The length of a chess game

[Lyudmil Tsvetkov]

How long does an average chess game last?

If you look at human databases, it should be about 40-45 moves on average.

If you compare that with latest TCEC, the length rises to some 75-80 moves on average.

Well, if you look closer at TCEC, 1/3 of those moves might be still due to shuffling, either in the middlegame or endgame, but the other 2/3 are definitely due to better understanding of the game and making use of all available resources on the board, as well as tougher defence. So that, objectively, shuffling removed, the average TCEC length should be around 70 moves, still close to double the length of an average human game.

If we assume that with rising quality of computer chess in the future the average length will rise further, it is quite probable that the average length of a chess game by perfect play will reach some 120 moves, or even more.

Well, one of the reasons that human games last shorter is of course the fact that humans get tired at some point, they simply would like to take a break, offer an early draw, etc., but we are not concerned with psychological implications here.

My personal opinion is that, by perfect play, the average chess game should reach some 120 moves and above. So anyone wishing to play a perfect game better prepare to stand the pressure of a 120-mover!

According to you, how long is an average chess game by perfect play?

[bob]

How long does an average chess game last?

If you look at human databases, it should be about 40-45 moves on average.

If you compare that with latest TCEC, the length rises to some 75-80 moves on average.

Well, if you look closer at TCEC, 1/3 of those moves might be still due to shuffling, either in the middlegame or endgame, but the other 2/3 are definitely due to better understanding of the game and making use of all available resources on the board, as well as tougher defence. So that, objectively, shuffling removed, the average TCEC length should be around 70 moves, still close to double the length of an average human game.

If we assume that with rising quality of computer chess in the future the average length will rise further, it is quite probable that the average length of a chess game by perfect play will reach some 120 moves, or even more.

Well, one of the reasons that human games last shorter is of course the fact that humans get tired at some point, they simply would like to take a break, offer an early draw, etc., but we are not concerned with psychological implications here.

My personal opinion is that, by perfect play, the average chess game should reach some 120 moves and above. So anyone wishing to play a perfect game better prepare to stand the pressure of a 120-mover!

According to you, how long is an average chess game by perfect play?

I think you are WAY off. Perfect play by both sides will very likely last beyond 1,000 moves. Maybe all the way to the 5500 or so 50-move-rule limit... If you mean perfect play by one side, who knows.

[Ajedrecista]

Hello Lyudmil:

I calculated the average number of plies in 1-0, 1/2-1/2 and 0-1 of some computer games:

Average number of plies in {1-0, ½-½, 0-1}.

That data showed that in average black needs more moves to win than white. You can also use the tool called joined by Andreas Stable and collect some data from various PGNs.

Regards from Spain.

Ajedrecista.

[Dirt]

Perfect play by both sides will very likely last beyond 1,000 moves. Maybe all the way to the 5500 or so 50-move-rule limit...

What does perfect play even mean? Other than never losing, I'm not sure. Perhaps 10 move draws would be common.

[Lyudmil Tsvetkov]

Hello Lyudmil:

I calculated the average number of plies in 1-0, 1/2-1/2 and 0-1 of some computer games:

Average number of plies in {1-0, ½-½, 0-1}.

That data showed that in average black needs more moves to win than white. You can also use the tool called joined by Andreas Stable and collect some data from various PGNs.

Regards from Spain.

Ajedrecista.

Interesting.

65-70 moves on average, and that was 2 years back in CCRL, with engines of different calibre calculated.

So the length of games seems to be rising with increasing chess strength of the opponents. For comparison, the average length of the recent Carlsen-Anand WC match was 501 moves in 11 games, or very close to 45 moves per game. And in the 122-move game Carlsen also shuffled a lot trying to win R+N vs R.

No one has data for move average for current stage of TCEC (by far the strongest pre-final stage)?

The point is that quality chess means longer games.

[Lyudmil Tsvetkov]

Perfect play by both sides will very likely last beyond 1,000 moves. Maybe all the way to the 5500 or so 50-move-rule limit...

What does perfect play even mean? Other than never losing, I'm not sure. Perhaps 10 move draws would be common.

Perfect play means exploiting the available position resources in the best possible way, both for the attacking player and the defender.
And when the attacking player comes up with subtler plans, and when the defence is more stubborn, the game last longer.

Non-optimal or weak play exploits the available position resources suboptimally or weakly, and therefore the climax comes quicker.

Have you seen games of children starting playing for the first time? Usually the games end very soon after a serious mistake for one side. I guess also the game length in Graham's lower divisions is lower than that in CCRL top leagues.

[Lyudmil Tsvetkov]

How long does an average chess game last?

If you look at human databases, it should be about 40-45 moves on average.

If you compare that with latest TCEC, the length rises to some 75-80 moves on average.

Well, if you look closer at TCEC, 1/3 of those moves might be still due to shuffling, either in the middlegame or endgame, but the other 2/3 are definitely due to better understanding of the game and making use of all available resources on the board, as well as tougher defence. So that, objectively, shuffling removed, the average TCEC length should be around 70 moves, still close to double the length of an average human game.

If we assume that with rising quality of computer chess in the future the average length will rise further, it is quite probable that the average length of a chess game by perfect play will reach some 120 moves, or even more.

Well, one of the reasons that human games last shorter is of course the fact that humans get tired at some point, they simply would like to take a break, offer an early draw, etc., but we are not concerned with psychological implications here.

My personal opinion is that, by perfect play, the average chess game should reach some 120 moves and above. So anyone wishing to play a perfect game better prepare to stand the pressure of a 120-mover!

According to you, how long is an average chess game by perfect play?

I think you are WAY off. Perfect play by both sides will very likely last beyond 1,000 moves. Maybe all the way to the 5500 or so 50-move-rule limit... If you mean perfect play by one side, who knows.

Not that much.

The position resources get exhausted sooner, and the 50-moves rule is useful only in limited instances.

But I bet current development Crafty-last released Crafty version matches lasted a lot shorter years back...

[Lyudmil Tsvetkov]

Many might think that what I have written is very stupid at first glance, but it is quite the opposite.

It is a bout a very basic perception: how long does a chess game last?

Couple of years ago before the astounding improvement in engine strength, people would have said, well, it lasts 40-45 moves on average.
Now, we already know after TCEC and similar events that, with significant strength increase and improved play, a standard chess game at that level lasts some 70-80 moves.

I can bet that in 5 or 10 years, if engines are further perfectionised, a standard game with such engines will last some 100-120 and even more moves on average. That will be the average length of a normal chess game at the top level, and everybody will know that a standard chess game at that level lasts that much.

That is a very basic perception, it changes our historical vision of chess game length and chess in general at least 3-fold. Anything more sensational than that?

[Uri Blass]

Many might think that what I have written is very stupid at first glance, but it is quite the opposite.

It is a bout a very basic perception: how long does a chess game last?

Couple of years ago before the astounding improvement in engine strength, people would have said, well, it lasts 40-45 moves on average.
Now, we already know after TCEC and similar events that, with significant strength increase and improved play, a standard chess game at that level lasts some 70-80 moves.

I can bet that in 5 or 10 years, if engines are further perfectionised, a standard game with such engines will last some 100-120 and even more moves on average. That will be the average length of a normal chess game at the top level, and everybody will know that a standard chess game at that level lasts that much.

That is a very basic perception, it changes our historical vision of chess game length and chess in general at least 3-fold. Anything more sensational than that?

When both players play random moves the average length is more than 120 moves so better quality does not always mean longer games.

[Lyudmil Tsvetkov]

Many might think that what I have written is very stupid at first glance, but it is quite the opposite.

It is a bout a very basic perception: how long does a chess game last?

Couple of years ago before the astounding improvement in engine strength, people would have said, well, it lasts 40-45 moves on average.
Now, we already know after TCEC and similar events that, with significant strength increase and improved play, a standard chess game at that level lasts some 70-80 moves.

I can bet that in 5 or 10 years, if engines are further perfectionised, a standard game with such engines will last some 100-120 and even more moves on average. That will be the average length of a normal chess game at the top level, and everybody will know that a standard chess game at that level lasts that much.

That is a very basic perception, it changes our historical vision of chess game length and chess in general at least 3-fold. Anything more sensational than that?

When both players play random moves the average length is more than 120 moves so better quality does not always mean longer games.

I did not check that, but I also did not mean that.

Can a game be completely random?

I meant of course no shuffling, resigning when there is an evident win, relying on a relentless struggle otherwise.