Low Correlation between Puzzle Proficiency & Rating - Implications for Chess Improvement?

Made it through half of the original post....

My puzzle score fluctuates quite a bit... most commonly sitting between 2100 and 2200, but has been as high as the low-mid 2400s. My game ratings aren't anywhere close to this.

Why does it fluctuate so much? Clearly, there is a correlation between my patience/time spent per puzzle and my puzzle rating.. and I think this may be the answer to your question.

If I spend a TON of time per puzzle.. more than I'll ever spend in a game per move, (outside of correspondence play)... I see a positive effect. Maybe some people treat puzzles like this... feeding into your phenomenon. For me, I'd say that I benefit w/ increased time controls more than the average player... so perhaps I'm an extreme case.

I exemplify the lack of correlation between puzzles and play in actual games.

bullet: 1195
blitz: 1173
rapid: 1405

puzzles: 1964

mind you, the puzzles rating can't be directly compared to a game rating (see this: https://lichess.org/forum/general-chess-discussion/puzzle-rating-too-high)

but i think my puzzle rating is pretty decent and it does not reflect my game play rating. I think an important component of good game play rating is consistency. I let too many blunders thru during game play, but when i solve puzzles i am more focused and take enough time until i think i've solved it before making a move. I don't just guess. In a game i don't have that luxury as time is running.

So in addition to winning at chess being more than just tactics (its also good development, strategy & planning, positional understanding, endgame theory etc.) there is also that aspect.

@Blundered_the_queen said in #1:

Basic summary of a very long, rambling post: Why are players who are good at puzzles bad at chess, and vice versa? Does this mean that puzzles aren't the best way to train chess? If so, what is the best method to train at chess?

I've heard the following refrain numerous times both in real-life and on the Internet: Any chess player under 2000 FIDE doesn't need to bother with strategy, openings, or even endgame theory. All they need to do is keep doing puzzles until they get better at tactics and stop giving their pieces away, and they will naturally reach 2000 FIDE.

If this were true, then I think there is also an implicit assumption we could make of players under 2000; the closer they are to 2000, the better they should be at puzzles! People who are good at tactics will have a high puzzle rating, and likewise a high blitz/rapid/classical rating. You would expect there to be a very strong direct correlation between how good someone is at chess puzzles and how high they are rated, at least until they hit the highest echelons of chess and become a titled player.

And yet, whenever I look at the players I've played, I'm not entirely sure that any such correlation exists. At my current level of play (which is about 1200 blitz and 1300-1400 rapid), the puzzle ratings of my opponents can vary from as low as 1300 to as high as 2100 (no, that's not a typo). And I'm not talking about people who've maybe done a puzzle 5 months ago and improved a lot since then; I'm talking about players who regularly do puzzles and also regularly play games.

I've also seen people with a 1691 puzzle rating attain an actual game rating of 1759 rapid. I've also just seen someone on the forum a few days ago with a puzzle rating that consistently climbed into the 1700s and even 1800s before falling down to an average of 1600 complain about being perpetually stuck at 1300 classical.

The puzzle rating / Game rating of some players I've observed, in table form:-

Puzzle Rating Game Rating
1396 1372 (rapid)
2178 1414 (rapid)
1691 1759 (rapid)
1634 1393 (classical)

All of these players play both puzzles and games on a weekly if not daily basis (in the case of the last person, he just played 400+ puzzles today). This isn't a case of someone with an inactive puzzle/game rating from 5 years ago, these are actual puzzle / game ratings from active players.

I'm not naming the exact players since I'm not sure if Lichess would consider naming them on a forum post to be harassment. But I'm sure that everyone who's played around at the lower levels of Lichess has seen similar players before. These aren't rare standout players I've combed through a million games to find, they're people I've played literally today, or people on my best wins / worst losses list (aside from the poor guy who's been playing hundreds of puzzles per day; that guy's someone I've found on the forums). I've selected them out of maybe 20 or so players as illustrative examples of the point I am making, but I'm not going out of my way to find the weirdest players I can find, and I'm sure anyone who's around my level can reach around their game history and easily find a dozen or so players with similar rating disparities.

This brings me to the point I'm trying to make: if your puzzle rating doesn't correspond to your actual game strength, then what gives? Didn't your chess club, your local grandmaster, your favourite twitch streamer, your chess forum, and social media all tell you the only thing you need to get good at chess is puzzles?

I've thought of 3 explanations that kind of explain why highly-rated puzzle players are still bad at chess, but all of them feel unconvincing, even to me.

1. Chess really is all about tactics; the reason highly-rated puzzle players are bad at chess? It's because Lichess puzzles don't actually reflect their tactical ability, which was always terrible. These players just managed to get an inflated Lichess puzzle rating because they're using some kind of non-tactical calculation to defeat puzzles. Maybe they just got really lucky and guessed the right move a couple hundred of times without calculating the entire puzzle out in their heads.

2. Lichess puzzles really do reflect your tactical ability; but chess is actually about more than just tactics, even at 1300 Lichess. The players with a "tactics rating" of 2100 might suck at basic opening principles so badly that they just implode on the chessboard before they can bring their tactical skill into play. They do have a finely-honed sense of tactics, but they just never get to use it because they don't know how to get their pieces into position.

3. A player with a high puzzle rating really does have a high tactical strength, and chess really is all about tactics. Normally, players with high puzzle ratings / tactical ability would be good at chess, but the players you've observed are outliers who are unable to bring their tactical strength to the board because of some mysterious psychological hangup about how they play. When they play games, they play on autopilot, and can only really think about chess when they're looking at the puzzle dashboard.

All of these seem like terrible explanations.

With the first explanation, even if you're "guessing" your way through puzzles all the way up to 2100, that still reflects a high level of intuition and game sense. Maybe you suck at calculation relative to your puzzle rating, but you should still have a good enough sense of intuition to outplay someone who can neither calculate nor intuit the sort of positions you can find.

With the second explanation, I've analysed hundreds of my own games and I can confidently say that at my level, all games are filled with obvious tactical blunders from both sides. Even if I play the opening perfectly and my opponent is down -4 or -5 in the opening, I will always find a way to blunder away that lead by missing two or three tactics throughout the entire game. It's obviously, trivially true that chess is all about tactics at my level.

The third explanation basically sounds like bad pop psychology to me. If someone is dedicated enough to suffer through thousands of chess puzzles to improve their chess games, they're not the sort of person who would sleepwalk through a game of chess. Saying that good puzzle players lose to bad puzzle players in chess, is like saying the guy who hits the gym twelve hours a week can lose to a coach potato who only goes two hours every week in a powerlifting competition, because the first guy suffers from a mental block where he can only lift weights at the gym and can't do it outside the gym. This idea just sounds like plain nonsense.

So, logically, everything seems to indicate that chess puzzles ought to be the key to chess improvement. I can't think of any good explanation for why a good puzzle player can still be a bad chess player.

But all of the logical thought in the world doesn't matter in the face of empirical data. This is a bit like Zeno's paradox of Achilles and the Tortoise; it doesn't matter how good your theoretical framework is if it cannot stand up against empirical data. A single data point about a man walking across the room breaks the Achilles paradox. And there are probably thousands of data points on Lichess that break the "puzzles lead to improvement" paradigm.

I don't have any answers here; if I actually had the answer of "how to improve at chess" I probably wouldn't be rated 1200. But I do know that whatever the answer is, it probably isn't just or even mainly about puzzles. If anyone has found the holy grail of chess improvement and wants to share it in this discussion, please let me have it.
Well
In games you dont know that there is a tactic
While in puzzles you know there is a tactic and you spend more time finding it