Most differences in results are within chance.They do not justify criticizing players.

Terms · 45 metrics

Glossary

For how the numbers are calculated, see the methodology in detail (Japanese).

Mahjong terms

Riichi
A declaration made with a closed hand in tenpai.
Dama
Keeping a closed tenpai hand without declaring riichi.
Tenpai
A hand one tile away from winning.
Shanten
The number of steps a hand is away from tenpai; a tenpai hand is 0-shanten.
Deal-in
A discard that allows another player to win by ron.
Exhaustive draw
A hand ending without a winner because no playable tiles remain in the wall.
Tsumo
A win on a self-drawn tile; the word also refers to drawing a tile.
Ron
A win on another player's discard.
Ukeire
The tiles, or number of remaining tiles, that improve a hand.
Tile efficiency
Choosing discards to keep a hand moving toward tenpai efficiently.

Concepts

Estimate (skill estimate)
The actual record shrunk toward the league average. With few games luck weighs heavily, so the estimate stays close to the average; the more games, the closer it comes to the actual record.
Weight of the record
How much of the player's own record goes into the skill estimate (n ÷ (n + k)); the league average fills the rest. It approaches 100% as games add up. It is not the probability that the player is strong.
Count to half stability
The number of observations at which the weight of the record reaches 50%. A lower count means the metric reveals a player's individual tendencies sooner.
90% interval
The uncertainty of the estimate: built so that, across all players, the true value of at least 90% of them lies inside it.
Percentile
Position among the players with enough games in the same season (100 is best). Style metrics have no good or bad side, so there 100 is the highest value.
Style tags
Given when the career estimate is in the league's top (or bottom) 25% and the weight of the record for that metric is at least 50%.
Hand balance breakdown
The points won or lost per hand, split into wins (riichi, calls, dama), deal-ins (to riichi, open hands, dama), others' tsumo and the rest. They add up to the average balance per hand.
Tile efficiency match rate
The share of early and middle discards, with no riichi out, that keep the most tiles left that advance the hand (maximum ukeire). Speed-first hand building scores higher. It shows no link to results, so it is treated as a difference of style, not of quality. More (Japanese)
Tenpai turn (luck-adjusted), immediate riichi rate
In hands that reach tenpai closed by turn 12: the tenpai turn with starting hand and draw luck set to average, and the share of riichi declared as soon as tenpai. The immediate riichi rate is compared as the difference from each season's league average. More (Japanese)
Dama rate
The share of wins made without riichi or calls. Aligned with the Paifuya and Tenhou definition on 9/28.
Luck metrics
Differences from the league average; plus means luckier than average. Luck is not skill, so no estimate is made.

All metrics

GroupMetricUnitLeague averageCount to half stability
ResultsAverage placementgames2.50146 games (62–no upper bound)
ResultsPoints per gamegames0.0136 games (56–no upper bound)
ResultsPoints per game, luck removedgames0.098 games (41–no upper bound)
Results1st-place rategames25.1%128 games (62–no upper bound)
Results4th-place rategames24.9%121 games (51–no upper bound)
StyleWin ratehands20.0%793 hands (495–1,992)
StyleDeal-in ratehands10.9%938 hands (623–1,819)
StyleRiichi ratehands22.7%738 hands (450–2,228)
StyleCall ratehands23.0%66 hands (47–103)
StyleFirst riichi rateriichi76.0%452 riichi (180–no upper bound)
StyleAverage riichi turnriichi9.34170 riichi (113–409)
StyleAverage win valuewins6,590248 wins (138–1,398)
StyleAverage deal-in valuedeal-ins5,626294 deal-ins (105–no upper bound)
StyleWin efficiencyhands1,4748,611 hands (1,687–no upper bound)
StyleDeal-in losshands6392,297 hands (975–no upper bound)
StyleNet efficiencyhands8351,987 hands (913–no upper bound)
StylePush rate vs riichi (1-shanten)hands facing riichi40.8%182 hands facing riichi (106–810)
StylePush rate vs riichi (2+ shanten)hands facing riichi25.6%91 hands facing riichi (63–194)
StyleTenpai rate after the first callfirst calls25.6%62 first calls (41–138)
StyleAverage win turnwins12.384,109 wins (354–no upper bound)
StyleTenpai rate at exhaustive drawsexhaustive draws40.0%179 exhaustive draws (116–508)
StyleStaying in vs riichihands facing an opponent’s riichi14.1%2,028 hands facing an opponent’s riichi (530–no upper bound)
StyleTile efficiency match ratediscards72.9%337 discards (285–429)
StyleAverage ukeire loss (tiles)discards1.06595 discards (501–742)
StyleBreaking shape for valuediscards2.0%955 discards (755–1,299)
StyleTenpai turn (luck-adjusted)tenpai hands8.171,646 tenpai hands (288–no upper bound)
StyleImmediate riichi ratetenpai hands71.6%111 tenpai hands (73–204)
StyleDama ratewins13.3%62 wins (50–89)
StyleRiichi incomeriichi2,623371 riichi (187–no upper bound)
StyleRiichi win rateriichi46.0%175 riichi (111–380)
StyleRiichi deal-in rate Ariichi14.3%2,594 riichi (347–no upper bound)
StyleRiichi deal-in rate Briichi11.0%912 riichi (293–no upper bound)
StyleRiichi good-wait rateriichi49.7%299 riichi (168–1,001)
StyleRiichi exhaustive draw rateriichi19.3%242 riichi (152–700)
StyleCall incomecalls487595 calls (228–no upper bound)
StyleWin rate after callingcalls29.8%256 calls (139–836)
StyleDeal-in rate after callingcalls11.5%256 calls (153–647)
LuckStarting hand (shanten)hands0 (baseline)6,920 hands (1,468–no upper bound)
LuckDora and red fives dealthands0 (baseline)(almost all luck)
LuckDraw lucktile draws0 (baseline)(almost all luck)
LuckUra-dora luckriichi wins0 (baseline)918 riichi wins (188–no upper bound)
LuckIppatsu rateriichi wins0 (baseline)(almost all luck)
LuckAvoiding a chase riichi (/first riichi)first riichi0 (baseline)(almost all luck)
LuckMangan+ tsumo against dealer ratetsumo against0 (baseline)(almost all luck)
LuckPoints from luck (/game)games0 (baseline)2,424 games (133–no upper bound)