The last few days I've been spending all my free time figuring out how to extract useful data from the 25,000+ blitz games I've played online.
Apart from looking at where my rating started (511) and where it has gone since (as high as 1920), the most important thing in order to get there must have been playing games.
One thing I did was create a scatter chart with a data point for every month I played. Number of games played vs difference in average rating each month. And there's a trend: at least 50 blitz games per month and my rating tends to go up. Over a long period of time (nearly 9 years) this is obvious. But of the many months I played, 44 of the 102 featured my average rating going down.
Anyway -- it's clear: People should play games if they want to improve. You're gonna lose -- that's part of learning how to win.
The price of getting better is losing. Adults really hate losing and often try to treat chess like a knowledge contest. Once you know the basic patterns (tactics, mates) then it’s a process of improving your skill. That only works when you see what you don’t do well, what you fail at. Failure is the only way to get better
“If you avoid losses, you deprive yourself of the most valuable lessons you could be getting.”…this landed so much..such a switch in attitude and an excellent post. Thanks
I am of the opinion that the problem is not losing. It is ratings. If ratings did not exist many people would be upset over losing but without a metric to reflect it the pain would be considerably less. It takes a lot of emotional fortitude to play at Lichess with all ratings hidden.
I know ratings are a negative for many people, but for me, if ratings didn’t exist, I doubt I’d be interested in chess at all. The struggle for measurable improvement is what makes chess meaningful to me. I suspect many others are the same.
There is a lot of empirical evidence that measurable improvement is largely but not entirely a myth. Seeking a myth is bound to result in unhappiness for many, but admittedly not everyone.
What’s a myth? By measurable improvement I mean significant ratings increases. Ratings do stay flat for most players most of the time, but once in a while they go up. That’s not a myth.
For the vast majority of chess players who are beyond the beginner level significant rating improvement will not occur. People who advertise services and products to the contrary create a myth in my opinion. I did previously state largely but not entirely with regard to the myth comment.
Agreed, improvement is very hard and very rare. But that’s what makes it meaningful when it happens! Without ratings, we would not be able to identify these very rare achievements when they happen.
Was it Steinitz that went into a cave and came back in tournament with a new style? so much loss of information for the learning problem (data for theory), where did he lose games, obligation for learning and exploring new avenues.... And where were those less glorious games recorded. On what napkins? alone, really?
I find that the logic of losing is the same as that of winning, only switch the colors of what you were just experiencing.. still.. the social instinct of competition is why this is more motivating to make some effort than mere solo problem studying.. Although I can do offshoots of those from previlous game experiences where I notice some disconfort with respect to aspects I can generalize having a weakness (not the hindsight ones, but those I find I have little imagintion in planning or what to do, those where I can only use raw calculation, and being lazy, I will work on how to avoid, that expenditure, by narrowing down or abstracting up (or both) what it is that troubles me, and of course, errors in hindsight and SF toolery would ensue, but i find that having a more relax attitudes about winning or losing and playing very slow, to look at the scenery while playing, is condusive to such intro-spection in the foresight problem, being slower games there is more time to 1) make listen unproductive silly mistakes, and also more time to dream ahead and find such high level errors in that kind of thinking, not just computer, but imagination and assessment of what to choose as possible goals (tactical or not, my plans are usually always having to be revised).
but basically, that is how deep reinforcemnt leanring in A0 and Lc0 proceeeds.. wins and losses equally contribution to its statistical learning about the many positions visited upstream.
we don,t have the same patience and ability to redistribution causation from you outcomes to the previous positoins along the game in uniform way along the whole depth of games, so we make stepping stone theories of features on board we might have developped pattern recognition with or wihtiout shared words from population chess theory (in opposition to our internal evovling one, the current black box of learning theories in chess).
so we make plans and hypotheses of what is beneficial, but getting non favorable outcomes at that level, while evovling how we value such human limitation required non-turn by turn features sets awareness and desirablity as possible in-game future, is equally contributing to learning that emerging chess board logic..
Thanks for finding a way to distill that machine learning influenced understanding of mine without going as cryptic as I might have just done.
The last few days I've been spending all my free time figuring out how to extract useful data from the 25,000+ blitz games I've played online.
Apart from looking at where my rating started (511) and where it has gone since (as high as 1920), the most important thing in order to get there must have been playing games.
One thing I did was create a scatter chart with a data point for every month I played. Number of games played vs difference in average rating each month. And there's a trend: at least 50 blitz games per month and my rating tends to go up. Over a long period of time (nearly 9 years) this is obvious. But of the many months I played, 44 of the 102 featured my average rating going down.
Anyway -- it's clear: People should play games if they want to improve. You're gonna lose -- that's part of learning how to win.
I love this chess data stuff and it’s why I first subscribed to Nate in I think 2021. Unfortunately he doesn’t write about it much anymore!
If I can organize myself I may make a post and share the code for others to use.
The price of getting better is losing. Adults really hate losing and often try to treat chess like a knowledge contest. Once you know the basic patterns (tactics, mates) then it’s a process of improving your skill. That only works when you see what you don’t do well, what you fail at. Failure is the only way to get better
“If you avoid losses, you deprive yourself of the most valuable lessons you could be getting.”…this landed so much..such a switch in attitude and an excellent post. Thanks
I am of the opinion that the problem is not losing. It is ratings. If ratings did not exist many people would be upset over losing but without a metric to reflect it the pain would be considerably less. It takes a lot of emotional fortitude to play at Lichess with all ratings hidden.
I know ratings are a negative for many people, but for me, if ratings didn’t exist, I doubt I’d be interested in chess at all. The struggle for measurable improvement is what makes chess meaningful to me. I suspect many others are the same.
There is a lot of empirical evidence that measurable improvement is largely but not entirely a myth. Seeking a myth is bound to result in unhappiness for many, but admittedly not everyone.
What’s a myth? By measurable improvement I mean significant ratings increases. Ratings do stay flat for most players most of the time, but once in a while they go up. That’s not a myth.
For the vast majority of chess players who are beyond the beginner level significant rating improvement will not occur. People who advertise services and products to the contrary create a myth in my opinion. I did previously state largely but not entirely with regard to the myth comment.
Agreed, improvement is very hard and very rare. But that’s what makes it meaningful when it happens! Without ratings, we would not be able to identify these very rare achievements when they happen.
I appreciate Dan your comments. And acknowledge your impressive otb rating climb over the years.
Was it Steinitz that went into a cave and came back in tournament with a new style? so much loss of information for the learning problem (data for theory), where did he lose games, obligation for learning and exploring new avenues.... And where were those less glorious games recorded. On what napkins? alone, really?
I find that the logic of losing is the same as that of winning, only switch the colors of what you were just experiencing.. still.. the social instinct of competition is why this is more motivating to make some effort than mere solo problem studying.. Although I can do offshoots of those from previlous game experiences where I notice some disconfort with respect to aspects I can generalize having a weakness (not the hindsight ones, but those I find I have little imagintion in planning or what to do, those where I can only use raw calculation, and being lazy, I will work on how to avoid, that expenditure, by narrowing down or abstracting up (or both) what it is that troubles me, and of course, errors in hindsight and SF toolery would ensue, but i find that having a more relax attitudes about winning or losing and playing very slow, to look at the scenery while playing, is condusive to such intro-spection in the foresight problem, being slower games there is more time to 1) make listen unproductive silly mistakes, and also more time to dream ahead and find such high level errors in that kind of thinking, not just computer, but imagination and assessment of what to choose as possible goals (tactical or not, my plans are usually always having to be revised).
but basically, that is how deep reinforcemnt leanring in A0 and Lc0 proceeeds.. wins and losses equally contribution to its statistical learning about the many positions visited upstream.
we don,t have the same patience and ability to redistribution causation from you outcomes to the previous positoins along the game in uniform way along the whole depth of games, so we make stepping stone theories of features on board we might have developped pattern recognition with or wihtiout shared words from population chess theory (in opposition to our internal evovling one, the current black box of learning theories in chess).
so we make plans and hypotheses of what is beneficial, but getting non favorable outcomes at that level, while evovling how we value such human limitation required non-turn by turn features sets awareness and desirablity as possible in-game future, is equally contributing to learning that emerging chess board logic..
Thanks for finding a way to distill that machine learning influenced understanding of mine without going as cryptic as I might have just done.