Advice please

[Werewolf]

I'm looking for a way to compare the strength of human players from the distant past with players today. Some of those players may have lived long before the Elo system.

I'm looking for software that can analyze their games for accuracy and produce some kind of score at the end. I know Chessbase has a feature called "Centipawn Analysis". However, I don't think it can analyze games in batches and also the score at the end would need interpreting.

Any advice?
Thanks, Carl.

[gordonr]

I'm looking for a way to compare the strength of human players from the distant past with players today. Some of those players may have lived long before the Elo system.

I'm looking for software that can analyze their games for accuracy and produce some kind of score at the end. I know Chessbase has a feature called "Centipawn Analysis". However, I don't think it can analyze games in batches and also the score at the end would need interpreting.

Any advice?
Thanks, Carl.

I'm wondering if analzing games for accuracy will be a good way of comparing strength from the distant past. For example, what Tal done at his peak was best for that time but maybe wouldn't work so effectively nowadays. There are players who play more accurately than others but don't create enough complications, etc for their opponent. We see a similar effect in computer chess - some engines are very accurate, solid and drawish. Some play less accurately but may win more, especially against weaker opponents. So I don't think "accuracy" and "strength" are exactly the same.

[towforce]

Seems likely that the further you go back in history, the weaker the top players will be (accuracy, openings etc).

Some strong players have a knack of looking at a game and guessing the elo of the players with reasonable accuracy. The danger is that they will actually recognise the classic games of the old masters.

IMO one surprise for the old masters would be the number of women able to beat them!

[Werewolf]

I agree with both the posts above, but I need to start somewhere. Can you recommend any software?

[Ferdy]

I'm looking for a way to compare the strength of human players from the distant past with players today. Some of those players may have lived long before the Elo system.

I'm looking for software that can analyze their games for accuracy and produce some kind of score at the end. I know Chessbase has a feature called "Centipawn Analysis". However, I don't think it can analyze games in batches and also the score at the end would need interpreting.

Any advice?
Thanks, Carl.

It is not that difficult to create a python script to analyze games in parallel. Indicate how you would like the output to look like. Perhaps save it to a database like sqlite so that it is easier to query for what criteria you have in mind to compare the players. I might create such analysis tool or other members on this forum if the instruction of the output is clear. You need to have a criteria in advance so that the output can capture it.

Be aware that the strength of a player depends on the strength of his opponents. The best approach to this is to find common opponents for the player you want to compare. Gathering games is a challenge. You can use ordo to detect connectivity issues.

Aside from calculating centipawn loss, one criteria is to compare which player commits mistake first. In the eyes of the engine, the player that commits mistake first is weaker. If player1 commits suboptimal move at move 25 and player2 commits suboptimal move at move 30, player2 is better than player1 in this case. Collect some games and get the average and other measures. To increase criteria, you can specify suboptimal1 is [25 to 50] cp loss, suboptimal2 is [51 to 60] cp loss and so on. Put more weights to higher loss range. Player that commits big loss is weaker.

The other criteria is to compare the ability of a player to punish opponents. If opponent commits mistake, can the player find the best continuation? Does player find the top1, top2, top3, etc. move? You can increase criteria by classifying the loss, if opponent move loss is [25 to 50] cp, how the player exploits this? Can player find the top1, top2, etc move? Or measure the cp loss of his selected move. If move loss is [51 to 60], how the player would reply? If opponent makes a [51 to 60] cp loss, and player1 got 80% top1 reply move while player2 got only 70% top1 reply move, then player1 is better for this criteria.

Another criteria is the rate of top1, top2, top3, topn non-forced move.

There can be other criteria.

Then later you will weight which criteria is more important, criteria1 is 5%, criteria2 is 10%, criterian is x% and so on.