Everything I Thought I Knew About Improving At Chess Was Wrong
Maybe improving at everything else too
What I’d always believed (and assume been told) was that the way to improve at Chess was by learning to be systematic in analysis, acquiring understanding of deep positional principles and a systematic, disciplined approach to manually doing alpha-beta pruning yourself when confronted with a position, and start with long time controls and work your way down as things become more automatic. What you don’t want to do is learn a bunch of silly little tricks which are only rarely applicable and do so at quick time controls which aren’t developing much of any discipline.
That advice is about 80% wrong. There’s some utility to basic positional formulas like improved piece weights and of course looking ahead multiple moves is a core skill especially at long time controls. But most of the rest of the advice is wrong, more based on a fantasy of what Chess is all about that what it actually is. In reality Chess, first and foremost, is a game of tactics. A reasonable definition of a game being ‘tactical’ is ‘If you slap together a crude manually designed board evaluation algorithm and an alpha-beta pruner it will obliterate even very strong humans’. Another way to think about it is how close the game is to a wordfind. Chess is obviously not a wordfind, but it is a tacticsfind. At each move the player finds the best tactic they can, resulting in a stronger or weaker position. How good of a tactic they find is directly correlated to how many moves they’re looking into the future.
Of course as one gets stronger seeing basic tactics becomes more of a given. Probably around 2300 playing strength the main determiner of Chess playing strength switches from how good you are at tactics to how good is your opening preparation. (Sorry, couldn’t resist.) But at my level the differences in board value caused by subtle positional differences are so small compared to swings caused by tactical misses that they’re a bit lost in the noise. What I’ve been getting a lot of benefit from is using this tactics trainer for a few minutes a day (mostly instead of checking social media). The key to getting improvement out of it is not to spend lots of time on each puzzle but to give yourself only a few seconds for each one, then play through the solution when you miss it, come up with an explanation of how your could have figured it out, and commit that to memory. The more of them you get through, the more you’ll remember. Yes it’s only teaching you silly tricks, but that’s most of what Chess is. Positional concepts themselves are often just acknowledgement of tactical potential which isn’t quite manifesting yet, like noticing when pieces aren’t defended even when nothing’s attacking them, or they’re attacked even when they’re currently adequately defended.
Even with only a few seconds to evaluate each position It’s possible to do quite a bit of lookahead, particularly once you get good at pulling out candidate moves. Where human lookahead deviates a lot from slavish alpha-beta pruning is that especially in very tactical positions you often see something which almost works which then suggests something you didn’t initially consider as a candidate move, especially if it involves distracting an enemy piece so it’s no longer defending something or in the way of something. It’s also common for something to be an obvious candidate move a few moves down and the better move order to be the less obvious one.
My own natural skill level at Chess isn’t very high, owing to a serious lack of visual memory. Okay okay, it’s a low natural skill level on the scale of mathematicians, and I can ‘visualize’ some things very well but it’s all 3d not 2d and not very useful for 2d board games. In any case, I’ve played enough Chess in my life that I should be a lot stronger than I am. But after doing tactics training for a few minutes a day for a few months my estimated strength by tactics hovers around 2200 or so depending on how long I take for each one and how tired I am my current caffeine and blood sugar levels. Playing strength seems to be around 1750 or so. It’s weaker because I’ve barely been playing actual chess games, resulting in my time management being awful and also my consistency at easier tactics isn’t what I’d like. It would be nice if the tactics trainer threw in a lot more ‘easy’ problems so there were strong rewards for getting easy problems right almost all the time instead of being so focused on getting hard problems right half the time. It would also be nice if it were more consistent at keeping the problem going through the hard to see move and making the opponent always force you to play the hard to find move even when it’s a weaker line for them. But tactics trainers are hard to write, and that’s by far the best one I know of.
Around 2300 is where tactics puzzle seem to switch over to needing some positional understanding, in that the resulting position from the right series of moves is superior for some subtle positional reason instead of conferring a clear material advantage. It’s interesting watching Chess commentary after gaining some basic skills. It becomes very obvious which commentators are strong players themselves and which are weak players using Stockfish, because the weak ones completely skip over obvious-to-strong-player variants which don’t work out in the end for very involved tactical reasons.
Next you’ll probably ask whether cheesy math team problems are better for improving at mathematics than learning actual deep concepts. To that I say that teaching integration should skip ahead to Laplace transforms and not waste all that time on techniques which get subsumed by it.