"Best Way To Study" Experiment - [SIGNUPS CLOSED]


All good suggestions. I could probably manage an hour of study plus one ranked game a day, but I can understand that some people can’t spare this amount of time. Splitting option 1 into one ranked game and either tsumego or pro game memorization may be a good idea. Can anyone recommend a good sgf editor for Linux? I am using Quarry at the moment, but when I tried to memorize a pro game it created a new branch in the game tree even when I replayed a move correctly.

Does the “play a lot” option include reviewing? I review the majority of my games using Leela.

Teaching games and reviews from a higher ranked player is a good idea if we can find enough teachers and reviewers.

I haven’t tried playing Leela at 19x19 yet. Would you suggest even games or a large handicap?

There are plenty of youtube videos available. I’m not sure how useful watching a Nick Sibicky video, for example, a day would be though for someone at my level.

Anyway, I’m happy to be assigned to any study group. Are you going to make the selection at random?


1/ Is probably the most normal way of working on go. That’s what asian kid do right ?
For myself, i don’t really like tsumego puzzle but i would love to work on memorizing pro game;

2/ i already playing a lot, 10 correspondence game at the time, few live and blitz a day. So nothing new for me here

3/ Teaching game and review from a Dan or a low kyu is hard to have everyday. But i may find some people to do it for me. Also reviewing my own game, to see when i lost or what did i do to win is interesting. Looking for his own mistake or the mistake of the opponent.

4/ Bot-use, yeah why not but bot have a way to people that may be hard to deal with for a high kyu player. i’m mostly curious about that. A study about ddk playing with bot to train can be significant to see the real benefice of bot for studying.

5/ i Really enjoy the game of Dyrwin from his playlist ‘Back to Basic’, that really teach me something, i highly recommend it. It may be a way. (but i already did it, i need something else now)


Is it necessary to define fixed conditions? It makes me wonder whether there will be enough variation in the data. Perhaps defining a list of study activities would be enough. If each participant keeps a diary or log of how much time they spend each day on each study activity, the data could be analysed to see how time invested in each activity correlates with improvement.


Thanks for chiming in, everyone.

I’m considering to do something like what alemitrani said, but for other reasons. I would have to rename this project if I did exactly that, because it would not be an experiment anymore but an observational study. :slight_smile:

The problem with ‘no conditions’ is that people won’t feel the need to adhere to the menu, effects if any will be diluted and confounded to the core (selection bias, baseline bias, individual characteristics bias, time investment confound,…) and it’s hard enough to reliably track what people do (journals will already be necessary).

The most prominent limiting factor for this research seems to be time commitment per participant. But, as they say, limitations encourage creativity and I’m working out a way to turn this into a perk.

Stay tuned (and keep the comments comin’). :slight_smile:


I’m definitely up for a variety of the suggested activities, as I’m already doing most of it in one way or another.

  1. Right now I’m just playing a lot to get my rating to stabilize, but normally I’m also doing Tsumego and I already started memorizing games, even if it’s just the first few dozen moves.

  2. Since my online Go anxiety seems to have magically vanished after not playing live games for a month, I would also be down for just playing a lot.

  3. Not sure about the teaching game/review, seems pretty difficult to pull off on a daily basis for a month.

  4. Probably one of the least popular options, but I could also imagine training with Leela and reviewing with Go Review Partner. But I’m not sure if at my level I’d have to use a really high handicap to stand a chance, which might have an effect on how I play in my daily ranked game.

So 1,2 and 4 are fine for me.

Incorporating books or youtube lectures might also be a good idea, but a lot of these videos are already close to 1 hour long, which might be too much for a lot of people who only have 1 hour available every day.

Personally I really like books, but except for tsumego books, I have a tough time actually learning from them.
I’m also not sure how to use those in this experiment, maybe work through specific chapters of a free book each day?

Anyways, it’s starting to get interesting, so I’m looking forward to any new information. And I’m back to my 20 rating-stabilizing games. I hope I can make it.


It would be interesting to take into account roughly how much participants sleep. I find myself trading off time spent sleeping and playing Go, and I’m not sure that it is productive. I wonder whether I might improve faster if I played less Go and started sleeping more. There is a load of existing research on the benefits of sleep so a basic hypothesis could be that studying Go is productive, when done in moderation, i.e. when it does not incurr loss of sleep.


Of course you’re only going to be receptive of input if you can focus, but what’s more important is regular, deliberate practice (most stable result supported by research on learning).

In order to generate intuition for good moves, you need to be regularly exposed to good moves (pro game memorization/replay). In order to reliably visualize move sequences, you need to make a habit of visualizing move sequences. Tsumego help because they are forced sequences (they do branch, but predictably so). In order to be able to handle live games, in addition to the above, you need to get accustomed to time controls. The more you practice these, the easier it is for you to find a rhythm (spend x seconds on assessment, y seconds on reading, z seconds on new assessment, …).

Psychological factors boil down to focus/confusion, confidence/anxiety, calmness/nervousness/anger.

This is why it would be desirable to have a controlled experiment with random assignment, to balance out groups in terms of individual characteristics. Alternatively, to include (necessarily self-report) measures of these individual characteristics. This seems to be nigh impossible considering the fact that some methods simply require a minimum of time that lots of people don’t seem to have.

I’m working on finding a reasonable solution to this, but as indicated, I might have to drop the “experiment” badge in turn.

The trouble with people being unable to play 1 game a day makes it difficult to measure progress (fewer games = fewer possible results, less reliable data). The trouble with taking out the requirement to play 1 game a day to accomodate is that I would have to have you guys play a set of games after the study period. Naturally, people who play more games will improve in the tangentially related fields (time control management, ranked game anxiety) which might confound the results accordingly. One game a day was meant to reduce that hypothesized effect.

Either way it’s a compromise. I just want to make sure that we learn something interesting in this 1 month. :slight_smile:


Well, as an archetypical subject (25K) I can say that, whatever we learn from the study per se, some/all of us will be learning go in a concentrated fashion over the period. That’s a win all 'round.


The first option is ideal for me. I have the time available for it and it seems to be the well rounded option.


Alright everyone. It’s decision time. Which of these options do you prefer? Please only vote if you’re a participant. :slight_smile:

  • A) A single game a day, everyday
  • B) No games, only practice as prescribed for a month, then 10 games in 1 week

0 voters

ONLY IF you voted for option A, please choose:

  • “The Menu” - [1 game a day+] Memorize and replay pro games, do tsumego
  • “The Grind” - [1 game a day+] Play as many games as possible, at least 2 a day.
  • “The Network” - [1 game a day+] Play and review at least 1 game a day with a strong bot

0 voters

ONLY IF you voted for option B, please choose:

  • “The Minimalist” - Memorize pro games, do tsumego
  • “The Autodidact” - Read books & articles, play (teaching?) games, take a break, but do NOT memorize pro games an do NOT do any tsumego for 1 month

0 voters


Before selecting anything I must ask , why not start with the easier approach?, I mean, is hard enough to play a daily 19x19 and studying tsumegos or memorize pro games everyday, since it requires at least 2 hours per day , then even If you are engaged with the experiment I seriously doubt that that anyone will gonna stick to follow a plan that requires more than 1 and a half hours per day every day, for a month.
why not starting with some not that demanding approach like, doing 30 minutes of tsumegos (or if you dont like tsumegos then maybe just do one game), every day for 1 week.
Then the next week you can add to the initial 30 minutes , another 30 minutes to do some other activity (e.g. if you where doing tsumegos , then when you finish it review pro-games for 30 min. ).

I think most of the people (and this can be applied from go to learning languages, or any discipline ), struggle to do some activity for long periods since, from the very beginning they start overworking, then after a couple of day everyone get tired and quite. I think, what is important first of all is building a habit, then when you have a habit is actually quite easy to repeat it every day.


I have to disagree. I have found that people like challenges. This is not new we were ask if we can do this up front. Plus this is somewhat of a scientific experiment for find the best way to learn go.


Okay, I got confused and selected something at the bottom.

To be honest, I don’t see fitting anywhere really. :confused:


Well, as I’ve said previously, there is really only one way to measure progress for a large number of people halfway reliably: tracking rating over time. This can either be done continuously (1 game a day) or in a before-after fashion (10 games in a week after the practice period is over).

Since I cannot even claim to reliably measure progress if I a) don’t have a reliable baseline and b) have very little data per player (your rating can only improve so much after 10 games), this is already a compromise. I would like everyone to spend roughly the same amount of time on each method.

Since it is impossible (and, incidentally, undesirable) to force everyone to do exactly 2 hours every day, there will necessarily be some variability in time spent. That’s why I need everyone to meticulously time their practice schedules per day.

However, there has to be a common factor. And that will be the game a day. The only alternative is to do 10 games after the study period. I would prefer more, but it will be those with very little time on their hands who choose this option, and they won’t be able to do more than 10 games. Extending the period to more than a week is problematic, because by then they will have exceeded 1.25x the duration of the study and it becomes hard to attribute rating changes to their study during the 4 weeks.

And yes, if you can only fit in memorizing 25/50/75 moves of a game a day, so be it. As long as you record your time investment. You are responsible for doing the best you can, not for doing everything on the list to the max. Even though it would be awesome. :blush:

Also, if you can play 2 games a day or have time for a bot game and a review, I need you in camp A for the comparison’s sake. :slight_smile:

It’s impossible to please everyone, I understand. Still, if you have more concerns (or if you think I haven’t answered your question), keep 'em coming.

@korni If you get confused, don’t vote, ask first.


@smurph, well, I wasn’t confused when I voted. :stuck_out_tongue:

Anyway, if I manage to join, I will probably go with the option of studying then play a seies of games.
What makes it hard to imagine, is that when I study a section/chapter, in a day or two, I then like to play one game to try and apply what I learned. Then move on… At the end, I would play many games to try and apply all what I’ve learned over a period of say one week.

I still haven’t figured out what the length of the study will end up being.
But overall, I find it personally quite restrictive at the moment.


Well, without restrictions, there will be 36 different approaches and none will give us a relevant result. If you want to participate but the least restrictive method, there is B - “The Autodidact”.

As for the study period, it will be limited to 4 weeks.


Yes, I have gone for that option.

Small technicality, but option B leads to 5-ish weeks (study a month, play a week). Just saying, that wouldn’t be an issue for me.


I went for A-Network, which isn’t necessarily my top choice, but it’ll still be enjoyable for me. I think if everybody will take the same route, there won’t be any data to work with except “more time = more progress”, so I went for a less popular option.

Just for clarification: All other Go activity would be more of less prohibited during the next month? So 1 game a day is set and I only have influence on how much of the chosen activity I do?


Long story short: I left the group.

My head wasn’t in the right place to play and qualify for the required ratings/game numbers. In addition, the recommended time setting the matches had to be played with is for me (right now) a bit too stressful and I’m not yet confident in my playing abilities to handle it well, not without probably losing all of the required games (badly). It’s not working for me. I am truly sorry for the problems I might have caused with this.

I will, however, continue to study, learn and play more Go and become a better player… no, a better version of myself.

Thank you, smurph, for this opportunity. <3


I picked option Network. It seems fun. It is something i havent tried before. Not sure if I got this correct- we play 1 ranked game a day against a human and then review it with bot or do we play games against Leela and review those (sort of like training with Leela) and then play a ranked game against a human or maybe both?