ARE CHESS TACTICS TOO HARD? 📈
The Eighty Five Percent Rule for optimal learning and my results after 10 years of use.
In my last newsletter, I mentioned the article ‘The Eighty Five Percent Rule for optimal learning’. I think it’s a really interesting article in relation to chess training. Here is the abstract of the article:
Researchers and educators have long wrestled with the question of how best to teach their clients be they humans, non-human animals or machines. Here, we examine the role of a single variable, the difficulty of training, on the rate of learning. In many situations we find that there is a sweet spot in which training is neither too easy nor too hard, and where learning progresses most quickly. We derive conditions for this sweet spot for a broad class of learning algorithms in the context of binary classification tasks. For all of these stochastic gradient-descent based learning algorithms, we find that the optimal error rate for training is around 15.87% or, conversely, that the optimal training accuracy is about 85%. We demonstrate the efficacy of this ‘Eighty Five Percent Rule’ for artificial neural networks used in AI and biologically plausible neural networks thought to describe animal learning. 1
It might also be useful to listen to the explanation by Andrew Huberman (neuroscientist and tenured professor at Stanford):
The video should start at the point (1:56) about the optimal difficulty for learning.
After reading and listing to the Huberman podcast my first thought was that chess tactics trainers would be a good sandbox for comparison since it measures error rate for the training. So I have looked at my stats on chess.com over the last 10 years:
I have progressed from around a 1400 rating to now 2600 after 2,516 puzzles. The error rate is a staggering 43 %. That’s 27 % higher than the recommended should have been for optimal learning according to the study. I can’t stop wondering if my progress would have been different/faster if the puzzles would have increased in difficulty with 85 % correct answers as a constant.
It’s not just on tactics trainers chess players try to solve hard problems. Often it’s recommended to solve Kasparyan studies or just other studies in general:
(For an easier introduction to studies check out ‘Endgame Studies 101’ by IM Kostya Kavutskiy.)
I did about two pages and solved maybe one or two studies correctly before I concluded this was simply too hard. It is also my impression in the chess community that we recommend really hard books to each other. Calculation by Aagaard Domination by Kasparyan, Perfect Your Chess by Volokitin, and John Nunn's Chess Puzzle Book by Nunn. And now I wonder if grandmasters could score 85% on these books? (I will admit that my own books also contain some difficult puzzles and are not for beginners.)
Nunn’s book actually has a scoring table in it and if you are rated 1855 FIDE (approx. my rating) you are expected to get 26 points out of 120 on the test ~ 21 % correct! The prober rating target for the book according to the table combined with the eighty-five percent rule is players with a 2573 FIDE rating! That is around a top 300-400 player in the world. Coincidentally GM Nunn is rated 2568, so the book might be a perfect fit for himself.2
I talked to Martin Bennedik, who runs the chess tactics site chesspuzzle.net. He tells me that an 85 % success rate would equate to solving puzzles 300 Elo below your current puzzle rating.
That is actually one thing that is possible to test on chesspuzzle.net using the filter-mode.
I have tested it out and I had a session with an 87 % solve-rate (29 correct / 4 failed), which actually was kind of a confidence boost, and something I will experiment with more in the future.
Searching for the correct and optimal difficulty of their puzzle diet could be something many chess players might benefit from. Finding the sweet spot where the puzzles are stretching your mind while not above one’s current capabilities. It is here where I think many, me included, take on hard books/tactics in order to boost our egos or maybe just because it is what others recommend to us.
Taking on hard puzzle sets will put chess players in the anxious territory in the above figure, while solving mate-in-one-puzzles might be boring and useless unless you are a beginner. And from what I see in the chess community many are in the anxious zone, while a few are in the boredom zone. Finding flow in your training is the key according to The Eighty Five Percent Rule if you want to optimize learning gains.
It is also something I want to work on and implement when I publish The Tactics Ladder series since it will be an opportunity to find the right step on the ladder when the whole series is published ranging from 800 Elo to master-level. In the book series, I set it as a goal to solve 90 % of the puzzles after the third cycle.
Post a comment with your thoughts about finding the right difficulty below, and if what you think the correct solve-% should be.
If you liked this post please share and/or subscribe to this newsletter.
/Martin
Wilson, R.C., Shenhav, A., Straccia, M. et al. The Eighty Five Percent Rule for optimal learning. Nat Commun 10, 4646 (2019). https://doi.org/10.1038/s41467-019-12552-4
FIDE Rating, GM John Nunn, https://ratings.fide.com/profile/400017
I agree the focus is on too hard puzzles for another reason: the easier tactics are far more common. At a certain point in your training you get very complicated tactics in wild and confusing positions.
I checked a few hundred blitz games for tactics, expecting to find some of those bizarre patterns but almost non came up...
Difficult tactics do help to concentrate and visualize, but your blindfold books are a more enjoyable tool for that to me :)
For tactics I use lichess where you can filter for easier tactics, try to stay focused and get as high a rating as possible. It is similar to a real game in the sense that you can do very well and throw it all away with 1 blunder which you should have seen if you were not a lazy detective at the end of your game.
Some comments on finding the right difficulty level on problems. I believe that the 85% level makes sense. The first question I need to ask is why in the world is a person doing these puzzles. If a person is doing these puzzles I assume it is so, they can have a better understanding of what happens in a chess position and problem, and become more proficient at using these skills to solve slightly more difficult problems as they learn from the ones they just did.
Before I go any farther I would like to state what I believe is one of the major reasons that students in general are not able to do anywhere near 85% of the problems in a given section. Take a trip back to the start of the section in a chess book. It may be on knight forks and shows some basic ones, a few a little more difficult and then the one’s that are pretty difficult and that not many students understand. The writer has shown the basic problems (level 1), a few the next level on (level 2 and maybe a 3), and then the problems jump to levels 8 and beyond.
Where is the opportunity for instruction to from go from step 3 to 8 so someone begins to understand the difference between the problems and what sort of things to look for in problems. Not last but least learns how to go about solving it. Also, are the problems leveled in such a fashion as to let the student learns from each set of problems and grow while they are working. What we do not want to happen and all to often is, students are asked to urn a four minute mile when you haven’t run a 6 minute mile yet.
I just was working on a tactics book from a well-known publisher who said something to the effect, If you find these problems too difficult then pay some more games and come back and they will be easier. If all I had to do was to play some more games and then I could be a tactics genius I would be one already. Authors need to make their problems fit with what they teach. And if they do not teach these ideas that are needed for the problems do not a large percentage of these hard problems.
I will use one of my own personal examples. I am pretty good in math, at least in high school. I really had to listen to the lesson to understand the material. I am not that way in chess. Let people learn while they do the problems, that is the reason for the problems. Don’t write chess books that have problems that everybody can do and then ones only the naturals can do. There are more middling chess players out there like me that there are naturals.