playing against (weak) computers

[gordonr]

The point is that with a puzzle you know that there is some move that is much better than the others. That makes it easier to find.

That's a good point and it used to put me off doing puzzles. But then I heard a GM comparing it to a tennis player practising their serve without an opponent. Sure, it's not the same as a real game but if you can't get it right when it is easier, you're not going to get it right during a game.

If being good at tactics is a combination of: tactical alertness, pattern recognition, calculation, visualisation, etc. then it's fair to say doing puzzles won't train all of these equally but it may still be beneficial to train some of them. I'm guessing "pattern recognition" is one main focus of doing lots of puzzles.

[Ras]

I'm guessing "pattern recognition" is one main focus of doing lots of puzzles.

Even if one misses them when the opponent makes a mistake, it's still helpful to build one's own position up with such patterns in mind, e.g. recognising when one is about to overload a defender and not doing this in the first place.

[Dann Corbit]

Puzzles did help me for sure. But you're right, during the game, no one is going to tap your shoulder and tell you: hey, you have a winning tactic here!

Play like you have a winning tactic for every move. You just have to find it. Oversimplification, of course.
But it is a summarization of:
If you find a good move, look for a better one.
Some things my dad taught me:
Look ahead at least seven moves on offense.
Always consider at least four moves on defense.
Never, never move until you are sure of the reason for your opponent's last move.

Unfortunately, some of the most beautiful tactical moves are actually refuted by modern hardware (e.g. it is better not to take the gift).
Analyzing Morphy's games or the evergreen game or the immortal game with a computer is just sad.
But I find that going over them with my mind is exhilarating.
I love tactical puzzles and I have literally millions of them.
I do not have any idea if they make me a better player, but I love to try to solve them.

[DomL77]

Its best to work on your calculations and not just 1 move puzzle solutions. Find the continuation of solutions that go for 2-3 moves long like the Encyclopedia of Chess Combinations, Woodpecker method etc..The stronger your opponent the less tactics you'll see.

Work on your mistakes make in a game via the puzzle themes. You can use the lichess theme option or the offline version on github https://github.com/brianch/offline-chess-puzzles and focus on a theme like defending etc..

Strategic/middle game part good books to read are - Mastering Chess Strategy by Johan Hellsten or How to Reassess your chess by Silman

Endgames you can find many online or do endgame sparring matches against an engine.

[lucasart]

I like to play against the strongest programs on my high end hardware. Of course, they bash my brains in like an angry elephant stomping on a grape. I have no hope of victory or even a draw. My goal is to "last a while" or try new ideas.

But after enough pounding I eventually lick my wounds by playing a very weak program like the old version of Golem on a single core.

That method did not work for me at all. The problem is that it teaches you to play passively, just trying to build an *almost* impregnable fortress and cower behind it for as long as possible.

I found that playing active chess, against equal opponents, is what improved my chess.

There is, however, an area where playing against invincible computers is helpful. Learning to convert winning positions.

Sometimes, at the end of a puzzle, once I got the "clearly" winning position, I wonder: how do I win this? Not against Rufus and Dufus, but against Stockfish. If I understand how to win, in a way that eliminates counter play completely, not even Stockfish should be able to swindle me.

I play, one move at a time, slowly, no clock, but no take backs. It's a good way to learn, even when you fail, you should try to understand what the flaw was in your logic, why the computer's recommended method is more effective, explore alternatives etc.

Same can be done after a frustrating loss, when you had an obvious win, and blew it under time pressure, for example. Start from the winning position, and carefully try to beat SF from there. Removing the stress from time pressure, the ego element from human vs human chess (you know that SF is not even in the same league as humans, nothing to prove there). That way you turned frustration into learning. Much better than mindlessly hitting the rematch button, and going into another game with frustration and vengeance in mind.

[jefk]

'Doml77' mentioned:

lichess theme offline version on github https://github.com/brianch/offline-chess-puzzles and focus on a theme like defending etc..

while positive feedback in this forum usually is a rare thing (besides nitpicking or other, sometimes valid
or less valid criticism), this is an excellent piece of software (with elo and theme selection and
still plenty of puzzles to choose from).
Training tactics via repetition within a certain theme is the basis of the Dutch steps method
(at least steps 2-5 or so) and is the way to improve pattern recognition (via recognizing
a certain tactical theme). ofcourse it's not like in a real game (where you don't get
a hint about a tactical them applicable) but after a lot of training and repetition you
build up a subconscious (tactical) pattern recognition which also will be useful later (in
games) On the longer run this seems to be more effective than doing random (tactics)
puzzles as online on chess.com or lichess (which also are different from real games
anyway because you know it's a puzzle (in a real game maybe one in five-ten moves
contain important tactics, with maybe two or three critical moments in a game
and then you can/should take more time; but you never know in advance when this
is going to happen in the game... post mortem analysis helps a little but also sometimes
is discouraging with a super engine because you always notice these tactical moments
only in retrospect. For adult improvers if you know the basics (lets say step 2 also as chess
tutor 2 on Shredderchess) i'm not sure how effective it (this offline lichess them based
training) would be for further improvement but at least for me it's worth a try

[DomL77]

'Doml77' mentioned:

lichess theme offline version on github https://github.com/brianch/offline-chess-puzzles and focus on a theme like defending etc..

while positive feedback in this forum usually is a rare thing (besides nitpicking or other, sometimes valid
or less valid criticism), this is an excellent piece of software (with elo and theme selection and
still plenty of puzzles to choose from).
Training tactics via repetition within a certain theme is the basis of the Dutch steps method
(at least steps 2-5 or so) and is the way to improve pattern recognition (via recognizing
a certain tactical theme). ofcourse it's not like in a real game (where you don't get
a hint about a tactical them applicable) but after a lot of training and repetition you
build up a subconscious (tactical) pattern recognition which also will be useful later (in
games) On the longer run this seems to be more effective than doing random (tactics)
puzzles as online on chess.com or lichess (which also are different from real games
anyway because you know it's a puzzle (in a real game maybe one in five-ten moves
contain important tactics, with maybe two or three critical moments in a game
and then you can/should take more time; but you never know in advance when this
is going to happen in the game... post mortem analysis helps a little but also sometimes
is discouraging with a super engine because you always notice these tactical moments
only in retrospect. For adult improvers if you know the basics (lets say step 2 also as chess
tutor 2 on Shredderchess) i'm not sure how effective it (this offline lichess them based
training) would be for further improvement but at least for me it's worth a try

Puzzle rush is good for intuition training and faster time controls. Deeper calculations training that involves 3-5 minutes calculation is great for longer time control and planning what positions you want in the middle/endgame. Encyclopedia of chess combinations or finding mate in 3 puzzles with 3-5 minutes to solve is torture, even if you don't solve them

The guys from chessdojo stated that if you can't solve more than 40% of the puzzles then its too hard and if you solve more than 70% its too easy.

[chesskobra]

There is a large collection of endgame and opening positions here https://github.com/jhorthos/lczero-trai ... ning-Books. The positions are evaluated by an Lc0 net on a scale -1 to 1 (q value - I don't know if it roughly corresponds to the probability that Lc0 wins a position using the particular net). You could use unbalanced positions at different levels of the q value to see if you can win these positions against strong engines. I am using these positions to test different Lc0 nets and weaker (or weakened) engines. In Stockfish 16 vs Stockfish 16, one side wins 100% of the endgame positions at the higher end of the q scale in about 30 moves on average, but some weaker engines struggle to win the positions against Stockfish 16 even in 40-50 moves. The Stockfish eval at the higher end of the q scale is often 5+.

[Uri Blass]

Lichess has a very good puzzle collection (essentially of infinite size). I find it very enjoyable, but it does not improve my play. The point is that with a puzzle you know that there is some move that is much better than the others. That makes it easier to find.

Puzzles did help me for sure. But you're right, during the game, no one is going to tap your shoulder and tell you: hey, you have a winning tactic here!

Playing games is important for that. But you have to play slow chess, like 15+10 or more. It's the only way you have time to apply a systematic though process on*every move* (even the seemingly obvious ones like recaptures):

• What are the pros (threats) and cons (weaknesses) of my opponent's last move
• Scan all forcing moves, even dumb looking ones. If nothing works...
• Look now at patient improving moves. You must force yourself to think of several moves, not just the first one that comes to you. Select the best one. This requires mental organisation to remember everything analysed so far, what the outcome of the lines were.
• Now you have your move. Verify before playing it. Try to solve the puzzle from the opponent pov after the move. If you discover a problem, go back to the previous step.

Needless to say that it takes time. You can't play like that in blitz.

I think the problem is that people see strong players playing bullet and blitz, so they think they should do that to get better. But it doesn't work. The reason you see these GM posting videos and streams of fast chess, is because it's more exciting to watch, attracts more views (hence more ad revenues). What is missed here is that they didn't reach that level by playing bullet chess like an FPS video game. They built their level by playing slow and boring chess. Analysis, study etc.

Analysis of one's games is of course extremely important. The first thing is to ban the rematch button. No matter the result, frustration, don't play consecutively. Take time to analyse. Force yourself to do the mental calculations before asking Stockfish for the answer, etc.

There are people who improve also when the only time control they play is blitz.
improving only by playing is impossible after you reach some level regardless of the time control and you need to analyze your games in order to understand the mistakes that you did.