Pattern I noticed in streamer games

Intro

I am a lazy bum, so rather than play chess, I often prefer to watch other people do it. And while I generally like the theory videos more (what happened to those? Everybody's doing speed runs now!) I watch other stuff as well. And I noticed something about the people playing. I don't like it's like that, but what to do? Brains will brain.

Thinking, Fast and Slow

Daniel Kahneman wrote a book called Thinking, Fast and Slow that, to my shame, I did not get to read yet, but I know the gist of it because friends and people all over are telling me about it (Including the people behind Jev, whose company is called System One...). So I am going to share it with you, pretending I know what I am talking about.

Basically, the brain works in two systems, the fast one that was built through trial and error and instinct and the slow one that involves conscious thought. The fast one is REALLY fast. When you avoid a swerving car or take your hand away from something hot, that's system one. And when playing chess, that's the part that gives you the intuition to find a move or not.

Pattern

So back to the streaming masters (LOL, yeah, a weird turn of phrase). I see them talk about "I like this move", "The rook wants to go here", "How about this, does it work?". That's system one at work. They have the intuition to select the moves, and only then do they think about the consequences.

I am not really a habitual chess player. I don't play continuously to nourish and grow my intuition or pattern recognition. I just occasionally play after I get a taste from a video I watched or a blog I've read or a study I enjoyed. This is the mental equivalent of exercise - which I know is really good for me, both physically and mentally - which I also refuse to do because I don't enjoy it at all. However, it's obvious that when I am in some positions I am blind to moves, patterns that had I trained my intuition would be clear as day.

But now I finally understand grandmasters saying they don't see a lot more moves than the average chess player. It's a bit of a lie, too, but not that big. What they have is intuition and visualization. Their brain picks candidate moves not only really fast, but also of high quality. They are not aware of why it happened! Yet their brain is working in the background based on the "muscle" they have trained relentlessly. Then the visualization skill - which I also don't possess - allows them to "see" the board after the move and then pick other good candidates moves from that position.

So while I struggle to even find the candidate moves, they already have candidate lines to think about. Only then does the slow system get to work and find if things are actually promising or not.

Algorithmic side quest

This makes me think of computers. Hell, I can't help it, I am a dev. But with the (heh!) intuition I got about how things work, I can now compare various chess engine designs with how people think.

Stockfish sacrificed the accuracy of candidate moves for speed. It is the equivalent of a very basic intuition of what good candidate moves are, but imense capacity for visualization. Poor candidate moves, but many long lines that it can examine.

Leela Chess Zero maximizes the accuracy of candidate moves, sacrificing speed. That's why it looks so slow, yet it generates some very interesting lines that Stockfish would not even dream about.

Other chess engines may be somewhere in between. Dragon (from Bruno LUCAS), that old thing, is an example. When watching top chess players battle it out in classical games, Dragon was the one often picking the same moves as them, not Leela, not Stockfish.

I believe engines should have several sliders instead of the normal "level". They should pick the intuition (which would probably change the model - like Maia 1800 instead of Maia 2500), the visualization (how deep they can go on a branch), the memory (how many branches they can explore at a time) and only then something like thinking level, which would be the slow system considering the candidate moves.

Well, food for thought.

The missing link

I already see some readers go "I play bullet all the time, I train with puzzle rush and racer a lot, but I still lose most of the time and my rating stays the same!". There is a missing link here. I can't pretend to know it for sure, but I suspect it's actually looking at your mistakes and understanding where you went wrong. This is system two at work to complement and enhance the learning of system one. If you don't learn from your mistakes, you are doomed to repeat them.

But the brain has a mechanism of trying to avoid negative things. The last thing it enjoys doing is examining in depth how much of a blunder you made. And it wasn't just one, there were several. And they were so incredibly bad, even when you thought you played a brilliant move that won you the game. Someone with half a Stockfish brain would have wiped the floor with you, idiot!

No, no one enjoys to go through that. So here you are, not only reading that you have to do a lot of exercise, but even the slow thinking that most people associate with the game of chess has to be focused on the painful moments. No pain, no gain? Hell, I hate what I wrote already and I am the one writing it!

Ideas for training

As much as I dislike it, I have to agree that the fast system is the thing that we train the least and that we need first. Playing a lot, training fast moves, but also taking the time to retrospect where we went wrong and what would have been a better move.

So what can be done to make this less painful to train?

Right now I am thinking of a list of candidate moves, something chosen by an engine at every level from 1 to 5, let's say. The goal of the game is to show the player a board and the list of candidates. They have to click multiple candidate moves, then click Next. They would be scored not only on whether they caught good moves rather than bad, but how many good ones they found from the ones that are not losing and how much time it took.

The problem with this is that it has no depth. There is no payout at the end, just a numerical score. You can't "play" like this. Maybe play the best move out of the candidates you chose and go on? I don't know. The other problem is the disconnect between your mind seeing the move and the notation. Maybe one could make the moves on board, have only a Next button to click. You make a move, it gets recorded, the position resets. Kind of like drawing arrows on Lichess.

But then there is the question of analysis. After a run, maybe go into a kind of Learn from your mistakes phase?

If you want to contribute with ideas, I've created a GitHub issue.

Conclusion

If - and that's a big if, as I realize some people play chess just for the adrenaline and don't really care about learning - you want to improve in chess, intuition comes before thinking. It's not like in the movies! If you are very smart it doesn't mean you will beat everybody else at chess. You have to work on it.

But work smart.