![]() ![]() Thus your conclusion would be "I'm as good as a 1100 rated player on _this_ site". Also, let's say you make an account on lichess then play stockfish level 2 in let's say 3+2 games and you win like 9/10 games. If you're a 900 rated player and play Stockfish Level 2 and you win and win and win then that's surprising because that shouldn't happen if it had 1100 rating => surprise. That's a basic design principle: Principle of Least Surprise. ![]() If a user on lichess plays Stockfish Level 2 which is displayed as having 1100 rating then surely an average lichess user would expect that Stockfish Level 2 plays at the level of a 1100 rated player on lichess. It's not that difficult I don't get what the problem here is. If 10 900 rating players can beat Lvl2 in at least 7/10 matches each then it's already very unlikely that lvl2 actually plays at 1100. Either lichess guys will run some analysis through their database or somebody else does it (since it is public) and publishes some more data or we just have a bunch of people playing lvl2, lvl3 and publish the result here. ![]() Of course this is just one data point and me saying subjectively stockfish level 3 makes mistakes not even a 1100 player would make isn't scientifically convincing but it's not like it's uncommon for way lower rated people beating stockfish (at least based on rough search results) (of course, they could've cheated who knows) but I think there's enough doubt/reason to investigate this and adjust the rating so it's actually accurate. I win about 2/3 of the games (~10min) against Stockfish Level 3 which has been given a 1400 rating yet my rapid rating isn't close to that and the conclusion thus would be that Stockfish 3 doesn't have 1400 rating. If 900 players on lichess can consistently beat Stockfish Level 2 then the 1100 rating it's given doesn't accurately reflect the strength of level 2. We're not discussing rating point systems but whether the rating given to Stockfish levels are a rough reflection of the playing strength accross different time controls. However, that is again completely not relevant for the point. If you were to create a plattform for elite players only and have them start at an initial rating of 800 none of them would reach >2500 ratings points because since they're fairly equal in skills they'll probably stay in the 600-1200 range. Meaning Stockfish Lvl 2 is about 200 rating points too high.Īnyway technically you can't compare ELO ratings or other ratings between different plattforms anyways because it's not an absolute measure of skill but a relative measure of skill because the rating is only valid within the rating pool. Doesn't really matter if it's ELO, Glicko, Glicko2 or some other rating system because you're comparing players within the rating pool and within that rating pool Stockfish Lvl2 plays around 900 rating points - not 1100 rating points. The point is that Stockfish Lvl 2 does not play at 1100 rating points. Thanks to all code contributors, fishtest contributors, and supporters for their contributions to this release.That doesn't actually matter.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |