Playing12 min read

Chess Clock: How It Works, and What Increment Changes in 2.77M Games

A chess clock gives each player a bank of time that only drains on their own turn. We read 2.77 million rated Lichess games from 1 October 2026 to see what an increment changes, how the clock empties by move number, and when time pressure turns into blunders.

The starting position

A chess clock is two timers joined by a switch. Each player starts with the same bank of time, and only your own timer runs on your turn. You make a move, press your side, and your clock stops while your opponent's starts. When a timer reaches zero that player has lost on time, unless the opponent could not have mated anyway. The rest is how time is added back, and we read 2.77 million rated Lichess games from one day to see what that does.

The short answer

The clock decides more games than most players think, and an increment changes which games it decides. At equal base time, 27.7% of 3+0 games in our sample ended on time against 21.6% of 3+2 games. The gap is 11.7 points under 1000 and about 2 points at 2200 and above. At move 20 the median 3+0 player has 67% of the clock left. Under 10% of the base time, the blunder rate runs 1.28 times higher in blitz class games.

Analog and digital clocks

The first clocks were mechanical. Wikipedia credits Thomas Bright Wilson of Manchester Chess Club with the invention, and says it was first used in competition at the London 1883 tournament. An analog clock has a small flag that falls at the moment time expires.

Digital clocks and internet play spurred more varied time controls, including increment and delay. Bobby Fischer filed US patent 4,884,255 in 1988, awarded in 1989, for a new type of digital chess clock, and increment is also called Fischer time.

Increment, Bronstein delay and simple delay

All three exist to stop a won position being lost to the flag. They differ in what happens to the seconds you do not use. Take a 5 minute control with 3 seconds added or delayed and a move that takes you one second.

  • Increment (Fischer). A fixed amount is added to your main time after each move. The move costs one second and pays three, so your clock goes up by two. Unused time is banked.
  • Bronstein delay. If you use more than the delay, the whole delay is added back. If you use less, only the time used is added back. The one second move leaves your clock unchanged, and nothing is banked.
  • Simple delay (US delay). The clock waits out the delay before your main time starts counting down. The one second move costs nothing, and nothing is banked.

Wikipedia treats Bronstein and simple delay as different mechanisms, though one Chess.com forum poster calls them 100% equivalent. At the board the effect is near identical: neither lets you build a reserve. Only increment does.

Chess.com's clock article calls Fischer bonus time common worldwide and never mentions delay.

Chess clock rules: FIDE and US Chess

Most rules agree. Where FIDE and US Chess differ, we say so rather than pick one. We read FIDE's text on its handbook site, but did not open the US Chess rulebook itself: the US Chess points below come from forum and blog quotations of it, so check the current rulebook before you rely on them in a tournament.

  • Press with the hand that moved. FIDE says a player must press the clock with the same hand with which they made the move.
  • Clock readings are conclusive. FIDE says every indication given by the clock is conclusive in the absence of an evident defect.
  • Flag fall. FIDE defines it as the expiration of a player's allotted time, and says a flag is considered fallen when the arbiter observes it. US Chess differs on who raises it: a Lichess blog quoting the rulebook says only a player in the game may point out a fallen flag, and a director must never start a time forfeit claim.
  • Time out against a bare king. FIDE draws the game if the position is such that the opponent cannot checkmate your king by any series of legal moves. US Chess wording as quoted on a forum says the player pointing out the flag wins provided they have mating material. Forum posters dispute exactly when the draw applies, so ask the director.

We found no sourced rule for forgetting to press the clock, so we do not state one. Ask the arbiter.

How to set common time controls

A control written 5+3 means five minutes of main time and a three second increment. Delay controls are written differently by clock and event, so check the event rules.

What you call the control differs by site. Lichess adds the base in seconds to 40 times the increment in seconds and calls the total bullet up to 179, blitz up to 479 and rapid up to 1499. FIDE counts a game as blitz if the allotted time plus 60 times any increment is 10 minutes or less per player, and rapid if that figure is more than 10 and less than 60 minutes. Chess.com's article puts bullet under three minutes and blitz at three to ten.

By that arithmetic, 3+0 (180 seconds), 3+2 (260), 5+0 (300) and 5+3 (420) are Lichess blitz, while 10+0 (600) and 10+5 (800) are rapid. So 10+0 is blitz by FIDE's wording (600 seconds) and rapid on Lichess: name the platform when you compare. Our blitz chess guide and bullet chess guide cover those two speeds in detail.

Online clocks, apps and over the board

Online, the server keeps the time. As of November 2026, Chess.com's clock article says it forgives up to one full second of lag and abandons a game if a player is disconnected for 10% of the estimated total clock time. Lichess has a button that adds 15 seconds to the opponent's clock, which a forum describes as hideable in preferences. We found no official Lichess over the board clock, and its OTB mode post was an April Fools joke.

Over the board, as of November 2026, retailers call DGT the standard make, and DGT describes its 2010 model as an official FIDE clock. For home use, check that a clock offers increment and delay. Chess.com's free iOS Chess Clock app supports Fischer, Bronstein and simple delay. Our guide to playing chess online for free compares the main sites.

What we measured

We streamed the Lichess open database file for October 2026 (CC0 license) and read its first 3,000,000 games, all played on 1 October 2026. After dropping BOT games, games with a rating gap over 400 between the players, and controls outside the nine we studied, 2,768,350 rated games remained. Bands use the average of both players' Lichess ratings. The nine controls were 1+0, 2+1, 3+0, 3+2, 5+0, 5+3, 10+0, 10+5 and 15+10.

Two limits matter. Lichess clocks are online clocks, and over the board play may differ. The blunder figures come only from the 281,213 games where someone asked for computer evaluations, a self selected subset.

What an increment does at equal base time

The cleanest test is the same base time with and without an increment. Time forfeit share is over all games in each cell.

PairGamesTime forfeit, no incrementTime forfeit, with incrementGap
3+0 vs 3+2563,360 and 264,23127.7%21.6%6.0 points
5+0 vs 5+3370,171 and 146,73621.2%18.1%3.1 points
10+0 vs 10+5368,790 and 70,46911.4%10.3%1.1 points

Pooled across bands, the increment helps in every pair, and less as the base grows. By rating band, the gap in points between the no increment and increment control:

Average rating3+0 vs 3+25+0 vs 5+310+0 vs 10+5
Under 100011.75.82.8
1000 to 13998.23.21.6
1400 to 17995.92.60.9
1800 to 21995.02.1no gap
2200 and above1.92.62.7

Under 1000, 34.7% of 3+0 games ended on time against 23.0% of 3+2 games, an 11.7 point gap on 17,530 and 12,572 games. At 1800 to 2199 in the 10 minute pair the two controls match, 11.5% against 11.6%. The 2200 and above cells for 5+3 and 10+5 rest on 609 and 599 games, so read their gaps as rough. The 3+0 and 3+2 cells there have 41,375 and 13,986.

This is not a controlled experiment. The two controls draw different players, and the same player may behave differently in each. What it supports is narrower: if flagging is how you lose, choosing the increment control removes more of it the weaker you are.

How the clock empties by move number

For each control we took the clock after a player's own move 10, 20, 30 and 40, divided by the base time, and took the median. Both colours are pooled, and only players in games that reach that move count. Increment games can exceed 100%, because added seconds count.

ControlMove 10Move 20Move 30Move 40
3+091%67%42%24%
3+296%68%39%23%
5+091%70%48%32%
5+395%70%45%30%
10+093%77%60%47%
10+596%79%61%50%

Medians are read from bins half a point wide, so treat them as accurate to about a point. Move 10 counts between 132,236 and 1,090,447 players per control, and move 40 between 39,747 and 393,430.

Look at move 40 in 3+2: the median player has less of the starting time, 23%, than in 3+0, 24%, even though they gain two seconds a move. By band, it comes from players at 1800 and above:

Average rating3+0 at move 403+2 at move 40
Under 100020%28%
1000 to 139924%27%
1400 to 179927%26%
1800 to 219924%20%
2200 and above21%13%

Below 1400 the increment leaves players with more time at move 40, as you would expect. At 2200 and above it leaves them with less, 13% against 21%. A median over survivors is not a median over games. In 3+0 a slow player flags before move 40 and drops out of the count, so only players who kept pace remain. In 3+2 the increment keeps slower players alive to reach move 40, and they arrive short. Strong players may also simply spend the increment. The data cannot separate these, and it does not show that the increment costs you time.

A benchmark for your own games

The 3+0 median at move 20 by band is 54% under 1000, 63% at 1000 to 1399, 68% at 1400 to 1799, 69% at 1800 to 2199 and 67% at 2200 and above. A player under 1000 is about 15 points of base time behind an 1800 player after only 20 moves, and the median 3+0 game lasts 33 full moves. Reaching move 20 with well under two thirds left means you are on the slow side.

How often games reach real time pressure

Here we counted games where at least one player's recorded clock ever fell under 10% of the base time. That is 18 seconds in a 3+0 game and a minute in a 10+0 game.

ControlUnder 10001000 to 13991400 to 17991800 to 21992200 and aboveAll
3+038.8%34.5%34.3%40.1%44.7%37.3%
3+230.6%31.9%36.2%45.8%57.7%38.7%
5+026.0%24.8%27.0%31.9%39.7%27.5%
5+324.2%27.3%32.8%42.2%52.4%31.3%
10+08.9%9.9%12.3%18.7%30.2%12.7%
10+57.1%11.8%18.6%29.7%54.9%16.8%

Two readings. First, the share is highest at 2200 and above in every row, even though stronger players lose on time less often. Strong players reach the last 10% often and survive it. At 2200 and above, 54.9% of 10+5 games saw a clock below 10% of the base, on 599 games, against 30.2% of 10+0 games on 4,167. Second, the increment does not mean you stay out of trouble. In every control from 1400 up, the increment version reached time pressure more often than the no increment version. With a bonus on every move, players seem willing to run the clock down and trust it to refill.

Time trouble and blunders

A blunder in our data is a move where the mover's winning chance, on Lichess's win percentage scale, drops by at least 20 points. We compared the blunder rate on moves played with under 10% of the base time left against all other moves, by Lichess speed class. Blitz and bullet tables by band are in our blitz chess guide and bullet chess guide, so here we only report the pattern.

  • Pooled across bands: blitz class moves with low clock blundered 7.9 times per 100 moves against 6.2 otherwise, a ratio of 1.28. In bullet the ratio is 1.40, and in rapid class games 1.07.
  • The ratio grows with rating in all three classes. In blitz it runs from 1.09 under 1000 to 1.88 at 2200 and above. In bullet it runs from 1.22 to 1.93, and in rapid from 0.98 to 1.70.

That second line is the one to remember. Weaker players blunder a lot all game, so time pressure adds relatively little: under 1000 in rapid there is no penalty, 0.98. A strong player is accurate with time and loses that when it goes. In blitz at 2200 and above the rate is 6.5 per 100 moves on low clock against 3.5 otherwise.

What to do with this

  1. Pick the increment if you flag a lot. At 3+0 against 3+2, players under 1000 lost on time 11.7 points less often with the increment. It is the cheapest fix on offer, if flagging is how you lose. At 1800 to 2199 on 10 minutes, our data shows no gap for the 5 second increment.
  2. Check yourself at move 20. In 3+0 the median player has 67% left. If you are well under, you are paying in the opening for moves you could have played from memory. Our guide to learning plans instead of memorizing openings shows what to spend the time on instead.
  3. Bank time early, spend it when the position changes. The penalty for low clock grows with rating, so stronger players have the most to lose from arriving at move 30 short.
  4. Treat the last 10% as a different game. 37.3% of 3+0 games reach it, so practise simple conversions before you need them.
  5. Review where your clock went. Our guide on how to get better at chess places review inside a training week, and the free chess game review guide shows what a useful review looks like.

Where Chessdrive fits

Chessdrive is not a clock or a game server. Afterward, a Stockfish review in Chessdrive marks each turning point, so you can see whether the position was lost before your clock ran low or only after. Ratings here are Lichess ratings, which differ from Chess.com and FIDE, as our guide to the chess rating system explains.

Method and sources

Data: Lichess open database, standard rated games, October 2026 file (CC0), the first 3,000,000 games in the file, all from 1 October 2026, streamed with no local copy. We counted 2,768,350 after dropping 177,165 games in other controls, 7,296 with a BOT, 44,331 with a rating gap over 400 and 2,858 with bad headers. Blunders use evaluated games only and skip the first 10 plies. Rules: FIDE handbook, Wikipedia, Chess.com. US Chess points come from a Lichess blog and the US Chess forum, not the rulebook.

Frequently asked

How does a chess clock work?
It is two timers joined by a switch. Each player starts with the same bank of time. You make your move, press your side of the clock, and your timer stops while your opponent's starts. A player whose time reaches zero loses on time, unless the opponent could not have checkmated by any series of legal moves, in which case FIDE calls it a draw.
What is the difference between increment and delay?
Increment adds a fixed amount to your clock after every move, so unused seconds are banked. Simple delay makes the clock wait before it starts counting down, and unused delay is lost. Bronstein delay adds back the time you actually used, up to the delay amount. FIDE recognises increment and delay modes.
Does an increment really reduce losses on time?
In our Lichess sample it did at equal base time. 21.6% of 3+2 games ended on time against 27.7% of 3+0 games, and 18.1% of 5+3 games against 21.2% of 5+0 games. The gap is largest below 1000 and mostly shrinks as rating rises. The controls attract different players, so this is not a controlled experiment.
How much time should I have left at move 20?
In our 3+0 games the median player who reached move 20 had 67% of the starting time left, which is about two minutes. Under 1000 it was 54%, and at 1800 to 2199 it was 69%. A big gap below that benchmark means you are spending your clock in the opening.
Are blunders more likely in time trouble?
Yes, and more so for stronger players. In our blitz class games, the blunder rate with under 10% of the base time left was 1.09 times the normal rate under 1000 and 1.88 times at 2200 and above. These come from the roughly 281,000 games with engine evaluations.
What is a good chess clock to buy?
As of November 2026, retailers call DGT the standard make, and DGT describes its DGT 2010 as an official FIDE clock. For casual play any digital clock that supports increment and delay will do. Chess.com also publishes a free iOS Chess Clock app with Fischer, Bronstein and simple delay.

Play on the clock, review it off the clock

Chessdrive is not a clock or a game server, so play your timed games on Lichess, on Chess.com or over the board. When you want to know which moves cost you the game, 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