Systematically Using All Puzzle on Lichess

Has anyone tried systematically using all the puzzle types and puzzle features on Lichess? I mean by trying to do all the puzzle options in a systematic and rotating way with spaced repetition for reinforcement of learning? Yes, it could be a huge and maybe impossible undertaking but hear out my thoughts on this.

There are a great many puzzle types. I haven't checked how many types there are but I know there are more than in my intial list of 72 types from Lichess. Let us say there are 90 distinct types. That would be close. Remember there are tactics, motifs, checkmates and special game phase puzzles like endgame puzzles for example. If there are 90 properly distinct types, I could theoretically do 3 puzzle types a day in 3 x 1 hr sessions and get through the 90 types in a month. Thus, I would rotate through once a month.

Is once a month enough for spaced repetition reinforcement? I think maybe the answer would be "yes" when we add in the fact that many puzzles feature two or three of any of tactics, motifs or "phase types" in any one puzzle. So reinforcement will also come from the sheer volume of puzzles done, thus creating cross reinforcement of learning. As stated, I envisage doing 3 hrs of puzzles a day to get the volume. I admit, I don't know if I can handle this volume day after day for months or years on end. And I would still have to find the time to study the other aspects of chess.

Doing a large and valid statistical sample of puzzles in this way should mean familiarisation with close to all the tactical possibilities of chess. This is leaving aside visualization, calculation depth and many other factors. It seems to me that if doing lots of puzzles (a large and broad statistical sample ultimately in the tens of thousands) gave me a statistically valid puzzle rating and improved my chess ability then my puzzle rating and playing rating should eventually converge. If other types of chess study and playing live games also assisted in, then my playing rating might exceed my puzzle rating. I don't mean exceed my current puzzle rating. I have noticed my puzzle rating is now falling as I attempt advanced tactics and motifs compared to the basic ones. If I keep this up I expect my puzzle rating to fall even more; maybe for months before a rebound.

You will note that my hypothesis above (that puzzle rating and playing rating should converge) is probably not what most people experience. We tend to take more time on puzzle positions than on game positions (except for puzzle storm and puzzle racer I guess). Most puzzle types (except the healthy mix) also give a big tactical hint. Games don't have these hints nor do they signpost "hey you have a tactic here for the win or to save the game". So puzzle rating might always exceed game rating by a certain percentage.

Of course, I might find the proposed daily puzzle volume too hard to keep up. I haven't even worked up to the full rate yet. But it could be interesting (to me at least) to see if the hypothesis above will hold and to see if doing puzzles (mainly not exclusively) actually works much at all. In my case, my age (70) will be a confounding factor. I may prove unable to retain new stuff to any significant degree. But at least one study out there says learning / restarting chess and in old age helps to prevent or minimise cognitive decline. Can I build up the puzzle rate and keep it there? Or will it drive me mad? ;)

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6617066/

Footnote 1 : The title was meant to refer to all puzzle types, not all puzzles.
Footnote 2 : The recent rapid decline of my puzzle score seems to relate most to me hitting the advanced puzzles group and finding some of the most serious relative weaknesses in my tactical perceptions.

Sunk without a trace.

How can you remember most of them?

@Wodjul said in #1:
That's a lot to take on so how do you manage the timing of the spaced repetition? Some kind of software?

@ELO5287 said in #4:

That's a lot to take on so how do you manage the timing of the spaced repetition? Some kind of software?

Unfortunately, nothing so clever. I simply do one type of puzzle (tactic. motif or game phase) per day from the extensive list in Lichess. I do that type of puzzle three times in the day: one hour each in the morning, about noon and early evening per day. Each puzzle is simply served up to me at random by Lichess puzzles. So I am not repeating the exact puzzle but repeating the tactic, motif or game phase. I will work through the whole list (about 90 themes, motifs etc.) in a month and then repeat the cycle each month. My spaced repetition occurs on the day and then on the day a month later. The repetition is of the theme of the puzzle, not of precise puzzles except by chance. At the same time, these puzzles mostly contain several themes in one, so cross-repetition happens (randomly) in this manner too and with considerable frequency.

Currently, I start the puzzles on easiest setting in each hourly session and work up. When I get five in a row correct I move up a level. I allow myself time to work them out but I still try to get something like 20 puzzles done per session. Puzzles that accomplished players could do in 3 seconds can quite literally take me 3 minutes or more per puzzle on average though I see easier ones more quickly (mostly).

I try to see calculation lines through to the solution but I fail in that fairly frequently. Nevertheless, I will recalculate at each step, try to get through it and still call it a solution if I stumble successfully to the conclusion. I have to crawl before I walk. Can't expect perfection.

There is a lot of statistical noise in this approach but the human brain is adapted, I think, to make sense of somewhat open-ended ambiguous input. Out of the noise, it will pick/develop its ways of seeing the patterns. This is all in theory of course. I could be quite wrong about this approach. But if my brain is capable of learning at all, this approach might yield better sight of tactics over time. It's all a bit of an experiment of course. I don't mind the time and I enjoy doing puzzles except for the ones I am completely beaten by.

Doing the puzzles this standard way with all types will mean my puzzle rating becomes a metric of seeing all tactical patterns but only in the puzzle environment. My playing rating will become a metric of seeing tactics in real games. Perhaps they should, in theory, converge closer over time while both hopefully trending up. This may all fail but as I say it's an experiment to make the meta of doing a lot of puzzles more interesting for myself.

So far, I can see only a brick wall when trying to visualize positions more than 3 of my moves ahead. This puzzle method may or may not solve that problem for method reasons or because my brain is simply not innately good enough for the task.

@Wodjul said in #5:

I will work through the whole list (about 90 themes, motifs etc.) in a month and then repeat the cycle each month. My spaced repetition occurs on the day and then on the day a month later.

Repeating the whole cycle monthly is not what spaced repetition is though. In order for it to be spaced repetition the intervals between the repetitions must increase.
For something like this an implementation of SR would probably involve doing the puzzle themes you find the hardest most often and the ones you find the easiest at a decreasing rate.

@ELO5287 said in #6:

Repeating the whole cycle monthly is not what spaced repetition is though. In order for it to be spaced repetition the intervals between the repetitions must increase.
For something like this an implementation of SR would probably involve doing the puzzle themes you find the hardest most often and the ones you find the easiest at a decreasing rate.

Thank you, that is an excellent suggestion. I am not achieving proper SR except sometimes by chance. First, I think I need to standardize my approach more with respect to time. I think I need an "egg-timer" on my puzzle attempts. Three minutes maximum per puzzle would standardize matters and ensure I attempt at least 20 puzzles per hour. Of course, I could do easier puzzles faster. Then, as you suggest I need to record, via speadsheet maybe, my rise or fall in puzzle rating per session per type of puzzle. Then I need to weight the frequency of doing each puzzle type to do it more frequently or less frequently as required.

It is obvious now, as I hit the harder puzzle types in my test run, that I am much worse at the advanced puzzles. My puzzle rating is plummeting. I am making more dumb blunders which could be just a bad day but could also be a true reflection of how bad I am at seeing certain tactics, motifs etc and the patterns they involve.

This will take more thought but your suggestion gives me an excellent lead into refining this training method. Thanks again.

I will continue thinking here on this topic. I hope @dailyinsanity will also be interested, whether or not he thinks these initial ideas are on-target or off-target.

Summary of Projected Method.

1. I will be using all the discrete Lichess puzzle types (tactics, motifs, themes or TMT for short) as my "flash cards". The puzzles are random as we know but the chief theme and hence pattern for pattern recognition is consistent for each puzzle type. So each TMT "flash card" is a basket of (new) random puzzles involving the chief stated theme, like forks or pins for examples. The "flash card" is for the fork pattern or the pin pattern and so on.

2. For 3 x 1 hour sessions per day and for six days a week, I will start at rated level "easiest" and work in each TMT spending a full hour on one TMT. Five "wins" means going up to the next puzzle level. I will score each session by how much it raises or lowers my puzzle rating score. An individual puzzle will be a fail if I cannot successfully complete it in 3 minutes (on an egg-timer system).

3. I will follow a Leitner System of seven "boxes" (maintained on a spreadsheet) with a puzzle type scoring below my current weekly average puzzle score staying in box 1 or being demoted one box from a higher box as the case may be. Equal or above the average in score, it goes up one box. The boxes will equate to initial Day 1 learning introduction or re-learning, Day 2 re-scheduling, Day 7 re-scheduling, Day 14 re-scheduling, 4 week re-scheduling and 8 week re-scheduling.

There are multiple ways for this attempted method to go wrong and not work for learning and scheduling. I don't think I can even predict the ways things could go wrong or be unworkable. I am going to trial (and error) the method and see how I go. If nothing else, I will do a lot of puzzles (unless I quit) and there might be some method in my madness.

Suggestions are welcome.

Here is an excellent document on spaced repetition learning and the Leitner system:
https://traverse.link/spaced-repetition/the-optimal-spaced-repetition-schedule

@Wodjul

Rather than doing this experiment with the huge Lichess puzzles collection, maybe it would be more informative to use the "Intermediate tactics" and "mating patterns" sections of the training page, here:

https://lichess.org/practice

• a smaller, more controlled set of exercises that should be challenging enough for beginner to intermediate players.

@Panagrellus said in #9:

@Wodjul

Rather than doing this experiment with the huge Lichess puzzles collection, maybe it would be more informative to use the "Intermediate tactics" and "mating patterns" sections of the training page, here:

lichess.org/practice

• a smaller, more controlled set of exercises that should be challenging enough for beginner to intermediate players.

Those exercises introduce the main tactical ideas, usually with about 10 puzzles per tactic. Once done once (if you get my meaning), doing those tactics over and over again will provide little extra training. An extra and practically endless supply of tactics puzzles in variations is required. The player needs to learn a feel and understanding for multiple variations of standard tactics in a myriad of situations. At least, that is my understanding. Also, doing the tactics the way I propose, in a standardised, timed way in large numbers, should make my puzzle score a good metric of tactical improvement which ought to work effectively in games, in theory.

Of course, I could be wrong and this training method may prove unworkable or ineffective. It also may be the case that I am untrainable in chess, by myself or by anyone else, beyond just being an old patzer who knows the moves and a few basics. Or maybe I will crack and be unable to do such a volume of puzzles.