Playing13 min read

Bullet Chess: What It Is, and How 682,934 Games End

Bullet chess is anything under three minutes a side, and the clock decides about half of its games. We read 682,934 Lichess bullet games from 1 October 2026 to see how they end at five rating bands, and how often the player who flags was winning.

The starting position

Bullet chess is chess on a clock so short that the clock decides about half of all games: 52.4% of the 682,934 Lichess bullet games we read ended with a flag, not a mate. That fact shapes everything about the format, and none of the ranking pages we read puts a number on it. We read the first part of the Lichess database file for October 2026, all games from 1 October, and sorted the bullet games by rating band, by time control and by how they ended. Below are the definitions, the rules that bite, our data, and an honest answer to whether bullet is good for you.

The short answer

Bullet is chess where the flag is a main opponent. On Lichess it means a clock of 179 seconds or less by a formula that counts increment, which makes 1+0 and 2+1 bullet and 2+2 and 3+0 blitz. In our data, more than half of games are lost on time at every band below 2200, and between 27.3% and 43.1% of the players who flagged had a clearly winning position at the last evaluation. Whether bullet makes you better is an open question that experts answer differently, and our data does not settle it.

What counts as bullet chess

Bullet is a label used by the big sites, not by the rule book. As of November 2026 the two sites that matter draw the line in slightly different places.

  • Chess.com. Its terms page defines bullet as time controls faster than 3 minutes per player. It names 1|0 and 2|1 as the common bullet controls and describes hyperbullet as less than a minute with no increment.
  • Lichess. Each game gets an estimated duration, the starting clock in seconds plus 40 times the increment in seconds. Its FAQ puts 29 seconds or less in UltraBullet and 179 seconds or less in bullet. Blitz runs to 479 seconds.
  • FIDE. We did not find a bullet category in the FIDE Laws of Chess as checked in November 2026. Blitz there means 10 minutes or less per player with the increment counted at 60 seconds for each second, so every bullet game also sits inside FIDE's blitz definition.

Which controls are bullet on Lichess

The arithmetic is ours, from the Lichess formula. The increment is where people get caught.

ControlEstimated durationLichess label
0.5+0 (30 seconds)30 secondsBullet
1+060 secondsBullet
1+1100 secondsBullet
2+1120 secondsBullet
2+2200 secondsBlitz
3+0180 secondsBlitz

So 2+1 is bullet while 2+2 is already blitz, and 3+0 falls one second over the line. Chess.com's wording, faster than 3 minutes, draws the same line for games without an increment. A control with only 15 seconds a side works out to UltraBullet, and we dropped those games from our data. We did not source whether each site offers every control in the table, so check the lobby.

The rules that bite in bullet

No piece moves differently in bullet. Three rules matter more here than in slower chess.

  • The flag. A player who runs out of time normally loses, even from a won position.
  • A flag against too little material is a draw. FIDE says the game is drawn if the position is such that the claimant cannot checkmate the opponent's king by any series of legal moves. Lichess says the same in its FAQ. If your flag falls but your opponent has only a bare king, or no way to mate you, you draw. We counted these in our data, and they are rare.
  • Premoves. Chess.com has a support article on premoves, where you enter a move before your opponent has replied. We could not verify the details, such as how many you can queue, so we do not state them. Community guides we found say to premove only when you are certain of the reply.

How bullet games end: our data

We streamed the Lichess open database file for October 2026 (CC0 license) and read its first 1,953,110 games, all played on 1 October 2026. After dropping UltraBullet games and games with a rating gap over 400 between the players, 682,934 bullet games remained, alongside 879,452 blitz games. Lichess's own classification decides the speed. Game level figures use the average of both players' ratings, which are Lichess ratings. A game counts as a time forfeit when its Termination header says so, a checkmate when the last move ends in a mate sign, and everything else goes under resignation or other: resignations, stalemates, agreed draws, repetitions and abandoned games.

Average ratingGamesTime forfeitCheckmateResignation or otherAverage plies
Under 100020,76064.3%19.3%16.4%45.3
1000 to 1399113,89157.3%21.2%21.5%54.7
1400 to 1799218,30353.6%22.2%24.2%62.0
1800 to 2199235,94750.9%22.7%26.4%68.3
2200 and above94,03344.7%22.2%33.1%76.6
All bands682,93452.4%22.1%25.5%64.5

Source: Lichess open database, October 2026 file, games of 1 October 2026, CC0. Every row sums to 100%. A ply is one move by one side, so 64.5 plies is about 32 full moves.

What the table says

  • The flag falls less as rating rises. Time forfeits drop from 64.3% under 1000 to 44.7% at 2200 and above. We read that as better players finishing the game before the clock does.
  • Checkmate barely moves. It stays between 19.3% and 22.7% at every band. What changes is that resignation takes the share the flag loses: 16.4% under 1000, 33.1% at the top.
  • Games get longer with rating. The average game lasts 45.3 plies under 1000 and 76.6 at 2200 and above. Games under 1000 are shorter, and more of them end on the clock.
  • Draws are rare but grow. The draw share rises from 1.5% under 1000 to 5.2% at 2200 and above.

For blunder rates by mover rating and by remaining clock, which we published yesterday for bullet, blitz and rapid, see our blitz chess guide. We do not repeat those tables here.

Who flags while winning

The numbers above say who loses on time. The next question is whether they were losing anyway. For the bullet games that ended on time and carried computer evaluations, we took the last available evaluation from the flagged player's side. A score of plus 200 centipawns or better, about two pawns, counts as winning, minus 200 or worse as losing, and the middle as roughly equal.

Average ratingFlagged games with evalsFlagged while winningRoughly equalFlagged while losing
Under 10001,57643.1%11.9%44.8%
1000 to 13995,13938.7%13.3%47.7%
1400 to 17996,55634.1%14.7%50.7%
1800 to 21994,84032.8%14.8%51.0%
2200 and above1,67327.3%15.8%54.6%

Under 1000, 43.1% of players who lost on time were ahead by two pawns or more at the last evaluation. At 2200 and above the share is 27.3%. The small remainder in each row is games drawn on time, which are the cases where the flag fell and the opponent could not mate. They make up 1.3% of flagged games at 1800 to 2199 and 2.3% at 2200 and above.

Here is how we read it, and it is an interpretation. Weaker players reach winning positions often enough but do not convert them in the seconds that remain, so a big lead turns into a flag loss. Stronger players convert faster, or simplify, and lose on time mostly when the position was already bad. A player at 2200 and above who flags is more likely to have been the one defending.

What limits this number

  • The last evaluation is not the moment of the flag. It is from the last analysed ply. A position can be winning on the board and still impossible to convert in the seconds left, and the evaluation says nothing about the clock.
  • Evaluated games are self-selected. Someone asked for computer analysis, which skews toward people who want to improve and toward games that went wrong.
  • It is part of one day. All games are from 1 October, so another day could differ.
  • The ratings are Lichess ratings. They are not comparable to Chess.com or FIDE ratings. Our guide to the chess rating system and how ratings differ across sites explains why.

What the increment changes: 1+0 against 2+1

The cleanest bullet comparison is the two most common controls. Across all bands, 1+0 has 507,493 games in our sample and 2+1 has 125,132.

Measure, all bands1+02+1
Time forfeit56.0%37.0%
Checkmate22.7%19.1%
Resignation or other21.3%43.9%
Average plies64.763.9
Blunders per 100 moves4.033.44
Flagged while winning37.6%22.8%

The games last about as long in plies, 64.7 against 63.9, but they end differently. With one extra second per move, 19 points fewer end on the flag, and resignation takes the difference. The flagged-while-winning share falls too, from 37.6% of flagged games with evals in 1+0 to 22.8% in 2+1. Those rest on 14,770 and 3,038 flagged games with evals.

The gap holds inside each band, so it is not just a rating effect. The 2+1 row at 2200 and above rests on 5,090 games:

Average rating1+0 time forfeit2+1 time forfeit
Under 100073.2%47.3%
1000 to 139966.0%38.6%
1400 to 179959.5%35.3%
1800 to 219953.0%36.2%
2200 and above45.1%33.6%

This is not a controlled experiment. The two controls draw different player mixes: people who choose 2+1 may be more patient, or may play it when they want longer games, and the same player may behave differently on each. What the data supports is narrower. Games on 2+1 end by the flag far less often at every band, and the increment changes the type of game you are playing. 1+1 and 0.5+0 are also bullet. Among 474 flagged games with evals in 1+1, 27.6% had the flagged player winning. In the 0.5+0 half minute games, 42.3% of 435 did.

Bullet against blitz: first moves

The openings people choose also change with the speed. This is White's first move, by the average rating of both players:

Average rating1.e4 in bullet1.e4 in blitz
Under 100061.8%64.5%
1000 to 139960.7%64.4%
1400 to 179956.3%61.5%
1800 to 219948.4%55.4%
2200 and above42.4%48.8%

Stronger players play 1.e4 less in both speeds, and bullet players at every band play it less than blitz players. At 2200 and above, 1.d4 takes 28.3% in bullet and 31.5% in blitz, and 1.Nf3 is 8.6% and 8.4%. Below 1000, 1.e4 is nearly two thirds of games in either speed.

Black's replies follow the pattern: under 1000, 50.1% of bullet players answer 1.e4 with 1...e5, while at 2200 and above the most common reply is 1...c5, at 24.0%. If you want to understand what these openings are for, our guide to learning plans instead of memorizing openings is a better start than a fast game, and our look at the best first move in chess shows how little the first move decides.

Is bullet chess good for you

We found disagreement among named people, and no study. Here is who said what.

  • Against. GM Avetik Grigoryan, in a ChessMood article we read in November 2026, says "Bullet isn't just useless, but also harmful." His reasoning borrows from Kahneman's Thinking, Fast and Slow: true learning needs slow, deliberate thought, and bullet gives you only the fast kind. He cites no chess study, so this is an expert's opinion, not evidence. He is more relaxed about blitz for stronger players and says beginners should avoid short controls.
  • For blitz, not bullet. NM Jeremy Kane of Chess.com argues blitz is the best form of training. He is speaking about blitz, and we do not attribute that to bullet.
  • Community posts. One Lichess blogger with about 15,000 one minute games says bullet is fun but not real training, because the games are rarely reviewed. Others say it trains quick reactions and calm under time pressure, while admitting that many games are decided by flags and premoves. A Chess.com member blog by Pawn_to_Glory1 says to stick with longer controls to improve. These are anonymous opinions.

What our data adds is narrow. It shows that bullet is a different game from slower chess: half of games end on the clock, and a player under 1000 who flags is nearly as likely to have been winning, 43.1%, as losing, 44.8%. It does not show what happens to a player who plays a lot of bullet over months, because we compared different games and different people on one day, not the same players over time.

Our honest position is that bullet is fine as entertainment and as a speed exercise, and a poor use of your only improvement time. If you play a lot of it, you can get more from it by reviewing your games. Our guide on how to get better at chess places fast games inside a training week.

How to get better at bullet chess

These follow from the data, plus the folk advice we found, which we mark as such.

  1. Take the increment. If you want to win by playing chess and not by clock, 2+1 is the easier control. 56.0% of 1+0 games end on the flag, against 37.0% in 2+1, and the difference shows at every band we compared.
  2. Convert a won position simply. Between 27.3% and 43.1% of players who lost on time were ahead by two pawns or more. Folk advice, not tested in our data: when you are up material, trade pieces and avoid long calculations. A queen up with 8 seconds left is a flag problem, not a chess problem.
  3. Know the draw rule. If you flag against an opponent who cannot mate you, the game is a draw. If your opponent flags against your bare king and you cannot mate, so is theirs. The rule is rare in practice, though, 2.3% of flagged games at 2200 and above.
  4. Use a short, known opening. The first move changes with rating and speed in our data, and strong players spread across more first moves. Community guides say to memorize your lines so moves five to ten cost no clock. That is folk advice, not expert consensus.
  5. Premove only when the reply is forced. Community guides say to premove when you are sure what comes next, mostly with under 5 to 10 seconds left. A forced recapture is the safest case. A premove on a position that can change is how pieces get lost.
  6. Review three games a week. Pick losses where you were winning at some point, and find the move where the evaluation turned. Our free chess game review guide shows what a useful review looks like.

Do the best players play bullet

Yes. Chess.com holds a Bullet Chess Championship each year at 1+0. In the 2026 edition, played from 24 to 27 June with a $50,000 prize fund and $10,000 to the winner, GM Nihal Sarin beat defending champion Alireza Firouzja, 15.5 to 7.5 in the Grand Final and then 9.5 to 4.5 in the Grand Final Reset. Chess.com's coverage quotes Nihal saying "I definitely did not think I would beat him, and I just wanted to try to fight."

Magnus Carlsen has played too: Chess.com's headlines have him losing the 2023 Grand Final to Hikaru Nakamura. Chess.com's 2026 coverage does not mention him, and we took the older results from headlines only, so we leave out their details.

Where to play bullet, and where Chessdrive fits

Lichess and Chess.com are the places to play bullet. They have large pools and separate ratings for each speed, and Chessdrive is an improvement and analysis app, not a bullet server. Our guide to playing chess online for free compares the main options.

Chessdrive fits after the game. A Stockfish review checks every move and marks the turning points, so you can find the games you flagged from a won position and see where the lead started to slip.

Method and sources

Data: Lichess open database, standard rated games, October 2026 file (CC0), the first 1,953,110 games in the file, all from 1 October 2026, streamed with no local copy. Time controls use Lichess's classification, base plus 40 times the increment, with UltraBullet dropped. Rating bands are under 1000, 1000 to 1399, 1400 to 1799, 1800 to 2199 and 2200 and above, using the average of both players' ratings. Games with a rating gap over 400 were skipped. Time forfeit comes from the Termination header and checkmate from the last move. Flagged while winning uses the last available evaluation from the flagged side, with plus or minus 200 centipawns as the cut. Definitions: FIDE Laws of Chess Appendix B, the Lichess FAQ and tournament help, and the Chess.com bullet terms page. Opinions: ChessMood, Chess.com and Lichess community posts. Championship results: Chess.com news. All numbers in the tables come from our run of that file.

Frequently asked

What is bullet chess?
Bullet chess is chess played on the shortest standard clocks. Chess.com defines it as time controls faster than 3 minutes per player. Lichess counts the starting seconds plus 40 times the increment and calls the game bullet if that total is 179 seconds or less, so 1+0 and 2+1 are bullet while 2+2 and 3+0 are blitz.
How is bullet chess different from blitz?
The clock decides how the games end. In our Lichess data 52.4% of bullet games end on time, and the share falls from 64.3% under 1000 to 44.7% at 2200 and above. Blitz has its own table in our blitz guide.
Is bullet chess good for you?
We found no study showing it either way. GM Avetik Grigoryan of ChessMood argues bullet is harmful for improvement, while community posts say it trains quick reactions and composure under time pressure. Our data describes how the games go, not what they do to a player over months.
How many bullet games end on time?
Just over half. Across all bullet games in our sample 52.4% ended with a time forfeit, 22.1% with checkmate and 25.5% with a resignation or another result. In 1+0 games the time share is 56.0%, against 37.0% in 2+1.
What happens if you run out of time with only a king left to beat?
The game is drawn. Both FIDE and Lichess say a flag fall is a loss unless the opponent could not checkmate you by any series of legal moves. At 2200 and above, 2.3% of the bullet games that ended by flag in our evaluated sample were drawn on time.
How do you get better at bullet chess?
Play with an increment when you can, because 2+1 games in our data ended on the flag far less often than 1+0 games. Look for a fast way to finish a winning position, because 27.3% to 43.1% of flagged losers were up two pawns or more. Review your games afterward, because that is where the learning happens.

Play bullet where it is fast, review it where it is slow

Chessdrive is not a bullet server, so play your one minute games on Lichess or Chess.com. When you want to know which moves lost them, a Stockfish review in Chessdrive checks every move and explains the ones that mattered at your level. On iOS and Android, $5.99/month or $39.99/year on the US storefront with a free trial. Prices vary by region.

Keep reading

All guides

Updated