I tried a few games at level 5. Like all programs it doesn't make for a good comparison to human play.
But I noticed it doesn't focus on development in the opening. Also I offered material over and over, sometimes it took it, sometimes it didn't. In one game I took its rook and it didn't recapture. I don't know any site or OTB rating that would rate this at 1300. I would guess under 1000. But this sort of odd play is hard to compare to a human opponent (the blunders are ok, but not consistent.)
You are right about the odd play. It's like most of the computer levels (except the highest ones) intentionally make a few blunders, however, the tactical knowledge is pretty decent around Level 4 and above (although blunders make many levels do poorly). I'm thinking maybe the below approximate ratings for the levels.
Level 1: 300
Level 2: 500
Level 3: 800
Level 4: 1000
Level 5: 1200
Level 6: 1500
Level 7: 1700
Level 8: 2000
Level 9: 2100
Level 10: 2300
Of course, I know that at my own level of play I probably cannot make very accurate estimates. Perhaps a higher-ranking player (2000+) can review this?
Level 1 is 400. It goes up 200 per level except level 10, it goes from 2000 to 2600.
It might also depend on your hardware (it runs using your computer). On this (very) old laptop, level 9 plays terribly (although it did draw me a few opposite color bishop endings when it was a few pawns down).
So for you level 5 may be a lot better than when I tried level 5... I'd say the level 9 I played was maybe 1500-1600 heh.
Err... not that 1500 is so terrible, but when you're expecting 2000 and it's dropping pawns every single game...
The computer I am using is quite good. I have played Level 10 and it definitely plays at the GM level (2600 or so). I'm guessing Level 5 is somewhere around 1200 based on the fact I could beat it.
On what basis are y'all making any of these guesses? I guarantee you that I could't tell the difference between a 2600 rated player and a 2200 rated player just by playing against them... they'd destroy me equally, and I don't know the nuances in the highest level games. Similarly, a 500 rated player and a 1000 rated player are basically equal as far as I can tell by playing against them. Can you really tell the difference between someone you would beat 98% versus one you would beat 99.6% (for example) just by observing their play?
Yes, you can somewhat based on how actively they are attacking you and how often they leave their pieces en prise (and other basic blunders). A 1000 player typically has basic knowledge of tactics and applies them to the game, although his plans are often poor. A 500 player plays with no plan and at best avoids leaving pieces en prise (but he does have trouble with counting).
Long time ago, I have tried playing the Level 10 computer against Chessmaster 9 just for fun, and it whips the Chessmaster even when it moves instantly while the Chessmaster moves at blitz time like 4/1. Assuming the Chessmaster's elo is about 2650, I am guessing the computer has an elo of 2800, at least (at full strength). It has very good modern positional understanding, which old engines do not.
But it cannot beat Naum engine which has an elo of about 3000 and is considered an old engine that is no longer updated by the author. And compare that to Stockfish and Komodo at 3300+ just for perspective....
Just for kicks I took on level 9, and made it more challenging by playing as black. In some ways it was one of the better games I've played, but I still got mated in just over 30 moves because the AI played perfect chess, with only a single inaccurate move.
So I agree that level 9 is definitely rated at least 2000.
"Is 21.b6 not a blunder?"
Stockfish/lichess considered it an inaccuracy, but not a blunder.
It's almost like you were playing a different computer than I was. Granted, for much of the game it played quite well, but from moves 49 to 59 it made four blunders, according to Stockfish.
It's almost like you were playing a different computer than I was. Granted, for much of the game it played quite well, but from moves 49 to 59 it made four blunders, according to Stockfish.
I was playing against it on my Xbox One. If I was playing against it on my computer it probably would have played much better.
Because you are not timed when playing the computer, you are under less stress and have more time to think. Because of this, you play more strongly in those matches than in, for example, rapid chess. It means that on slower time controls (probably in OTB games) you would be better than here. I have the same situation myself.
Try playing level 7 on a PC. You will be obliterated.
These computer matches use your PC for the processing, you can easilly see it by checking your task manager. So rating them is useless.
Not to mention the unbelievable blunders that levels up to 5 are making, not even an 8 year old just starting chess would fail to simply exchange pieces that are next to each other.
There are really two factors that determine the Computer Level's performance: computer processing power and the decision-making model. As you go up in levels the decision-making model improves, but if you are playing the game on a weak computer it won't play as well because it is trying to meet a time deadline to make the decision. If you're playing on an iPhone 5 instead of, for example, a desktop computer, the computer probably won't play as well for you as it does for me. On my computer the Level 5 plays at approximately a 1200 level (since it plays about as well as the 1200 rated human players I have played).
Hello,
I was wondering how strong each level is for the chess.com computer. I can beat Level 5 without too much trouble (I think it may be ~1300 - 1400 or so), but after research I have seen quite a few guesses. I would like to figure out the approximate ratings to figure out my playing strength (since the human players at my rating level for some odd reason are way stronger than what their rating suggests (i.e., I see a great deal of variation in playing strengths of players with the same rating). Any ideas?