Approximate Ratings of chess.com Computer Levels?


I just played Computer Level 1. 35 move game. . The computer One scored 22 excellent 6 good 1 inaccuracy 1 mistake 1 blunder 3 forced. That is hardly a level 400 rated player!

So overall Chess.com is amazing. To have a site like this is just wonderful. But there are areas where there seems to have been little oversight. In Tactics Trainer some of the target times are beyond ridiculous. And to have a Computer Level 1 playing at a rate where in a 35 move game they only make 1 blunder. That is supposed to be level 1, easy? If I were truly a beginning player I would find it very discouraging, and would move on to some other way to learn chess. (not that I am actually very good )

Excheqquer: the same way you are able to solve 1600 rating tactics but still be able to lose consistently to players with 1000 rating in blitz. that is the same logic used to calculate the ability of easy computer

I just played the level 1 computer twice. It played a lot of horrible moves. Idk if it was as low as 400, but it was pretty bad.
I had a somewhat comfortable draw with Computer Level 9: https://www.chess.com/analysis-board-editor?diagram_id=3728050

I have played extensively vs. the computer and I will concur that level 5 makes a whole lot of mistakes that a normal 1200 level player wouldn't. It doesn't take advantage of hanging pieces, etc...Level 6, on the other hand, is more than 200 points above 5, I have only beaten 6 a couple of times; I beat 5 rather easily.

i think computer level 3 is easily 1000 or higher, it makes blunders but is a very strong player. I'm over 1000 and level 3 beats me all the time, unless it makes a blunder.

Level 1 is quite bad but is good at checkmating or seeing blunders sometimes, I can beat a Level 2, 3, 4 and 5 on Computer but for some reason, I can't beat level 1. Here is one of my games:
You might be thinking, wow, that guy blundered at (A MOVE) but everyone blunders.
Anyways, you can see how I lost to a level 1 and how good they are at spotting checkmate and blunders, the problem is, sometimes they respond to them in the wrong way, if level 1 computers worked on their moves overall, they would be level 8 or 9.
wtf in my opinion only difference between computer level 1 and 5y old kid who's just been taught chess rules and makes completely random moves, is, level 1 plays 100% legal moves and 5y kid still only 50% legal moves

The chesscom computer program with 10 levels is unique. It botches up game at all levels just like humans do. It's a great advantage and nice feature. It also sees tactics extremely well like all engines do. In order to compare its levels to ELO ratings we should remember that there are several ELO systems in existence:
FIDE for long games, for rapid games, for blitz games;
USCF for long games, for blitzes;
CCRL 40/40 for engines, and CCRL 40/20 for engines.
I'm not gonna be touching upon the Internet rating systems based on ELO principles though. You can easily measure your FIDE ELO by playing rated games against a Fritz or cheap phone Shredder (use the arithmetic mean value for the highest precision) but no takebacks are allowed! Consulting opening books is not allowed. Evaluation is not allowed. Postponing or pausing the game is not allowed. Time control should be very long. Second board is not allowed either. Hints are not allowed.
I've performed some assessments of the levels of the chesscom computer program but before I write my findings and approximations, FIDE vs USCF ratings should be explained just a little bit, I mean the main difference in reality. But before even that, we also need to address differences between long games and blitzes. Well, as you can see in the official sources, the FIDE ELO for long games goes as high as 2850-2900 while for blitz games it goes as high as 2950--3000 for the upper end. For the lower end it's considered to be reversed, and the whole ELO scale line is longer for blitz games. It means lower ELO ratings for weaker players in blitzes than for the same weak players in long games. At the same time ELO ratings are higher (NOT LOWER) for the top players in blitz games than for long games. (I'm omitting the reason here for shortness.) Now than, we also have USCF ELO ratings vs FIDE. The USCF rating line is even longer. It starts from the very bottom, like ELO 400 or even lower, while FIDE starts at around 800-1000. So you can see how huge the differences are. I mean a FIDE 1100 vs USCF 1100 or USCF 900. They are separated by hundreds of points especially for beginners. I mean a USCF 1000 is way stronger than FIDE 1000. Anyway the both systems are mostly used for professional players with ELOs above 2000. At around 2000 points of ELO the both systems (FIDE and USCF) show the similar numbers. At the upper end USCF gives gives slightly higher ELO points (perhaps, around 50 points) for long games. Now we should bear in mind the enormous differences in FIDE ELO vs USCF ELO for beginners, which most chess players don't care for since most tournaments have participants with ELO higher than 2000.
Now I'm gonna compare the 10 levels of the computer program on this website with FIDE ELOs, not USCF. This is needed for consistency. One more remark: the computer here always plays instantaneously just like in bullet chess. The quality of game drops by 500-600 ELO points if you are given 3 minutes for the game instead of 3 hours. But I'm gonna compare it to long game ELO points of FIDE (just like if the chesscom computer spent a lot of time before making a move) -- that is the question here asked by the OP. So here are my approximate findings (+/-100 Elo points. Yes, that's not precise at all but at least you can have the approximate strength):
Levels 1, 2, 3 are for novices who just learn the rules and values of pieces. I'm not even going to rate them as FIDE also doesn't have very low levels. It starts somewhere around 1000. Most tournaments are way above that.
Level 4 -- FIDE 1100
Level 5-- FIDE 1300
Level 6 -- FIDE 1500
Level 7 -- FIDE 1750
Level 8 -- FIDE 2100
Level 9 -- FIDE 2300
Level 10 -- FIDE 2650
I found a large gap between level 7 and level 8. The same gap exists between levels 9 and 10. Also the computer botches up things at all levels. At level 10 it just doesn't choose the strongest moves and gives a strong grandmaster a fighting chance in a very long game. At level 9, the computer may make slightly inferior moves (or even worse) but the level is astronomical in tactics. Sometimes it says mate in 24 or something like that! Level 8 is similar but a couple of very slight inaccuracies might be present in critical positions. So you can potentially beat it. Level 7 is much easier; it's around FIDE ELO 1750; and very strong amateurs or average club players can beat it in long games. Level 6 is not difficult at all. All amateurs can beat it but "I’m taking back that move" doesn’t exist in serious chess! So beating the chesscom program should be fair. No “take back”! I’m saying this because people often clam: Hey, programs are stupid (yes, they are strategically very weak even Stockfish -- it's just a very advanced calculator) and we can beat them at low levels but taking back moves should not be allowed. That’s how beginners beat level 5 and 6 – they take back moves! You can’t do that in an OTB tournament. Now some claim that the chesscom computer heavily depends on the hardware. It may be so but I don't see difference for levels up to 8-9. It's very consistent and good enough. I also think that level 10 doesn't really depend on hardware, maybe just a little bit. I'm well aware how much improvement hardware can make. I saw it with my own eyes. It's just that level 10 is way below the current levels in chess engines due to intentional moves that are not overly precise. It has to give you a fighting chance at drawing the game if you are a grandmaster.
Although I've said that the chesscom computer program is good, I think it can be further improved on with more levels added so that people can select their levels better. Chess clocks should be added too. Right now it doesn't allow enough levels for amateurs to practice with.

Some clarification in accordance with the FIDE system is necessary. As I said it all relates to long games only! Level 10 of the chesscom program will beat any grandmaster if he tries play bullet or blitz against that level. The same holds for lower levels of the program when pitched against lower FIDE players. So here the strength of the chessom computer program as if it were playing long tournament games:
1000-1200 ELO is razryad 5 (grade 5) – beginner. Level 4 of the chesscom program.
1200-1400 ELO is razryad 4 (grade 4). Level 5 of the chesscom program.
1400-1600 ELO is razryad 3 (grade 3). Level 6 of the chesscom program.
1600-1800 ELO is razryad 2 (grade 2). Level 7 of the chesscom program.
1800-2000 ELO is razryad 1 (grade 1)
2000-2200 is a candidate master. Level 8 of the chesscom program. (I have a Saitek dedicated chess computer (portable) from a bygone era which plays at this level)
2200-2300 is a FIDE candidate master. Level 9 of the chesscom program is a tad stronger than that.
2300-2400 is a FIDE master.
2400-2500 is a FIDE international master.
2500 and above is a FIDE grandmaster. Level 10 of the chesscom program corresponds to a strong grandmaster.
Women have a different system of course, and titles are not granted just by reaching some ELO rating. It’s much more complicated than that.
The CCRL 40/40 doesn’t correlate well with this system. It is way, way above in terms of playing strength. Kramnik showed that when he played against Deep Fritz 8 (a very week program even on two Intel Core 2 Duo CPUs as it was played). As to Kasparov against Deep Blue, I have doubts that everything was fair and square. Kasparov claimed that some grandmaster or grandmasters were helping out Deep Blue. Kasparov didn’t receive the printout of the game fast enough when the computer supposedly missed a 3-move tactic. I think Kasparov was right. I believe Kasparov was stronger than Deep Blue, which was not that strong after all, probably much weaker than Kramnik’s Fritz 8.
The most difficult question is how the number of plies (half-moves) calculated ahead corresponds with the ELO rating. As Kasparov modestly said he doesn’t miss 6 ply tactics, but playing against Deep Blue which calculated 14 plies in middle game and even more in end game was a tall order. I think Kramnik faced a much more sophisticated program in that respect. Nowadays chess engines are rated by CCRL as 3300-3400, and they compute 25-30 plies. That is too weak!!! I mean ELO 3300-3400 is too weak for 25-30 plies. I mean we know that Fritz 8 was on par with Kramnik, yet newer and much, much stronger programs that were released 5-6 years later are not rated as much higher than that. I mean they are not rated as 3200-3300. Only the modern ones (2016-2018) are rated that high. I believe all engines after 2006-2007 are now underrated by some 200 points of ELO. I also think that too many draws wreak havoc in the whole business of CCRL ELOs as they have to use blitz time controls more and more.
Bottom line: you can only win against modern chess programs when the latter play slightly inferior moves, or when the depth of computing is restricted by a certain number of plies, or when you have some odds like a couple of pawns. It’s no longer restricted by hardware. I don’t believe that the 10-level chesscom program underperforms on certain PCs even at level 10 because said level 10 is as low as 2650 FIDE ELO or in other words it is a low as computing only 14 plies ahead. Weird, but it looks similar in strength to Deep Blue (Deep Blue was claimed to compute 7 moves = 14 plies ahead) -- a little weaker of course -- because the chesscom program plays more varied game that gives a grandmaster some chances. Sometimes it sees a tactic that is too complicated though. Even too complicated and too long for Deep Blue! As I said, even at lower levels it sometimes shows something like mate in 24 (24 moves for black + 24 moves for white calculated in a couple of seconds = 48 plies ahead). So, it may be said the ELO jumps up and down at all levels, even at levels 5, 6 or 7. Sometimes the chesscom program makes moves at a level as low as 2400, and then it starts to compensate by making moves at ELO level 3200 (for level 10). Sometimes it may irritate you when the program starts to see way too far ahead or make stupid inaccuracies. It’s across all levels: from 4 to 10. I didn’t test levels 1, 2, and 3. They are just for people learning moves and the value of pieces in my opinion. They don’t (in my opinion) have any good correlation with ratings.
For testing purposes I used the Saitek Travel Champion, mobile phone shredder 1.2.2, Fritz 6 and Fritz 11. I also tested these computers/programs against each other and against myself (I’m a very weak player. I only once played in an OTB tournament and got only razryad 2 (grade 2) which corresponds to FIDE Elo 1600-1800. I didn’t play any further ‘cause I was not ready for higher levels and a bit more serious grades/titles. I guess I need to prepare a little bit if I’m to participate in any amateurish OTB events.) Grandmasters told me it’s very easy to reach Elo 2000 but I guess you need a bit of practice in long games before you reach that level. Bullet chess is fun but not conducive to improving your game. If I find time I will improve my weak play but playing long games against a program is a bit boring. Regretfully, there aren’t many opponents here who wish to play long games. As to bullet chess, there are as many opponents as you wish but I feel like the more I play bullet chess, the worse player I become. Bullet games are very addictive but you need to move the mouse faster

wtf in my opinion only difference between computer level 1 and 5y old kid who's just been taught chess rules and makes completely random moves, is, level 1 plays 100% legal moves and 5y kid still only 50% legal moves
Wrong, a 5y kid would play one opening which they got taught and then COMPLETELY random moves.
On average, with stockfish L1 does 50% moves which are quite decent...