Make new mistakes: why your flashcards should be your own blunders

You know the position. Maybe not the exact same one, but close enough that the same idea gets you every time, and you've said “again?” out loud at your screen more than once. Everyone makes mistakes. Getting better mostly means your mistakes start to change. That's what we mean by “make new mistakes”. And there's decent research on how to get rid of an old one for good.

If learning science agrees on one thing, it's this: trying to recall something makes it stick better than looking at it again. In the study that made the effect famous, students read a prose passage and then either reread it or wrote down what they could remember. Five minutes later the rereaders looked better. A week later the students who had tested themselves remembered about 61 percent of the material. The rereaders were at about 40.¹

It holds up beyond one study. A meta-analysis of 159 effect sizes puts the average advantage of testing over restudy at half a standard deviation, and bigger when the test makes you recall the answer instead of just recognizing it.² Retrieval even beat concept mapping, which is an elaborate study technique, on both fact and inference questions. The students themselves had confidently predicted the opposite.³ And the most cited review of study techniques rates practice testing as one of only two “high utility” methods.⁴

In chess terms: a flashcard hides the answer and makes you find it. So does a puzzle. Clicking through the engine line again after a loss is rereading.

The other “high utility” method in that review is distributed practice, which just means spreading your study out over time instead of cramming it into one sitting.⁴ A synthesis of 317 experiments found spaced study reliably beats cramming. It also found the best gap between sessions depends on how long you want to remember the thing. Need it next week? Review in a day or two. Need it next season? A couple of weeks out.⁵

We should be upfront about a limit here. That synthesis is about verbal recall. Nobody has worked out the ideal schedule for chess positions, so any app (ours included) is applying a well-supported principle, not a proven chess-specific interval. The principle is simple enough. Get a position right and it comes back later than last time. Get it wrong and it comes back sooner. Every serious spaced-repetition system works this way, and the exact numbers are mostly engineering choices.

Interleaving

Shuffle the deck

Doug Rohrer and Kelli Taylor gave students math problems either grouped by type or shuffled together. The shuffled practice felt worse and scored worse during the session. But a week later the shuffled group did far better. They'd practiced the skill the grouped drills never test: figuring out what kind of problem you're looking at.⁶

Most tactics trainers are grouped by design. “Forks, 20 puzzles.” Once you see that label, you already know what to look for, and that was the hard part. A deck built from your own games mixes itself. An opening slip, then a middlegame tactic, then an endgame you misjudged, in whatever order you played them, and nothing telling you what to look for. It's harder, and a lot closer to a real game, where nobody warns you a fork is coming.

A surprising amount of chess skill is memory. Show a master a real position for a few seconds and they can put it back together almost perfectly. Novices can't. Show both of them a random jumble of pieces, though, and the master's edge nearly disappears.⁷ The explanation that has held up is that strong players carry a big library of familiar patterns, later refined into flexible templates, and they see every position through that library.⁸

What matters is which patterns go into yours. A puzzle collection gives you positions somebody else reached. Your own games give you the ones you reach: your openings, your opponents' habits, that pawn grab you keep thinking is free. A card made from the moment your last game turned is a pattern you will see again. It comes with a story, too: who you were playing and what you were thinking.

Mistakes help, too. Janet Metcalfe's review of the research concludes that making an error and then getting corrected is better for learning than never being allowed to make it, especially when you were confident you were right.⁹

How Chessalyz does it

Every questioned moment becomes a card

When Matty questions a moment in one of your games, you can turn that position into a flashcard in one click. It gets a name and a link back to the game. The review gap stretches each time you answer a card cleanly and shrinks when you miss it. The deck mixes itself because it's drawn from whatever you played, and since every card is from your own game, you remember where it came from.

For the patterns you haven't run into yet, there's the skills trainer. It serves puzzles at your rating, and the hints point you at the idea without giving away the move.