puzzle rush ability vs OTB rating strength ?

GM_Kenny_Ji

Vicariously-I
Elroch wrote:
Vicariously-I wrote:

I've gotten 29 a few times but I haven't broken 30 yet. I have a 23+ average though. I tend to go for accuracy rather than speed. I hate getting problems wrong.

This is interesting to compare because I do the same.  My best at that time (January) was 28 and my average just below 21. With 6 months more, Vicariously-I has got his average up to 24.55 and his best up to 32, while my average is 22.15 (>23 in the last 100) and my best 31, so he remains slightly more consistent than me, with almost exactly the same improvement in best and average.

Yeah cool I guess we have both improved our best and average. I was stuck on 29 for a while but now I can't seem to break 32. I came close a few times today though taking advantage of the 24 hour unlimited puzzle rush.

Ziryab
GM_Kenny_Ji wrote:

 

 

Where did you get that data? That does seem to offer prospects for a degree of correlation, contrary to my claim.

But why choose high score and not average?

MangoMankey

Average can be inaccurate for people who quit immediately when they get one of the first few puzzles wrong, see for example https://www.chess.com/member/hallvardhf , https://www.chess.com/member/anthonyatanasov , https://www.chess.com/member/wolfbewwi

Vicariously-I
Ziryab wrote:
GM_Kenny_Ji wrote:

 

 

Where did you get that data? That does seem to offer prospects for a degree of correlation, contrary to my claim.

But why choose high score and not average?

I think high score is more reliable than average score for most people. Look at @penguingm1, for example. His best score is 56 but his average is only 19.68. This is because he starts over when he has a bad start. I hear that many people do this.

Ziryab

Maybe such people rage quit blitz games, too.

drmrboss
GM_Kenny_Ji wrote:

 

Around 2000 in blitz, Puzzle rush 31. (Similar to your graphic distribution) happy.png

btickler
Vicariously-I wrote:
Ziryab wrote:
GM_Kenny_Ji wrote:

 

 

Where did you get that data? That does seem to offer prospects for a degree of correlation, contrary to my claim.

But why choose high score and not average?

I think high score is more reliable than average score for most people. Look at @penguingm1, for example. His best score is 56 but his average is only 19.68. This is because he starts over when he has a bad start. I hear that many people do this.

And I bet that same impatience impacts their classic time control ratings on average to some degree  wink.png...

odyson

When I do tactics trainer, I can make a pot of tea and enjoy it while training. When I do puzzle rush, my tea grows cold. This is probably the main reason I haven't done puzzle rush more than 15 times.

Besides, I have a gnawing suspicion that it leads to superficial evaluations of positions.

Elroch
Vicariously-I wrote:
Ziryab wrote:
GM_Kenny_Ji wrote:

 

 

Where did you get that data? That does seem to offer prospects for a degree of correlation, contrary to my claim.

But why choose high score and not average?

I think high score is more reliable than average score for most people. Look at @penguingm1, for example. His best score is 56 but his average is only 19.68. This is because he starts over when he has a bad start. I hear that many people do this.

True but, as an outlier, the high score is theoretically quite a high variance statistic for everyone.

A test is to compare the correlation coefficients between blitz rating and the two PR statistics.

MangoMankey

@GM_Kenny_Ji How did you collect the data? Was it done manually, or did you write a script to collect data, presumably using chess.com api?

Ziryab

My rating and PR performance are not represented.

Elroch

The ratings of people with the same puzzle rush best score vary by about 1000 rating points in the middle of the graph.

Ziryab

The two far right dots on my PR score's line represent the range within which rating usually falls, although it is in the upper half of the range 90% of the time. The rest of the dots represent ratings that I haven't had in years, and mostly that I had only on the way up. The line marks a spot where I was in 2014, briefly, after a loss of more than 100 rating points during a slump.

All this is to say that I don't rust the data.

 

Among other things, the data should be controlled for level of activity in both blitz and Puzzle Rush. There is no indication that such factors were considered.

btickler
odyson wrote:

Besides, I have a gnawing suspicion that it leads to superficial evaluations of positions.

I played for a few hours yesterday during the all day event, and by the end of it, I was not getting better, I was getting worse on average.  Doing serial puzzles that long actually made me feel like rushing more and guessing faster on problems I did not immediately solve.  It has the same effect as playing blitz, in that respect.  I don't think this kind of ADHD-affirming mechanic is really good for getting better at the game past a certain point.  It has diminishing returns and then starts to impede overall chess ability if abuse continues wink.png.  Call it the Nakamura Effect, if you will.