6

I wonder if there is a possibility to calculate/estimate the Elo rating Napoleon would have today by analysing his known games? Is it possible at all to calculate an Elo rating only with some documented games, if so, how many games would you need to give a good estimation? Is there a list of Elo estiminations for historic persons which are known as good Chess Players before Elo rating was used?

I wonder if it is somehow possible to estimate the strength Napoleon would have even it is not an Elo value. If you could theoretically play a game against him today, with all the chess-knowledge and strategies discovered after his death, how good would you need to be to have a chance?

  • 3
    Quite difficult to estimate Napoleon's strength as a general would be incentivised to lose to him - for one thing. – magd Jul 24 '15 at 14:15
  • 1
    The Regan method relies on the strength of the rated player's moves vs. optimum moves calculated by a strong engine. The opponent's strength is irrelevant, at least directly. (Although you can make a case for a player not expending maximum effort vs. a far inferior opponent.) – Roger Krueger Jul 24 '15 at 22:11
  • Perhaps, but we can all play at infinite strength after the moves 1.f3 e5 2.g4 – magd Jul 25 '15 at 5:30
4

There are efforts to calculate the playing strength of pre-Elo players on the basis of their games. See for example this paper by Kenneth Regan.

But if you look at the "intrinsic performance rating" of historic players on page 5, you will find that they vary quite a bit even for the same player at different points in time.

Personally I find it quite unlikely that the performance of Zukertort or Steinitz varied by over 300 points against different players.

To estimate the playing strength of a player who left only very few (and selected at that!) games, these methods seem to be nowhere precise enough.

Edit: I found another paper but this method seems to end up with a rating between 2700 and 2800 no matter what, which doesn't inspire much confidence.

3

The Ratings of Chessplayers Past And Present by Arpad Elo tackles that question. It can be done, Elo does a little of it (before harping about his methodolgy, though, remember when he wrote that he had only an HP calculator to do all the calculations). Others have done it using different methods.

But the exercise itself is rather pointless. An ELO of 2495 today does not mean the same strength as an ELO of 2495 from decades ago (unless you are seriously going to try and make the point that any of the top 100 or so grandmasters of today could easily beat the Booby Fischer from 1963 that went 11-0 in the US championship, or that Wesley So or Fabiano Caruana today could beat a Fischer at the peak of his abilities).

Ratings are useful for comparisons of contemporaries, but the farther apart in time the players get the more meaningless the comparisons are. A rating is a measurement of strength relative to the whole, and even among players of the same year, a reduction in the player pool will change the numbers (though probably not the relative order).

  • One result of the papers I linked was, that the intrinsic performance rating of players of a certain Elo stayed remarkably stable over time. So one can certainly argue that Wesley So or Caruana are on the level of 1970-'72 Fischer and that is without the opening! – BlindKungFuMaster Sep 10 '15 at 9:49
  • One can try to argue anything, including that the world is flat. The foundation of the system is a bell curve. While players near the center of the bell may be steady the extremes (at either end) will extend farther as the population under the curve expands. More FIDE-rated players must therefore result in higher ratings at the top of the pile. Affecting this also is the opposition. When the nearest opponent is -100 from you, your rating is artificially limited. Context rules all. – Arlen Sep 11 '15 at 16:40
  • Playing strength is distributed in a bell curve. That's why Elo is as well. The top players of a larger population are stronger, that's the reason why they are higher rated. Cause and effect. – BlindKungFuMaster Sep 11 '15 at 20:20
  • Not all bell curves are equal. While almost everything in nature follows a rough bell curve distribution, it's never exact. This has implications for player strength, as well as other attributes. Thought experiment: chart the distributions of height on a college basketball team. Iteratively expand it to more and more teams. As you expand the population, you'll eventually "fix" the extreme heights, before you run out of teams. Yet an equation to predict height distribution would not understand such "real-world" limits, hence height "ratings" would exceed real height. – Arlen Sep 15 '15 at 14:09
  • The Elo system doesn't squeeze players ratings into a bell curve as you seem to assume. If you score 75% against opponents of a certain strength, your rating will be 200 points higher. The meaning of this distance is fixed. Given that the average rating is not dependent on the population size, distance from this average rating always means the same. If you are 1000 points stronger than the average, you are stronger than a player who is only 900 points stronger than the average. Population size doesn't influence that fact at all. – BlindKungFuMaster Sep 15 '15 at 14:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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