Alexandra Botez Bot seems a lot stronger than 2100.

I played a couple of games with the Alex Botez bot yesterday (one as white, one as black) where the time control for each game was 10 minutes and the CAPS accuracies in both of her games against me were 95%-96% which suggests a strength of roughly 2500 (400 points higher than what chess.com claims) and the bot only spent like 2 seconds per move.

For those of you who have gotten a chance to play with the Alex Botez bot, did the bot seem to play around 2100 strength when you played against her, weaker than 2100, or stronger than 2100?

I think you might have posted this in the wrong thread by accident.

I’ve not played either Botez bot, nor am I near your skill, but keep in mind that many bots are inaccurately rated, due largely to the fact that computers simply don’t play like humans.

Antonio bot drops pawns more than a 1200, but can see mate in 7 instantly.

Isabel bot sometimes plays the endgame like a GM, and sometimes I outclass it in the endgame.

These bots can play inhumanly perfectly, but they are programmed to sometimes blunder (often times in very telegraphed and unnatural ways) so as to separate them from full power engines such as stockfish.

The Botez bot likely is tactically sharper than the real Alexandra, bots seem to get tactical sequences down perfectly.

As opposed to humans, who will sometimes get the move order wrong, or will begin to panic under time pressure.

https://www.chess.com/game/computer/28143361

I missed the winning move but stronger then 2100? Not in this game

But we don't care about accuacies, it's the same algorithm as the maximum level of the same program etc. what is important to look at are the positions and errors.

example :

here 3...Qc6 is a blunder and even if after bot plays at 100% it doesn't mean that this bot is a gosu because he plays at 100% in a losing position.

Now if you don't play 4) Bb5 according to its handicap given by the programmers to lower the maximum level, the bot will still make more or less mistakes but if you don't punish it at the end it will be much stronger than a human player displayed at this level because its algorithm remains stronger than a human displayed at this level.   

Alexandra Botez played her own bot and it beat her up!
https://www.youtube.com/watch?v=xKZpfN4q404

Xd

They're always changing bots a bit, so it might be easier now than it was two years ago when this topic was made, but this game was pretty easy. I chose 10 minutes.

The opening was maybe 1500 strength (?) and then it gave away a pawn for no reason, followed by a rook, so, lower than 1500.

And that's website rating. In terms of FIDE it was like... a 1000 rating lol (I'm guessing if I played it 10 more times it would do better in at least some of the games).

 

Sort of, yeah. It's not only best moves and blunders though, they'll mix in some inaccurate moves.

But it's true one moment chess.com bots will blunder material for no reason, and the next moment they will dodge a tactic I spent 5 minutes calculating. It can be frustrating.