3

An engine which, in a given position, finds all the possible legal moves, and then selects one of them randomly (each move has the same probability of being chosen) and plays it on the board.

Its rating would very probably be negative. My guess is: around -1500 Elo.

Maybe someone can program such an engine and then make it play against another engine.

We would need to find an engine that would be weak (but it would still be much stronger than our random-engine) and that would already have a known rating. We would then make them play thousands of bullet games, with an increment so that neither engine can lose on time.

The rating of our random-engine can then be quickly calculated.

R is the rating of the other engine which served as its opponent. W is the winning probability of the random-engine. L is the losing probability of the random-engine. D is the draw probability. S is the score of the random-engine: S = W + D/2

Notes:

  • There is a time increment, and we must assume that both our random-engine and its opponent will never lose on time.
  • No player can resign or offer a draw. The game must be played until checkmate, stalemate, 50-move without capture and Pawn movement, or repetition.
  • No player can be disqualified by making an illegal move or something like that.
  • I am only looking for a very rough approximation, so even if the rating is off by 300 points it would still be good enough.
  • Am I talking about FIDE rating, USCF rating, chess.com rating, ICC rating, or some other Elo rating? It doesn’t matter since we’re only looking for an approximation anyway.
  • We must assume that the opponent of our random-engine does not know that it is playing against an engine which always plays random moves. Otherwise it would probably constantly threaten checkmate and harass the Queen, or just play ultra-safe waiting for our random-engine to blunder some pieces, which would increase its winning chances even more.
Fate
  • 4,786
  • 2
  • 34
  • 68
  • 2
    [What would be the elo of a computer program that plays at random?](http://chess.stackexchange.com/questions/6508) – Stephen Mar 23 '15 at 10:09
  • I doubt you'll find an engine with a known rating, that is able to lose even a single bullet game against your random engine. But you can circumvent this problem by creating a gradient of random engines, i.e. engines with x out of y moves played randomly. With x==0 you'll have a known rating and you can calculate the ratings for x=1…y by letting the engines play each other. – BlindKungFuMaster Mar 23 '15 at 10:14
  • Here's an engine that plays randomly: [Brutus RND](http://www.xs4all.nl/~vermeire/BrutusRND.rar) – Stephen Mar 23 '15 at 10:29
  • If you lose all your games ever, the lowest your rating will go is about 400 points below the rating of your lowest rated opponent. So in order to reach -1500, he'd have to find a -1100 opponent to lose to first. – RemcoGerlich Mar 23 '15 at 20:16
  • On the top it now says "This question already has an answer here". But it's clearly not the case. In the question linked at the top, the answers are completely useless and hopeless: at best they just state trivial and obvious things like "the computer will, in the long run, blunder [...] his king won't be safe and possibly not even castled" ; at worst they state things which are completely wrong like "His Elo will probably be 0 FIDE". – Fate Mar 24 '15 at 09:20
  • I agree with RemcoGerlich. Elo ratings are relative to the pool of players rated, so the lowest possible rating is 400 points lower than the second lowest rated player assuming that player can beat it 100% of the time. – A passerby Sep 29 '15 at 10:03

0 Answers0