im posting this from my blog: http://chessratios.altervista.org/

Surprisingly engines like Stockfish and Houdini (the top 2!), Are not very close in this table ordered by Captures.

Carlsen is 1 point under the other humans.

captures and checks

What do you think about this?


The Following data analysis of chess games where made by me, the Engines games where extracted from de CCRL page in the category of “40 moves in 4 minutes” , also known as “404”.

The analysis was made game by game, and move by move… so the information, I think, is very precise.

All the analyzes and indicators presented here do not mean that one engine or person is better than another, they are only characteristics of their personality.

Captures: This are amount of captures that the engine or player made in proportion to the other moves.

Checks: This are amount of checks that the engine or player made in proportion to the other moves.

Other: Represents all the other moves that are not Captures or Checks.

Rank: The Rank column is the rank in CCRL of the engine at the moment when the data was downloaded.

All calculations are averaged with the number of games of each engine/player.

I add 4 human players to see the difference against the Engines analysis of these factors. And also added a ‘middle strength’ engine name Zotron (Ranked 441), also for the same cause, to see how these factors behave in different Engines.

  • Could you explain the data a bit? That's percentage of all moves that are check/captures? Why don't the numbers add up to 100%? Your blog does not currently work for me. Feb 27, 2017 at 23:43
  • 2
    One explanation could be that human players resign earlier, so humans spend a larger percentage of moves in the opening/midgame where most captures happen. Feb 27, 2017 at 23:47
  • For more information see edit
    – pimi
    Feb 27, 2017 at 23:57
  • @user1583209 I fixed the link for OP.
    – SmallChess
    Feb 28, 2017 at 0:36
  • 1
    "What do you think about this?" In what way is this not asking for opinion based answers? That is not what the forum is for.
    – Brian Towers
    Feb 28, 2017 at 13:50

2 Answers 2


I don't think the human-engine comparison is fair, because the engine data is restricted to 40 moves in 4 minutes, whereas the human data is, as far as I know, not restricted in the same way.

Given this limitation, I will address the possible differences between the human-computer data:

  • Threefold repetition: the computer data has more checks, which may imply that more games end in threefold repetition
  • Endgame bias: three of the human players you've included are famous endgame players, therefore their games are likely to include a lot of captures. Maybe a player like Tal would have fewer captures but more checks?
  • Maintaining the tension: computers are better at maintaining the tension than humans. They just don't have the need to simplify the position with a capture to alleviate the strain of calculating many variations like humans have.
  • " ...are famous endgame players, therefore their games are likely to include a lot of captures..." Note that the data does not list the total number of captures but the "capture percentage" (relative to non-capture moves). I would expect that in endgames there are relatively fewer capture moves. Feb 28, 2017 at 11:57
  • @user1583209: but to get to an endgame, where material is reduced, there must have been a lot of captures. In a 40 move game, if you go to a K+P vs. K endgame, there are 29 captures, so 29/40 = 72.5% of (full) moves are captures.
    – user1108
    Feb 28, 2017 at 12:02
  • I believe (not sure) the data in this case would be calculated as 15/20 since it is for one player only. Anyway, if the game ends at the 40th move you are right (even though the example is a bit unlikely). I was more thinking of situations where you end up in some endgame and keep on playing without pieces being captured for a long time. Feb 28, 2017 at 12:10
  • @user1583209: Good point about ambiguity in the calculation. Another way to make the calculation is to use total half moves in the denominator (80 = 40 white half-moves + 40 black half-moves).
    – user1108
    Feb 28, 2017 at 12:22
  • @Bad_Bishop I like the thing about 'maintaining the tension' ... is true they do not need to simplify positions to get a good score about a future position.
    – pimi
    Feb 28, 2017 at 12:36

I'd be careful drawing conclusions from this data only. Most captures (in %) happen in the opening/mid game, so shorter games will result in a higher percentage in the "captures" column. I would expect this to be the main reason for the data you see in that column (and not Carlsen avoiding capturing pieces). Perhaps you could analyze the average number of moves of the players!?

There can be various reasons for shorter/longer games:

  • Humans resigning/drawing earlier than computers
  • If a player is playing at a much higher level than his opponents (which is not so much the case nowadays anymore, but perhaps was at the time of Kasparov and before), this could result in those players winning games out of the opening.
  • Different style; a more tactical player might have overall shorter games than somebody like Carlsen who likes to slowly build up an advantage
  • Particularly in recent years top players, when facing much weaker opponents, try to avoid exchanging pieces. This is done to keep the position complicated and not necessarily represents best play.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.