Analysis tool suggestion: Count of games played with certain accuracy

Hey,

Is there a way in the analysis tool to do the following:

x-axis: accuracy from 0% to 100%
y-axis: number of games played with each accuracy.

If there is a way to do this can you please explain how. If this is not possible, could you add this kind of tool into the analysis section.

My hunch is that this should be somewhat normally distributed around your mean against equal (plus or minus 100 rating) opposition. Of course this could/should be combined with other arguments like opposition strength, game phase etc.

This could be interesting way to analyze you own performance and other players as well. Only seeing your average does not carry too much information necessary.

Only if such an analysis were itself accurate. ;)

You could download your games and only select analyzed games (for example, https://lichess.org/api/games/user/kihihi?analysed=true&tags=true&clocks=false&evals=true&opening=false), then do it yourself fairly easily with python-chess and matplotlib. If you want to include all your games (not just analyzed games), you could download them all and analyze them yourself (python-chess will help again), but needless to say, this will take a lot of time and processing power.

@MrPushwood said in #2:

Only if such an analysis were itself accurate. ;)

Okey. I just asked for this distribution, didn't discuss the analysis itself where it could be used.

I agree that this distribution is not an analysis by itself. Neither is an average for example.

I would personally want to eyball the distribution rather than just statistics like average.

Matter of taste I guess.

@AsDaGo said in #3:

You could download your games and only select analyzed games (for example, lichess.org/api/games/user/kihihi?analysed=true&tags=true&clocks=false&evals=true&opening=false), then do it yourself fairly easily with python-chess and matplotlib. If you want to include all your games (not just analyzed games), you could download them all and analyze them yourself (python-chess will help again), but needless to say, this will take a lot of time and processing power.

thx. This is why I would wan't Lichess to do it ;)

@kihihi said in #5:

thx. This is why I would wan't Lichess to do it ;)

Well, Lichess is free so they can't devote a ton of resources to analyzing every single game. Even chess.com doesn't do that.

Well here is possibly some misunderstanding going on.

I am suggesting the following:

While they have an algorithm that counts the average accuracy per player per variant, they must have all the games (with accuracy) of that player already in the dataset.

Now in addition to the avarage they provide, they could provide players the distribution (from which they calculated the average) as well.

And this is propably 5 rows of code.

And anyone could use these distributions in the way they see fit. I am not touching this topic of analysis.

Hopefully my suggestion is clear enough now.

@kihihi Ok, yes, you're right, there is insight data on accuracy. That must only take into account analyzed games. Maybe this could help? https://lichess.org/insights/kihihi/accuracy/accuracy I'm not exactly sure what that represents, but it does have a table at the bottom with accuracy and number of moves.

@AsDaGo said in #8:

This accuracy by accuracy is not too helpfull in my opinion.

It looks like it gives an avereage accuracy (of moves played in this decile) per decile, which is roughly the middle point in that decile. Makes sense but does not give too much of an insight.

Also my average accuracy in blitz is ~65%, with total of 150k moves. And this accuracy by accuracy in blitz sums up to 90k moves. Does not make too much sense either.

So I have played ~5k games of blitz, with an average accuracy of ~65%. I would like to see how these 5k games distribute over the scale of accuracy from 0% to 100%, let's say separated with 2% from one another.

Could Lichess reply to this or tell me how to get in contact with them?

I find this rather hard to communicate with Lichess.