I'm really interested in the OP's comment about playing what appear to be vastly different playing skill in the <1200 rating range. Sometimes I play people who appear to have only the most basic understanding of chess moves, and they are rated 1050ish. Then I play someone at 1080 and I get matched with an 80+ move game full of nuances and beauty. WTF??
I wonder if there is a way we can get more accurate ratings than just guessing. How would it work? I imagine somebody with a powerful computer but under strict time control (so no ponder) have the different levels of chess.com computer play a bunch of against engines with known ELO (roughly speaking of course, perhaps engines that has played hundreds of games against other engines and high level players).
I've tested it against some other engines. It is slightly weaker than Stockfish 8 on 25 ply (only slightly). On the Elo scale as used in TCEC compared to Stockfish 8 I'd say the Level 10 engine is somewhere around 2900-3000. I've checked its moves and almost all of them match perfectly with Stockfish 8 (it occasionally makes a "microblunder" but that's because it doesn't search very far). Level 9 is also tough but I would put it much lower on the scale because I have decent chances of beating it (so I would rate it ~2100-2200 TCEC Elo).
Great thank you, I did imagine it must be better than 2600-2700 at least because an engine I was using that was rated around that would lose to it often.
Wow, if you can beat level 9 you must be pretty good. I play on my PC and I'm stuck on level 7, have yet to beat it.