Are Bot's ratings accurate?

landrew_JR

I love playing bots. My dad does the same

 It allows you to get back in practice if you have been out for awhile. My dad loves playing anna Rudolf. He rarely wins but has fun. We both find the bots fairly accurate. Especially in advanced and master levels. 

Theebigcheese
Yea I played a chess bot at my level and I felt like I was playing someone who had never played before
JazzyBellekitty
You said you looked at blitz look at daily chess rating
Metal_Pineapple

I think that all bots will never be correct, but they should be in the general area. I also played 1600 Devon, Dash, and other bots that supposedly are good. I don't have a crazy rating or anything so I was also a little surprised when I defeated Dash the Halloween bot first try. I also checkmated Pierre in 7 moves because I played 1. e4 f5 2. Bd3 fxe4 3. Bxe4 Na6 4. Qh5+ g6 5. Qxg6 hxg6 7. Bxg6# and won. I call this the Pierre F-Pawn Issue. I kept doing this mate until chess.com fixed it which was in a couple weeks. I think that chess.com should change the bots rating a little bit and try to fix issues like the Pierre F-Pawn Issue. Thanks for taking the time to read this!

- Metal_Pineapple

barbashka
Vibhansh_Alok wrote:

My chess.com rating keeps circulating around 1400(blitz).... on the other site it is around 1700. Whatever that's not the matter, Among the newly arrived chess bots... I tried a 1600 rated guy and got the victory, first I thought I was lucky, so I tried many times and won. If Robots are accurately rated, that means I should be a 1600+ rated player. If not then what are the excuse for keeping fake rated bots

Here's one of the game...

 

no they are not accurate

abigoda

I gotta tell you guys, i had a tough time beating 1300 Nelson...one thing i can tell is that at the assisted match this specific BOT made some ridiculous mistakes, and when i tried unassisted mode he was not fooling around. So my conclusion is that you can not use the assisted match mode to try to improve your game. sad.png

ReadyToMate69
Metal_Pineapple wrote:

I think that all bots will never be correct, but they should be in the general area. I also played 1600 Devon, Dash, and other bots that supposedly are good. I don't have a crazy rating or anything so I was also a little surprised when I defeated Dash the Halloween bot first try. I also checkmated Pierre in 7 moves because I played 1. e4 f5 2. Bd3 fxe4 3. Bxe4 Na6 4. Qh5+ g6 5. Qxg6 hxg6 7. Bxg6# and won. I call this the Pierre F-Pawn Issue. I kept doing this mate until chess.com fixed it which was in a couple weeks. I think that chess.com should change the bots rating a little bit and try to fix issues like the Pierre F-Pawn Issue. Thanks for taking the time to read this!

- Metal_Pineapple

 

I am new to the game with a rating only 950 and also found Dash (1600) weirdly easy. I seem to be able to beat Dash but struggle against David (1400) despite the latter supposedly being 200 rating points lower. Maybe the Dash bot is overrated.

jonnin

No.  

I have beaten the bots at the ~1800 level multiple times, and me below 1600.  Not just beth, but the christmas one and the wally fellow etc.   100 points too high, maybe 150?  Could also be nonlinear...   its hard to tell.  I destroy the next level down from 1800 easily, so those are far weaker than I am... 1400 or less. 

 

Computers have always struggled very hard to produce human-like games at lower levels.  Its very difficult to emulate a human's mistakes.  They generally use a mix of reduced depth (can only see 3-5 moves ahead or so) and randomization (pick from the top 3-5 best moves available randomly) and in some cases monkeying with the piece scores (make bishop worth 4 and knight 3, for a flavor) or material vs position preferences to emulate.  I used to play one called 'tal style' that made semi-sound sacrifices to wipe you out quickly (I think the 2000 normal bot plays like this!). 

WolfPaul
Vibhansh_Alok wrote:

My chess.com rating keeps circulating around 1400(blitz).... on the other site it is around 1700. Whatever that's not the matter, Among the newly arrived chess bots... I tried a 1600 rated guy and got the victory, first I thought I was lucky, so I tried many times and won. If Robots are accurately rated, that means I should be a 1600+ rated player. If not then what are the excuse for keeping fake rated bots?

Here's one of the game...

Yes, I think they are overrated a bit or insanely inconsistent.

I am around 1170, and I just casually defeated Beth Harmon at 1600, although I did get +2 pawn advantage by playing some risky moves early on and then rushed to the end game to rely on that advantage, I noticed, she was able to trick me a couple of times and would have easily equalized the material but did not for some reason, even towards the end game, he could have promoted a passed pawn, but wasted the chance by making a random move and allowed me to promote first.

So, yeah, I guess, tactically they are somewhat accurate since I did notice some really good middle game tactics but did not finish the job by playing strange.

 

S1Daniel

In my opinion every bot does respond to the same move with the same move except the openings, I mean you can try that if you once defeated it, and you premove all the moves you still win if the bot starts with the same opening again. So if you try to defeat a bot with the same opening always and win that wont mean you are that good, but if you start with many openings, you clearly have the rating they have. Thats my opinion, hope it helps.

barbashka

Yeah

Epiloque
S1Daniel wrote:

In my opinion every bot does respond to the same move with the same move except the openings, I mean you can try that if you once defeated it, and you premove all the moves you still win if the bot starts with the same opening again. So if you try to defeat a bot with the same opening always and win that wont mean you are that good, but if you start with many openings, you clearly have the rating they have. Thats my opinion, hope it helps.

yes that is called cheesing, and you can set it to 1|0 and do that for easy wins

barbashka

Hmmm

JackRoach

I don't believe so. Many bots rated 250-1400ish randomly blunder material. This is because it is hard to simulate how humans blunder if you think about it. So bot blunders I think are a lot more obvious. I don't really know what I am saying, but hopefully this makes more sense you all than it does to me lol.

kifyfey-shady

i don't know if thier rating is a blitz rating or a classical one

 

Pat_Zurr

I can only speak for the Christmas bots.  I beat the 1600 bot easily, lost mostly against the 2000 bot, but still had several draws.  Considering my rating, I don't think the bots are accurate, but then again, maybe I am a bit better than I thought.  I carried a much higher ELO when I was on playchess and I just started on chess.com after a long, long layoff.