lichess.org
Donate

The Woodpecker Method, Revisited

There is logic behind the woodpecker method, but the way the author lays it out is inefficient. If you do a puzzle, and come back to it 3 days later, that position will be ingrained more than if you did it only once. When we force ourselves to remember we ingrain the knowledge further into our long term memory. However Doing a whole bunch of problems 7 times over is really just wasted value. The woodpecker method has it flipped where the review periods get shorter, when they should be getting further apart (i.e. instead of waiting 4 weeks then 2 then 1 for puzzles to be reviewed it should look something more like 1 week, 2 week, 4 weeks). Overall, it does work, just not as well as it could and the woodpecker method isn't efficient enough to make a significant difference.
Our brains are very good at making associations. Doing new puzzles is not as beneficial when learning patterns because you have to relearn the pattern. Sure you can look at puzzles with the same pattern. but when you don't even have the grasp of one simple mate in 3 puzzle for example, seeing it again in a different position is like restarting. Revisiting those puzzles in a puzzleset, helps us remember the move sequence.

At least that's how I was able to become pretty good at tactics despite not seeing them as a kid. Anyways thanks for the content.
I also wondered about the Woodpecker method. It is stated that:

"Hans Tikkanen, the author of the original Woodpecker Method, got three GM norms in a seven-week period after using the method."

But let us not forget that Hans Tikkanen also co-authored the book. This would mean he did an enormous amount of research to find positions for the book and then he did the book exercises assiduously as well. This is not the same as a reader doing the book, the methods and just the precise positions in the book. Tikkanen did much more. Plus we don't know what other extra chess study he did for those GM norm events. This is not to criticise him. I am sure the method works up to a point and with opportunity costs as @CheckRaiseMate writes. (@NDPatzer also writes about cognition and chess cognition.)

I assume Tikkanen sifted games and got excellent, representative positions. We might almost call them archetype positions and archetype tactics. As such, they would be very representative of a lot of positions, situations and tactics that you would meet in real games. And the general exercise of working hard at calculation and visualization will fit the person to analyze more atypical positions and tactics as well.

I have started using Lichess puzzles in a high volume way (and in other ways too plus doing other chess training in a self-designed approach). It is far too early to tell if this approach will make me a better chess player. The high volume part of the training involves doing many easy puzzles first. For example, rapidly doing random 1 move checkmates on the easiest setting. This sounds pointless perhaps but I have already formed the opinion is not pointless for a slow-playing, low-rated player like me at 1470 about in Rapid 15 + 10.

I have noticed that such training makes me rapidly flick my eye over the whole board, better taking in the whole position. I am looking for not just checking and therefore mating moves but also for my supports for the attack and for the opponent's defences which prevent/refute specific attacking moves. In addition, I can notice my mind templating my control of squares around the enemy king and templating its escape squares or templating that it is an obvious smothered mate. The theory is that doing enough these one move mates (then other tactics and longer move sequences) will make me adept and rapid at seeing in this specific case all one move mates and executing them. But it will also make my mind able to template crucial aspects of positions and develop the habit of rapidly flicking my eyes around, taking in all aspects, and pitfalls, of the whole board.

Then it is on to rapidly doing two move mates on easiest in the same fashion. The theory is that with all possible one move mate patterns well fixed in my mind then two move mates will become easier and the templating happening in my mind will assist me to visualize that one move further ahead. Then the process will go on to 3 move easiest mates. After that I will have to assess where to go next in the checkmate training. And this is only checkmate training. There is much more to chess tactics and chess as a whole of course.

The idea will be to get similarly rapid and "templated", as I call it, in each tactic and motif. I will have to do a lot of rapid puzzles. I am hoping to build up to 1,000 puzzles a week. With one-move puzzles it may prove pointless to work levels up to easier, normal, harder etc. Those levels probably won't exist essentially for one move puzzles. For two moves of any mate or tactic (and beyond) those levels will certainly exist for a low rated player like me.

This is all an experiment and I don't really know if this will improve my tactics and calculation. But I have become so frustrated at my poor tactics and lack of calculation capabilities along with bad blunders and chronic time trouble, I thought I needed to go back to absolute basics and hammer away at the very simplest stuff first. After all, with calculation, if you can't make one step, you can't make two. If you can make two steps you can't make three... and so on. And if you can't make each step reasonably quickly you lose in the time scramble.
I made a video a while ago that tries to give a bit more argumentative bite to the ideas behind the so-called Woodpecker method: www.youtube.com/watch?v=1qicV-jaGFE

The main point of disagreement seems to be with the following:

"If I do the same problems over and over, I run the risk of simply memorizing those positions, rather than picking up on the underlying patterns, which is what I really want. In machine learning and AI this is called overfitting. If I want to learn the underlying patterns it would be better to see the same pattern played out in a variety of positions."

While I understand that concern, I’m not entirely convinced that repeatedly solving the same problems only leads to memorizing positions without grasping the underlying patterns. I'd argue, just like in the video, that mere exposure to the pattern is enough to further cement that particular pattern on an intuitive level.

I’ve noticed some 'chess experiments' surfacing recently, and it would be great to see some empirical research conducted in this area. I imagine the experimental design would be relatively straightforward and certainly feasible.

I've had experiences of instantly seeing a tactical pattern in a game/puzzle, and then being able to recall the 'corresponding' puzzle (same tactical pattern) in my Woodpecker set. While these anecdotal experiences don’t carry much weight in a broader argument, I find them interesting nonetheless.

Regarding the idea that it’s better to see the same pattern in different positions, I might agree to some extent (although even here I don't think that it's immediately evident that variety should necessarily be better). Even if we assume that variety is better, I suppose the issue is then practical: repeatedly finding multiple positions that showcase the same highly specific pattern. Even if this approach is, say, 30% more efficient, the effort required to find such positions might outweigh the potential benefits.
I had the same hypothesis as in your last paragraph. It is a finite time structuring package for those that are driven by the end result. But do not mind the repetitive mind muscle stress, long term professional disease or injuries.

Also there are few sources that do not seem to have same definition of what it might be. I have heard it involved actually analysis de novo the set of positoin challenges and their problem solving.

But this blog version seems to be the most ambient version I would induce from crumbs about the "method" over the past year, listening on Lichess (the forum, mostly, less controlled or restrained scope than hanging around the lobby page since blogs have replace connecting with Lichess as a community of users who like chess, no offense to the blog authors that do bring some oxygen, while also coming from the scope target core culture.

I also think there are elements of acknowledging the part of our brain that is the boss of us, at least as I am writing this, and many other ramblings, I have come to accept my role as note taker, for that non-tip of iceberd part of my mind.. I just try to take notes of the flow of threads, that I could hope to understand and share....

So, that part of mind, has been long denied in chess theories (but not concieved as such) of learning (or improving, such a buzz word... like money back if it does not improve the rating, that learning method).

But I think if one is going to be serious about theories of leanring. They should not just do it half way. and only consider the "motor skills" honing.. (caricature). But the generalizatino problem.

The traget of rating performance is a bit of a tunnel vision here in such discussion.. As I think the more robust theory of learning in the more general problem of leanring not how to augment rating, but how to understand the mechanics of the board in general. I do not think that it is the same problem.

The limited life span scope of the well structured package that the blog does convey, seems limited, in its target (which might be also the appeal. not just that it sustains effort in well structured gymnastics tine scheduling, but the schedule boundaries themselves are not that long). maybe.

I think it is missing, the usual floating variable about the "extent" of the choice of positions to converge one's statisitical learning brain onto and its diversity, toward the problem of generalization. The target world of generalization being undefined (and probably suconcsiouly assumed to be undefinable).

I find that it could be salvaged. but it would need actual studying and injecting of more cognitive science and more than one person contributing new hypotheses of learning, than those that have learned already in hindsight of their preformance expertise record. Hindsight is great for uncovering certain hypotheses of explaination of how we got from before to after. But if the target problem being agreed upon in a "conversation" (at large: a many head one), then all learners can contribute. And we can start to make a theory of learning not just for one slice of the chess eager to improve in all threads of chess activies population.

I think the position set definition. and the pre-learning of those positions as problems to solve, is bit obscure rigth now.. I hope i did put all that I wanted in this rambling. I find the problem of generalization in chess culture not yet well posed. As it might be, for example in machine learning..
@DailyInsanity said in #6:
> I made a video a while ago that tries to give a bit more argumentative bite to the ideas behind the so-called Woodpecker method:

I take your points and have watched your video. I certainly don't assert that I know which method is better: database puzzles (random within a motif) or Woodpecker puzzles (set motifs repeated). Both, as they are presented, in Lichess and in the Woodpecker Method book respectively, have their own kind of limitations.

Lichess puzzle motifs (and tactics and themes) vary in specificity as you point out. A number of Lichess mate puzzles, for example, Anastasia's mate, Dovetail mate and Hook mate are specific enough for repetition training: still variation on a theme but the basic motif is very clear each time. However, many mate types are not in the Lichess list. For example, Greek gift is missing. The other categories are very diverse. As you point out, the forks category covers a wide field. It would be better if one could select "forks" and then say "pawn forks" etc.

On the other hand, the Woodpecker method may be too specific and not cover all tactics and motifs. (I am not sure about this assertion either.) The other issue is that the "noise" or "clutter" of the rest of the position may matter. You suggest it does not but I argue it may. It may matter sometimes in real play in introducing ancillary or opposite wing tactics that play off on each other. It may also matter in real play in the sense that noise or clutter can confuse or mislead a player (or at least slow apprehension and calculation). The "clean" nature of puzzles fails to match the messy nature of real games.

To sum up, database puzzles as currently presented, may be too varied and Woodpecker puzzles may be too specific. Too much variety runs the risk of losing the value of pattern repetition. Too much specificity runs the risk of gaps or holes in the training. The problem of course is the seemingly infinite variety and complexity of chess positions to the human mind, even the trained mind. If I had the time and mental energy I would train in both methods simultaneously. Lacking that, I intend to train with Lichess puzzles for a period, making them as specific as I can and doing high repetitions in the easiest category first to get my speed up and then lifting the difficulty while trying to keep the speed. At some point I hope to switch and try the Woodpecker Method book and concentrate on that for a while. I can't do both at once. My brain would be spread too thin, lol.

It would be wonderful, if the Lichess database or any database could provide radio buttons to tailor puzzles. Then we could select "forks" + "queen" or "traps" + "rook" (as trapped piece) and so on. I am surprised they haven't done this but perhaps the search overhead is too high? Even then, the permutations and combinations could be too high for humans to explore. Who could study them all? The study of chess is a huge field, that's the problem. We also don't fully understand how we learn. Each one of us also does not know the reach or limits or his/her cognitive endowment and life stage. My limits may be near or already here. I might study all I like and get no better from this point. After all, I am 70. But I continue the experiment for now.
I think that the key argument for the Woodpecker method & the earlier MDLM seven circles is Speed. By forcing yourself to see the pattern within an average of 5-10 seconds, you will be focusing only on the key elements of the positions - undefended pieces, tactical targets, alignments, etc... By making yourself see these very fast, you should, in theory, increase your overall calculation / visualization horsepower. You don't get this with a steady stream of different tactics.

I did repetition on a puzzle set of about 1000 Step 2 (from the dutch chess steps method) problems in the summer of last year and I actually did get thru 1000 problems in one day eventually. I saw my blitz rating go up a lot from about 1500 to 1800. However, I have not kept up the regimen and have not done 1000 in one day again, and my rating went down. I'm currently building out a larger puzzle set and hopefully will be able to return to a blitz peak...
@DailyInsanity Would you say this woodpecker approach has an advantage in terms of training economy over just solving easy puzzles (e.g. puzzle rush)? If we're scoring >80% anyway and only taking 5-30 seconds per puzzle, is there any point to the repetition part?
@spyputs said in #10:
> @DailyInsanity Would you say this woodpecker approach has an advantage in terms of training economy over just solving easy puzzles (e.g. puzzle rush)? If we're scoring >80% anyway and only taking 5-30 seconds per puzzle, is there any point to the repetition part?

I'd think that taking 5-30 seconds on a puzzle is still 'too slow' to have considered the corresponding pattern truly learned, in which case I think repetition would still be beneficial.