is there a maximum theoretical elo (2024)

#11

#12

It must be 4200 or 4242 Elo

The big question is how far away actual top engines are from perfect chess and how much draws they would score. 1 % is enough for 700 points difference. So perfect chess is 700 points higher than the actual 3500 Elo engines. ;)

#13

If we're talking FIDE Elo, then no there is no bound. The reason is that the Elo system is broken and to fix it FIDE fixed rating differences above 400 to be counted as 400 difference. (so no matter how much more rating than your opponent you have, you would always gain the e.g. 0.5 rating points for beating the weaker player)

If we're talking about a better rating system, like Glicko which is used by lichess, then there would be an upper bound since even against random play the random play will (very very rarely) play perfectly and get a draw. However just how high the bound might be is hard to judge since we humans are so weak that we already struggle to understand computer chess not to mention how perfect play would look like.

#14

Edgy atheists can't even stay on topic for a second and humor the idea of God for an entirely hypothetical thought experiment. Bravo, you can be proud of yourselves.

Rating is relative as is with all things measurable. It is established by comparison. If we take infinite prowess (that the right word?), then the performance and therefore rating is infinite and immeasurable. However, it can only be measured as relative to anything else.

I'm not sure if we can put a number on perfection, it would be a very abstract act.

#15

#16

There is definitely a perfect game or games in chess. That means that there is no way to play better than that or win opponent that plays on this level. Because there is a finite number of these games, the maximum theoretical ELO is also a finite number.

An interesting question is, is a game already lost, won, or drawn before the first move. If it is not drawn, it will cause significant ELO-fluctuation near the maximum ELO which is kinda hilarious.

#17

A god might have the ability to know its opponent and aim specifically for positions where it will make a mistake. An even more powerful one could directly manipulate its opponents into making mistakes.

A god that knows its opponents (or can manipulate them) will have a far higher rating than one whose godly powers are limited to making good moves.

I suspect most people who ask this question are really getting at something like "How much more improvement is possible in engines like Stockfish? How badly would the best possible engine beat the current top?"

That question would depend a bit on hardware and time control. Chess probably has a pretty generous draw margin, so a quite imperfect engine will still draw a lot of games against the best possible one if it is allowed deep thinks. Perhaps it's not possible to improve by 400 points over what we've got now (400 points = scoring about 90%) - winning with black from the start position is especially hard at this level.

However, there is still a ton of performance gain left to be had in performance per unit of energy & time used. Probably we can get way above a 400 point margin over Stockfish playing bullet on a smartphone.

#18

#19

#20

Chess might be solvable but perhaps it is really as deep as the mysteries of the universe. No one can answer some of those questions, and only the brightest of our minds have ever even understood that these questions exist.

What do I wanna say with this? Well we thought Stockfish was top of the crop. Until we saw the alien form of chess that AlphaZero brought onto our screens. My point is that maybe the more intelligent / deep the thinking -- you know kinda like how a hypothetical two-dimensional entity can not perceive the true form of a three-dimensional object.

But eh I'm just theorizing, by the numbers we have chess "figured out", there's some clearly perfect moves that we have discovered so it's not all esoteric mumbo jumbo of course. Just throwing around some ideas to chew on.