The calculated ratings are not intended to be compared to Lichess, Elo, or Andrejic ratings. Ratings in different time-control groups are also not intended for comparison. While the player ratings from different years are much less biased than those on Lichess, they still do not have the same relational comparative accuracy as player ratings of the same month.

All rated antichess games ever played on Lichess are taken into account. Bots and all players marked as cheaters or sandbaggers prior to August 2024 have been removed and their games discarded. Cheaters and sandbaggers marked in August 2024 or later as well as those not yet identified are not excluded. No berserk correction has been made. All games are divided into four groups, based on time control. The 0+1 and ½+0 time controls are considered Ultrabullet. All other time controls are classified the same as on Lichess. The maximum rating deviation to include a player in the table is 45.0 for all time controls, except for Rapid, where the threshold is 50.0. This means that inactive players may not be shown in the tables during periods of low activity. Lichess ratings are not included in the calculation. Ratings for different time controls are calculated separately and do not affect each other.

To calculate the ratings, the Glicko-2 rating system is employed along with the regula falsi. The suggested defaults by Glicko-2 are an initial rating of 1500, an initial rating deviation of 350, an initial volatility of 0.06 (100 %), and a group size of 10–15. Once 15 games has been played, the current rating is finalized. In the algorithm, the defaults as laid out by Prof. Glickman are used. As recommended here, the “current” rating is updated after each game and used to calculate the opponents' ratings. The rating is finalized either when the maximum number of games in the group is reached, or upon reaching a significant time interval between games. The rating deviation decay is chosen in the same way as on Lichess. The update accuracy is 1 second. The rating period is 1 day. τ is 0.5.

Main differences from the Lichess algorithm:

• Over the years, to calculate the ratings, Lichess has used different algorithms and different parameters. However, after adjusting the algorithm, the previous values have not been retrospectively updated.
• Lichess applies hard limits on the rating, its deviation, volatility, delta, etc.
• Due to the complexity of the recalculation procedure, Lichess does not properly recalculate the ratings of players who played against cheaters.

Our best explanation for why the highest Blitz ratings are lower than those in Bullet/Ultrabullet is that all major tournaments on Lichess are played in Blitz in Arena format. In this format, in order to win the tournament, players are forced to “berserk” against low-rated opponents. This lowers their ratings and also indirectly affects those few players who do not berserk but are forced to play against opponents with artificially low ratings.

Among players who reached a rating deviation not exceeding the limit, ratings range from 637 to 2 788. Volatility ranges from −13 % to +100 %. The lowest rating deviation is 29.6.