Are the bot characters’ ELO ratings accurate?

Sort:
p8q
324u wrote:

My elo rating is 1017 and I beat a 2200 Bot, when I checked the analyse it told me that the bot played like a 1400 rated person. No cheats. Conclusion : They rate the bots to high.

That game you played very good for your rating. It happens from time to time that we play a game above our rating, because humans strength varies (unlike engines strength). Also the game is full of exchanges and simplification, which favors the side that's winning after 16... Qd6 blunder.

16... Qd6?? is a blunder i've never seen Noa bot doing. Maybe a glitch that happens once every 500 games.

In a chess club that i visited weeks ago, they have a computer and i've seen Noa bot beating in 10 minutes games consistently to people who are 2400 rating in chess.com.

If you play that bot more than once (i've checked your archive and i see you played it only once), you'll see you beat it once every 300 games, if at all.

p8q
Lb0zOty wrote:

Im 400 in 30 minutes games but i can defeat and did defeat 2000 elo bot

In your archive i see only 3 games vs bots in your lifetime account.

And you lost 1600 bot

p8q
robertleonardo wrote:

Bots are rated way too high when compared to real game. I have knocked off two 2000 elo bots very easily. This translates to about 1/2 the elo in a real game.

Don’t be deceived in thinking as I was that “jeez I am beating bots left and right I am ready to play someone who’s Leo is the same as the bot”— lol you will Lose.

whatever bot elo your beating, cut the number in half—that’s probably where you are in live play.

Check the tournament in Mercosur (Argentina), year 2009, tournament chess engine vs Grandmasters.

"In this tournament HIARCS ran on a HTC Touch HD (ARM/528Mhz) mobile device." (a washer machine compared to any smartphone nowadays). Link:

https://www.hiarcs.com/Games/Mercosur2009/mercosur09.html

"Pocket Fritz 4 based on the new Hiarcs 13.0 engine". Link:

https://en.chessbase.com/post/breakthrough-performance-by-pocket-fritz-4-in-buenos-aires

Hiarcs 13.1 in CCRL list is 2893 rating. Link:

https://computerchess.org.uk/ccrl/4040/rating_list_all.html

So, according to your "cut the number in half" calculations, an engine 1446 rating anihilated the GMs like flies back in 2009, not losing a single game:

That's just one example. There's dozens of tournaments like this one: vs Kramnik, with engines 2800 rating beating world champions like nothing.

A 500 rated player might beat 2000 rating bot once every 5000 games, of course. If he does 100% of the times something is really wrong with software, configuration, etc...

p8q

Engines have the most accurate rating in chess history, they are calibrated in tournaments vs humans, and their strength barely varies 20 points in years, if at all. Humans strength depends on the day.

GMSrianshtheMMXXV428799

I am 600 and I can defeat a 1800

GMSrianshtheMMXXV428799

And once a 2200 with no assistance

Lb0zOty

I have 3 crowns against Li the 2000 rated bot

kingrogor

no

kingrogor

bot 1500elo< is the bad

kingrogor

I am 700 and i can defeat a2215

DeathicX

Yes, they are pretty accurate but it depends on your gameplay aswell!

kingofkings_97

lord no, in bots i ususally play the 1500 one and win more thimes than i lose. In humans id probably never go avove 1000 or less, mostly because most of low elo players just memorize a cheesy tactic from youtube and if you dont know by heart the exact sequence of moves you lose. AI will never use clickbait tactics unless by extreme chance. It prefers to play solid.

warsoaps

I believe the bots adjust their play a bit around you.

number1Magnusfan

im rated 500 and im trying to beat nora right now every bot before i have beaten.

JamalOnFent
p8q wrote:

I got useful information:

I talked on the phone with the president of a chess club that has computers and students often play vs chess.com bots. He said in his experience 1200 bots won 50% of the times vs his students that have 850 rating.

He also has Rodent IV in the computers (he installed it under my recommendation almost a year ago) and according to his experience that engine set at 1200 is 1200 classical and 1400 blitz 5|5, compared to his students rating.

Therefore, looks like komodo (chess.com bots engine) is less accurate (overvalued) than Rodent iv for a rating under 1600. But above 1600 they both match in strength.

There's hundreds of students in his club. So, I trust his figures. Plus he is comparing with his students rating, which is a more accurate comparison, since many of his students don't compete in FIDE tournaments, and others with this low rating didn't compete enough times to get accurate FIDE rating.

Thus, If I had to measure my rating from 1200 ahead, I'd use Rodent IV. No idea under 1200. I should have asked, but i don't care much under that rating.

Learn grammar dude

P1airo
Tbh I don't really know, I defeated 1100, 1300, and 1500 3 crowns, and 1600, 1800, and 2500, 2 crowns. They seem easy
Depending what the bots do