Forums

USCF/ELO Strength of Bots?

Sort:
LeoRostov

After decades away from OTB, USCF rated tournament chess, I hope to return in the next few months.  It would be really helpful to have some reliable estimate of the damage to my old USCF strength (left USCF at 1746) before returning to OTB.  My re-training plan will start with maybe 40-50 matches vs. the bots (along with opening, middle and endgame study + online matches).  Does anyone have an estimate for the bots' playing strength, ie., USCF or ELO strength?  In a recent match with PAL 9000 (Chess.com rating at 1200), the bot played a solid opening about 10 moves deep.....then hung a piece.  Then another 10 solid moves.....then overlooked a simple two move trap.  A few more decent moves, then hung another piece!  A 1200 USCF player will make multiple weak moves, even a few blunders.  But hanging 2-3 pieces?  No.  Maybe a 600-800 USCF player would do that and would add chess etiquette....namely to resign!  Matches like that one, are, I think, of limited usefulness, not even a boost for ego!  Specifically, what I am looking for is:  estimates of USCF ratings for Chess.com bots rated by Chess.com at 1600, 1800 and 2000.  Has anyone with a USCF rating, played enough matches with the bots at those levels to have some idea as to their actual USCF strength?  Any help on this.....very much appreciated.  (I have the same question for the Fritz 18 bots, which I may purchase).

  

magipi

So your plan is to play a match against some bots (okay), but first you want to know their exact ratings because... and at this point my imagination has failed, I can't figure out a reason.

LeoRostov

Not exact ratings, did not ask for that.  Hard to know if one is making progress if bot ratings are off by 100-400 points compared to real world USCF ratings, as indicated by some Forum posts.  Beating up on bots with ego booster ratings does not forecast at all, how one might do against USCF rated humans, which, I would like to know in advance if possible.  If I had some general idea, not exact, of my current USCF rating strength, I would know when to register for an OTC tournament. If I had some confidence that my current USCF rating were about 1600 or so, probably I would register.  At less, probably not.....need more training.  Not interested in playing a Class B OTB tournament and scoring 1/2 - 4-1/2!  Disastrous for my old USCF rating of 1746. 

Elon_Crust123

I think most of the time chess.com's bots have higher ratings than they should compared to normal chess.com ratings. As a USCF player myself, I believe around 2000 ELO for a bot could come to roughly 1600-1700 USCF, an 1800 bot 1300-1599, and a 1600 bot 1000-1299. The bots are relatively inconsistent with their level of play, but I believe this estimation is fairly accurate. I would say if you were 1746 USCF when you stopped playing OTB, assuming you are still at that level, then you should be able to beat bots around 2000-2400. Of course this all depends on the "personality" of the computer. If you really want to dig deeper into this question, I would suggest you play against the engine at different levels. Hope this helps!

basketstorm

Chess.com bots use Komodo and run locally using resources of your PC or smartphone

There might time limit on bot's thinking, so make sure to use a powerful PC without too much programs in background to get a consistent level.

Rating is calibrated to 5min + 5sec.

Winning against a bot 1 time means a little, doesn't mean your Elo is higher or something. Play with time control and play enough times and fairly count all losses then calculate your rating difference with the bot according to the table.

Quote:

Our intention is that for quick games (10 min, or 5 min + 5 sec increment) even a virtual beginner (say Elo 250) should be fairly matched with level 1, while the World Champion should have a tough time with level 23 (est. 3000 at 5’ + 5”). (Level + 1) * 125 would be a good estimate based on chess.com games. At level 24 we estimate that the World Champion would need a time limit like 15’ + 10” to be competitive. If the human is treating the games like serious tournament games (say 90 min + 30 sec) then the corresponding elo values would be a few hundred lower.