- Blind mode tutorial
lichess.org
Donate

Bug? Performance in Puzzle Dashboard can't be right

I sometimes look at the puzzle dashboard, and I always feel that the performance shown there can't be right.

If I take for example the last 30 days, it shows 45 puzzles played and a performance of 2070, but my puzzle rating raised in that time from 2370 to 2500 - so the performance can't be that low. I mostly do the puzzles of the day, but also played some puzzles of my actual rating, so there were a dozen puzzles rated between 2350 and 2500 that I solved.

Whenever I play in whatever tournament and I gain rating points during this tournament, my tournament performance shown is above my actual rating - only with the puzzles performance it's always way less than my actual rating - this can't be right.

I sometimes look at the puzzle dashboard, and I always feel that the performance shown there can't be right. If I take for example the last 30 days, it shows 45 puzzles played and a performance of 2070, but my puzzle rating raised in that time from 2370 to 2500 - so the performance can't be that low. I mostly do the puzzles of the day, but also played some puzzles of my actual rating, so there were a dozen puzzles rated between 2350 and 2500 that I solved. Whenever I play in whatever tournament and I gain rating points during this tournament, my tournament performance shown is above my actual rating - only with the puzzles performance it's always way less than my actual rating - this can't be right.

@out-of-mind said in #1
Your puzzle rating is not visible to other players. Its a big '?'. This happens when your rating of chess and puzzle rating exceed 800 points in difference.

@out-of-mind said in #1 Your puzzle rating is not visible to other players. Its a big '?'. This happens when your rating of chess and puzzle rating exceed 800 points in difference.

@KenulL_76 I know, but that's not the question - and to me everything is shown normal, just others see only the ? - and the problem I reported was also there before I reached the 2500 ... what does Lichess show as your performance on the puzzle dashboard - say for the last 10 or 21 days? Your puzzle rating has risen recently, but I guess the Performance shown is lower than your actual rating!?

Edit: the ? Appears when you reach 2500, but don't have a 1800+ rating in some fast games (or you reach 3000 and have no accordingly Higher rating)

@KenulL_76 I know, but that's not the question - and to me everything is shown normal, just others see only the ? - and the problem I reported was also there before I reached the 2500 ... what does Lichess show as your performance on the puzzle dashboard - say for the last 10 or 21 days? Your puzzle rating has risen recently, but I guess the Performance shown is lower than your actual rating!? Edit: the ? Appears when you reach 2500, but don't have a 1800+ rating in some fast games (or you reach 3000 and have no accordingly Higher rating)

I wonder if the (unrated) Puzzle Streaks affect the Performance Rating.

I wonder if the (unrated) Puzzle Streaks affect the Performance Rating.

@mcgoves interesting idea, possible, but I doubt any unrated puzzles are involved in any way - the dashboard shows the number of puzzles played in the selected time period - and those are only the normal puzzles and puzzles of the day

what performance is displayed on your puzzle dashboard for the past say 30 days and what is your puzzle rating change in that period (1 dec = 1936, 2 dec = 1947, 3 dec = 1958 - so 30 days period) should be something like +68 / + 57 / + 46 and performance should be something around 2075 - would be interesting to see what Lichess shows as performance

@mcgoves interesting idea, possible, but I doubt any unrated puzzles are involved in any way - the dashboard shows the number of puzzles played in the selected time period - and those are only the normal puzzles and puzzles of the day what performance is displayed on your puzzle dashboard for the past say 30 days and what is your puzzle rating change in that period (1 dec = 1936, 2 dec = 1947, 3 dec = 1958 - so 30 days period) should be something like +68 / + 57 / + 46 and performance should be something around 2075 - would be interesting to see what Lichess shows as performance

Do you often play themed puzzles, or easy (-300, -600) puzzles? This github issue may be relevant: github.com/lichess-org/lila/issues/10198 .

Your rating shouldn't decrease. Your performance rating, however, can decrease, as it's dependent on the average opponent (puzzle) rating.

Do you often play themed puzzles, or easy (-300, -600) puzzles? This github issue may be relevant: github.com/lichess-org/lila/issues/10198 . >Your rating shouldn't decrease. Your performance rating, however, can decrease, as it's dependent on the average opponent (puzzle) rating.

@mcgoves thanks for the link to the issue - and no, I only play puzzles of my rating (the standard mix thing) and puzzles of the day - ok those are often some hundred points lower than my rating - I just never observe a performance below my rating in tournaments when I gain rating, but always observe this with the puzzle performance - even on those days recently when most puzzles were in my range and just few below.

@mcgoves thanks for the link to the issue - and no, I only play puzzles of my rating (the standard mix thing) and puzzles of the day - ok those are often some hundred points lower than my rating - I just never observe a performance below my rating in tournaments when I gain rating, but always observe this with the puzzle performance - even on those days recently when most puzzles were in my range and just few below.

@out-of-mind said in #7:

I only play puzzles of my rating (the standard mix thing) and puzzles of the day.

Your profile's activity list only goes back a week, but in that time you did exactly one puzzle per day. My guess is that most puzzles you do are the puzzle-of-the-day, which are rated much lower than your 2500. So that would drag down the performance calculation. Consider skipping PotD for a few weeks, and see what happens.

Also, consider recording your rating before each puzzle along with the puzzle's revealed rating after it is solved, and calculate a performance on your own. I suspect the math done by lichess is misleading but correct.

I tried reading the code, but I'm neither a Scala nor Javascript guy, so I can't contribute there.

@out-of-mind said in #7: > I only play puzzles of my rating (the standard mix thing) and puzzles of the day. Your profile's activity list only goes back a week, but in that time you did exactly one puzzle per day. My guess is that most puzzles you do are the puzzle-of-the-day, which are rated much lower than your 2500. So that would drag down the performance calculation. Consider skipping PotD for a few weeks, and see what happens. Also, consider recording your rating before each puzzle along with the puzzle's revealed rating after it is solved, and calculate a performance on your own. I suspect the math done by lichess is misleading but correct. I tried reading the code, but I'm neither a Scala nor Javascript guy, so I can't contribute there.

@mcgoves said in #8:

Your profile's activity list only goes back a week, but in that time you did exactly one puzzle per day. My guess is that most puzzles you do are the puzzle-of-the-day, which are rated much lower than your 2500. So that would drag down the performance calculation. Consider skipping PotD for a few weeks, and see what happens.

Hi, I'm observing this performance thing for a while, not just the past few days. Probably I never observed a perfomance below my rating while gaining rating points because in tournaments I rarely have many opponents that are way below my rating. With the puzzles that's often the case due to the puzzles of the day. I probably underestimate the effect of "average opponents rating" that brings the calculated performance down to 2000 even in that week about two weeks ago where I had a dozen puzzles in the range of 2350-2500, but the puzzles of the day were on an average of maybe 1700. Though I got more or less everything right (high and low rated puzzles), Lichess seems to do some counterintuitive calculations.

Thanks again for the link to the issue with exactly my problem - I simply wasn't aware of the "average opponents" thing.

If this is reasonable or not ... - I don't know - when I win three 2400 puzzles and have a performance of say 2550, it doesn't make any sense in my opinion that the performance drops to 2000 by solving four more puzzles between 1600 and 1800 - but that's what actually seems to be happening.

I guess, I will just ignore the performance shown on the puzzle dashboard. I thought it was wrong, but it seems, it's unreasonable (to me) by design.

@mcgoves said in #8: > Your profile's activity list only goes back a week, but in that time you did exactly one puzzle per day. My guess is that most puzzles you do are the puzzle-of-the-day, which are rated much lower than your 2500. So that would drag down the performance calculation. Consider skipping PotD for a few weeks, and see what happens. Hi, I'm observing this performance thing for a while, not just the past few days. Probably I never observed a perfomance below my rating while gaining rating points because in tournaments I rarely have many opponents that are way below my rating. With the puzzles that's often the case due to the puzzles of the day. I probably underestimate the effect of "average opponents rating" that brings the calculated performance down to 2000 even in that week about two weeks ago where I had a dozen puzzles in the range of 2350-2500, but the puzzles of the day were on an average of maybe 1700. Though I got more or less everything right (high and low rated puzzles), Lichess seems to do some counterintuitive calculations. Thanks again for the link to the issue with exactly my problem - I simply wasn't aware of the "average opponents" thing. If this is reasonable or not ... - I don't know - when I win three 2400 puzzles and have a performance of say 2550, it doesn't make any sense in my opinion that the performance drops to 2000 by solving four more puzzles between 1600 and 1800 - but that's what actually seems to be happening. I guess, I will just ignore the performance shown on the puzzle dashboard. I thought it was wrong, but it seems, it's unreasonable (to me) by design.

@out-of-mind said in #9:

I thought it was wrong, but it seems, it's unreasonable (to me) by design.

I think the PR was invented for very specific reasons (GM norms; maybe tiebreaks). But since all the data is already in the computer, and computing power is basically free, we compute everything we can think of; and only afterward consider whether it is interesting or useful.

@out-of-mind said in #9: > I thought it was wrong, but it seems, it's unreasonable (to me) by design. I think the PR was invented for very specific reasons (GM norms; maybe tiebreaks). But since all the data is already in the computer, and computing power is basically free, we compute everything we can think of; and only afterward consider whether it is interesting or useful.

This topic has been archived and can no longer be replied to.