# Online resource for estimating my elo rating using my games?

There are resourses that can estimate your elo rating such as this one. Usually I need to solve some puzzles and the system will give me an estimation of my elo rating. Here I have a different thought: Can an online resource estimate my elo rating based on the games that I play?

For example, if I can upload the recent 100 games that I played under some standard time control, will there be a way to estimate my elo based on how I play? It seems feasible to me but I am not sure if such resource exists.

## 2 Answers

For example, if I can upload the recent 100 games that I played under some standard time control, will there be a way to estimate my elo based on how I play?

If, as part of that upload, you also supply the ratings of opponents in the same rating system (i.e. all FIDE or all lichess or all chess.com) for at least 5 of the games in which you score at least half a point, then yes. Of course the moves played would be irrelevant.

The point being that no rating system gives a measure of absolute playing strength. Every single rating system gives a measure of relative playing strength.

• Remember that "Elo" is only a system for calculating a rating (and also the name of the person who invented the system...), and the only thing the rating is useful for is comparing your success to other players in a given pool of players. You can't "estimate your elo" based on the moves in a game because it's not an absolute benchmark of skills (like, "if you have this rating, then you have the following playing skills") - it is literally only useful to compare yourself against other players in a given pool of players. Commented May 5, 2020 at 22:02
• Elo is not an absolute benchmark of skills, but that doesn't mean that your likely FIDE rating can't even be estimated by looking at your games without looking at the opponent's ratings. That's taking things too far in the other direction. The games of the typical 1400 player and the games of the typical 2400 player do not look the same.
– D M
Commented Feb 8 at 4:36

My suggestion would be to estimate your rating based on your ACPL (average centipawn loss). You should be able to find this by using Chess Compass and looking at the report tab. I would recommend analyzing only 10-20 games for the sake of saving your time, and also going to settings and changing lines to 1 and depth to 14.

After you have obtained your ACPL for every game, average all of your results to get a representative ACPL for all the games in your dataset. The next step is to use a formula to actually estimate your rating based on your ACPL. Despite various online searches, I could not find any information on the subject, so I decided to try making some myself. Note that I used a relatively small sample size in doing so (around 15 friends all within the ELO range 700-1900 and then some public info for several IMs and GMs).

Here are the results, but take these with a grain of salt:

Rating = 2500 - 16 * ACPL (for ACPL less than or equal to 120)

Rating = 2700 - 20 * ACPL (for ACPL less than or equal to 30)

Another concern with using purely just ACPL is that your ACPL will change based on who you are playing. For example, I would normally get an ACPL of 35 against a similar player. However, I could probably get that down to 20 by trying to play a specifically solid opening (like the London System), or possibly up to 50 by playing a very chaotic opening (like the King's Gambit).

Additionally, ACPL will change based on who you're playing against. If I were playing a 1000 rated player I would easily get a probably winning position, but more importantly a very playable one and thus have a lower ACPL. However, if I were to play Magnus Carlsen (or even worse, Stockfish), I would have a higher ACPL because I would get into a difficult and unplayable position and thus end up getting a higher ACPL.

To conclude, getting your rating based on some games is a huge pain and you should probably just play a few rated games on chess.com instead. Also, sorry for the really late answer.