to 24%
2026-10-1214214%6% to 24%
2026-10-1313811%6% to 24%
2026-10-1414610%6% to 24%
2026-10-1513510%6% to 24%
2026-10-1616318%6% to 24%
2026-10-1717522%7% to 23%
2026-10-1817215%7% to 23%
2026-10-1915110%6% to 24%
2026-10-2015715%6% to 24%
2026-10-2121417%8% to 22%
2026-10-2211217%5% to 25%
2026-10-2313320%6% to 24%
2026-10-2414321%6% to 24%
2026-10-255215%1% to 29%