This rating system is wrong! If I fail a puzzle rated 2000 I lose 15 points. But if I pass it I only win 1 point! You don't need to be a genius to understand that this system makes no sense. I was rated 2580 and I've already lost 100 points since the change. For God's sake!
I think that's because the puzzle is below your level
And how can I change the level? I don't see any way to do it.
It is not the fault of the puzzle but of the ELO/Glicko rating systems, which seem to be (at least in my naive view) imprecise measurement methods: Strong Players lose points when they play weaker players for a while and weak players win points when they play stronger players for a while. Also there is ELO inflation, but thats another issue. Finally it is simply uninspiring when i see Nakamura win 25 games in a row against another player and still lose points because he loses game number 26.
What is the problem supposed to be, exactly?
Stronger players lose more points for losing to weaker opponents than they gain by beating those same opponents (which makes sense, of course, because if you alternate wins and losses and score 50% against much weaker opponents, your rating should drop).
That's how ratings have always worked.
There may be some problems with the new puzzle system, but I'm confused about the claim here.
Stronger players lose more points for losing to weaker opponents than they gain by beating those same opponents (which makes sense, of course, because if you alternate wins and losses and score 50% against much weaker opponents, your rating should drop).
That's how ratings have always worked.
There may be some problems with the new puzzle system, but I'm confused about the claim here.
The problem is that they lose too much points.
Now that would be a specific complaint, but then instead of being confused, I just have to disagree.
If anything, losing 15 points for a loss to an opponent 500 points weaker and getting 1 point for a win against the same opponent seems almost exactly right.
That implies that your expected score is about 93.75%, since you would maintain your rating if you got that score against a long string of such opponents
It might be a smidge high or low, but not dramatically so. If I remember correctly, the empirical results I've seen show that 92% or 93% is about the actual score for the higher rated player when the difference is around 500 points.
You can of course slow down or speed up rating change in a system in general, but that's going to decrease rating gained also.
If a player loses rating points in a 26 game match after going 25-1 as in your Naka example, then that means that he has such a high rating relative to his opponent(s) that his expected score in such a match is greater than 25.
Put another way, he underperformed his rating.
There's no good way to make a rating system where you don't lose points for underperforming your current rating or gain points for overperforming your rating.
That's kind of the whole point of rating change :)
Now, you might be claiming that the rating system is overshooting the correct expected score for a rating difference of 500.
That's certainly possible, although as I stated earlier, the rating change cited implies an expected score very close to what's been found empirically (I've seen this a couple places, but the most accessible is an article by Sonas in Chessbase from a while ago looking at the expectancy tables where he looks at the empirical results for various rating differences).
Ultimately, it might be a smidge off (it's possible that a rating difference of 500 should be something like -13 for a loss and +1 for a win, for example), but it's not going to be a dramatic shift from what the OP saw.
If anything, losing 15 points for a loss to an opponent 500 points weaker and getting 1 point for a win against the same opponent seems almost exactly right.
That implies that your expected score is about 93.75%, since you would maintain your rating if you got that score against a long string of such opponents
It might be a smidge high or low, but not dramatically so. If I remember correctly, the empirical results I've seen show that 92% or 93% is about the actual score for the higher rated player when the difference is around 500 points.
You can of course slow down or speed up rating change in a system in general, but that's going to decrease rating gained also.
If a player loses rating points in a 26 game match after going 25-1 as in your Naka example, then that means that he has such a high rating relative to his opponent(s) that his expected score in such a match is greater than 25.
Put another way, he underperformed his rating.
There's no good way to make a rating system where you don't lose points for underperforming your current rating or gain points for overperforming your rating.
That's kind of the whole point of rating change :)
Now, you might be claiming that the rating system is overshooting the correct expected score for a rating difference of 500.
That's certainly possible, although as I stated earlier, the rating change cited implies an expected score very close to what's been found empirically (I've seen this a couple places, but the most accessible is an article by Sonas in Chessbase from a while ago looking at the expectancy tables where he looks at the empirical results for various rating differences).
Ultimately, it might be a smidge off (it's possible that a rating difference of 500 should be something like -13 for a loss and +1 for a win, for example), but it's not going to be a dramatic shift from what the OP saw.
I just give up. Thanks to those who bothered to reply.
@a_pleasant_illusion if you are referring to this article by Sonas (excellent one):
en.chessbase.com/post/the-sonas-rating-formula-better-than-elo-
Sonas says more or less the same what i said:
Sonas:
"There is a consistent bias in Elo's table of numbers against the higher-rated player. To put it bluntly, if you are the higher-rated player, a normal performance will cause you to lose rating points. You need an above-average performance just to keep your rating level. Conversely, if you are the lower-rated player, a normal performance will cause you to gain rating points."
and later:
"The bias also has an effect on the overall rating pool. It compresses the ratings into a smaller range, so the top players are underrated and the bottom players are overrated."
and now what i said (in #4):
"Strong Players lose points when they play weaker players for a while and weak players win points when they play stronger players for a while."
So why are you confused?
en.chessbase.com/post/the-sonas-rating-formula-better-than-elo-
Sonas says more or less the same what i said:
Sonas:
"There is a consistent bias in Elo's table of numbers against the higher-rated player. To put it bluntly, if you are the higher-rated player, a normal performance will cause you to lose rating points. You need an above-average performance just to keep your rating level. Conversely, if you are the lower-rated player, a normal performance will cause you to gain rating points."
and later:
"The bias also has an effect on the overall rating pool. It compresses the ratings into a smaller range, so the top players are underrated and the bottom players are overrated."
and now what i said (in #4):
"Strong Players lose points when they play weaker players for a while and weak players win points when they play stronger players for a while."
So why are you confused?
I'm down a hundred points too since the update. It seems the puzzles are just a lot tougher. Before the update I would solve 2200 rated puzzles with ease, now I'm struggling with 2000 ones. So I guess this is what OP meant.
Gaining one point per solved puzzle and losing 15 for failing is pretty ridiculous. That didn't happen before the update, because rated at around 2500 you would get puzzles in the range from 2200-2700, almost never under 2200, and so you would get 3-18 points for solving them. Which makes more sence than +1 per puzzle. I guess OP Just needs to wish farewell to the 2580 ranking for now :D
I think the puzzles are tougher because the puzzle ratings are less inflated by very at attractive second-best moves. What I mean is that earlier, many puzzles contained totally winning lines (e. g winning the queen) that still were second-best, because there was a variant leading to mate available. So good players would often fail those puzzles and so the puzzle ranking kept growing. Once your tactics ranking hits 2500 you get those kind of puzzles most of the time. Realizing this shortcoming of the puzzles and double checking for even better lines after every move, made it possible for you to exploit the inflated ranking and harness a high tactics ranking for yourself (like I did). Now that the puzzles don't contain any good second-best moves anymore the rankings of the puzzles are closer to their real difficulty and our inflated rankings come crashing down. I guess the only way forward is to improve so much that 2500 can be reached with the new puzzles as well :)
Gaining one point per solved puzzle and losing 15 for failing is pretty ridiculous. That didn't happen before the update, because rated at around 2500 you would get puzzles in the range from 2200-2700, almost never under 2200, and so you would get 3-18 points for solving them. Which makes more sence than +1 per puzzle. I guess OP Just needs to wish farewell to the 2580 ranking for now :D
I think the puzzles are tougher because the puzzle ratings are less inflated by very at attractive second-best moves. What I mean is that earlier, many puzzles contained totally winning lines (e. g winning the queen) that still were second-best, because there was a variant leading to mate available. So good players would often fail those puzzles and so the puzzle ranking kept growing. Once your tactics ranking hits 2500 you get those kind of puzzles most of the time. Realizing this shortcoming of the puzzles and double checking for even better lines after every move, made it possible for you to exploit the inflated ranking and harness a high tactics ranking for yourself (like I did). Now that the puzzles don't contain any good second-best moves anymore the rankings of the puzzles are closer to their real difficulty and our inflated rankings come crashing down. I guess the only way forward is to improve so much that 2500 can be reached with the new puzzles as well :)
This topic has been archived and can no longer be replied to.