Lately I've been wondering what could be maximum Elo achieved by a computer. I mean, we have super-computers, give super computer (1st in the world) chess program and let it play, it will probably pre-calculate first 30 (or maybe 50) moves within a second.

I went Googling and this site offered, kind of answer, particularly this one: https://chess.stackexchange.com/a/5252

Answer shortly was:

Given these assumptions, the machine will max out at about 3351 ELO1, where if it wins, its Elo will increase by less than half a point (+0.49 to 3351.49), which will be discarded (rounded down to 3351).

Okay, so there's 3351 Elo, which currently maxxed out value. Is there any other value that I could absolutely assign to human as well as computers? I don't want to sound cocky, but I'd really like to compare power of my processor to the one of human brain. Of course I know that processors as fast as human brain will come within 40 years and don't exist now. But I mean, I want to measure and compare on how my computer performs compared to champions. In absolute values, so it's not just limited because it's extremely relative. Just... you know, hard to explain, I hope you understand. Is there such absolute measure for human/computer calculation/prediction?

  • 1
    How is the last paragraph connected to the maximal rating value obtainable in a rating system?
    – JiK
    Feb 22 '15 at 16:35

My experience is with the ECF system, which is probably not technically elo but I understand it approximates it. With that understanding, Tony gave the correct answer, but I want to elaborate.

A simple explanation of how the ECF rating works, using elo units: when you beat your opponent, you get his rating + 400 points for that game, if you draw you get his rating and if you lose, his rating - 400. Your rating for a period is calculated by averaging out all the ratings points so obtained over that period.

As you can see, your rating is normally expressed as a function of your opponents' rating. (There may be exceptions like when your opponent has no rating, or a rating very far away from your rating, but that can be safely ignored for this discussion). That means that no matter how good you are actually, you can only be measured in terms of how good you are relative to your opponents.

The implication of this is that if all your opponents' ratings remain static (as it would be between rating set points), your maximum possible rating is the highest opponent's rating plus 400. This can only be obtained if you keep on playing the next best rated player and win every game.

Let's say someone with Magnus Carlsen's ability appears out of nowhere and enters a junior tournament where all his opponents have ratings of +-1400. Low and behold, he wins all his games - but his rating for the tournament will still only be +-1800, because it is calculated based on the strength of his opponents.

Hence, as you can see, there is no absolute maximum theoretical rating, but there is a relative theoretical limit. As your opponents' ratings increase, your theoretical maximum rating also increases, but never above the highest opponent rating + 400.

** Note to technically minded: ECF rating is convertible to elo using a formula established by the ECF as described here. How accurate this is is probably disputable, but according to the latest formula, I should have used 375 points instead of 400 as described above.

  • 1
    Which rating system uses the method you explain in the first paragraph?
    – JiK
    Feb 23 '15 at 14:57
  • That is a fair point - I described the ECF rating system, but it is my understanding that it is a reasonable approximation of popular elo systems within the context of the question. I may be wrong.
    – firtydank
    Feb 24 '15 at 7:30
  • Updated my answer to make it clear that it talks about ECF instead of elo.
    – firtydank
    Feb 24 '15 at 7:55
  • Yes, the links you provided are great and that might be exactly what I'm looking for, but, I tried to Google program which would measure it on my computer. There's formula to change ECF to Elo, but I don't know Elo of my computer nor it's ECF and I can't seem to be find a program that would measure ECF.
    – Anon
    Mar 3 '15 at 5:55

The maximum ELO (computer or otherwise) is 400 points above its highest rated opponent.

  • Good to know! +1 if you can you provide a reference and/or a calculation for this? Feb 22 '15 at 19:35
  • This is a good start for USCF ratings. uschess.org/content/view/7520/393
    – Tony Ennis
    Feb 22 '15 at 20:43
  • "Rating differences that exceed 350 points are figured as 350 points." That actually means there's no limit to how high your rating could get, using that formula.
    – D M
    Mar 14 '18 at 20:40

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.