The Right Feel

When I decided to learn poker, I learned a bunch of cards like this.

But even after I knew the cards well, I wasn't able to apply the knowledge when I played. My Anki-knowledge was totally separated from my real-world knowledge.

The reason they didn't transfer was feel. When you play poker in real life, you think "the guy who plays first raised, and there are three guys between me and him, and I have J9s." Poker in real life has a smell, a feel, and an appearance. You see the colors of the cards, of the felt, and feel the pressure of decision.

None of this will trigger memories of your terminology-stuffed 20-pt Arial black-on-white Anki card that you were reviewing at 3am last January.

In other words, this:

won't naturally trigger your memories of this:

You avoid this by making your cards feel more real. For poker, use a video of an online hand, cut right before the hero acts. Practicing making hero's decision from a video will make the intuition much more likely to stick with you when you actually play, even in person.

If you code in VSCode with the Fira Code font and the Monokai color scheme, use those to make your cards [0]. And especially for syntactical cards, you really want to practice typing them into the computer, to get the muscle memory.

For chess, use diagrams of actual chess boards. If possible, avoid contrived tactical puzzles. Using positions from real games will make the locations of the pieces feel more realistic.

For psychology books, use vivid characters from movies or TV-shows instead of the template "parent" from the book. Ron said this to Hermione; how should Hermione respond?

For languages, don't use sentences from textbooks. Get real sentences from the source material you want to read.

When I started SRS, lots of this stuff felt like fluff. Why should I bother getting a video of a poker hand where this happened instead of just writing out the sequence of events? But it turns out to be crucial if you want your knowledge from Anki to bubble up in the real world just when you need it.

Notes

0: Perhaps the best thing, much like with machine learning, is to jitter these parameters on every review: last time you reviewed in Monokai with Fira Code, this time in Solarized Light with Consolas, so on and so forth.