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

Ah, everyone has heard it by now. The best way to improve is obviously to lose n games as quickly as possible. No, actually it’s to play as many games as possible. Nono, actually it’s to do tsumego all day. Also not true, it’s memorizing and replaying pro games. Well, almost true, it’s… the list goes on.

I would like to settle this question once and for all: What is the most efficient Go study method? (Well, replications of the experiment notwithstanding.)

I have a bit of a background in research methods so I’d like to set up a proper ‘first’ experiment to give us an idea about the effect sizes we can expect and gather some additional data, insights and ideas, which would make future research in this direction easier.

In order to optimize the whole process, I would like to have a large set of participants, ideally beginners but any double-digit kyu is welcome. If there are too few people who’d be interested in participating, I would grudgingly accept SDK as well. :slight_smile:

The setup really depends on the number of prospective participants, because I will have to take dropout and incompleteness/loss of data into consideration. The more participants, the less of an issue that will be.

SO. If you’re interested in studying Go regularly, (ideally) strictly according to schedule for an extended period of time (again ideally about 3-6 months but I do realize that that’s just fantasy… worst comes to worst, 1 month of complete data should be enough to recognize trends).

What’s in it for you? Well, I suppose you’ll improve. :slight_smile: I’ll certainly offer at least 1 teaching game per participant. Depends on the turnout.

Interested? Please add this to your post:

  • your OGS nickname
  • your rating and overall rank
  • how many hours a day are you willing and able to spend studying Go (for at least 1 month)? <1 hr or 1-2 hrs
  • willingness to (privately) provide me with your age [yes/no]

If you are totally new, please read this post.

Thank you for your time! I hope we can finally settle this. ;D

Oh and - if you know any ddk, please spread the word and link them to this thread! The more prospective participants, the better!

P.S.: If you just want to comment, very well, but please keep your posts on topic

Currently: With our latest addition, fankylicious, there are… quite a few people, actually. :slight_smile:


I’m in if you’ll have me Smurph.

OGS Nick is Stresic, it says my rating is 1302, rank is 17k, but it will be varying around I’m sure… I’m willing to spend 1-2 hours a day or more, on most days studying go. Yes, I am willing to privately provide age.

Well, you know where to find me :wink:

1 Like

Rating: 1431 +/- 130 Rank: 13.7k +/- 2.7
willing to spend 1 hour a day for most days of the week
Age / gender available upon request

Question: how many games should we have in a week as a baseline (since I currently don’t really actively play)? Secondly, should the time spent on games be counted toward the studying?

Thank you for your questions!

  1. There is no required average number of games played, I might have to have some people play a few games to reach a certain total number though, say, if rating uncertainty happens to be unacceptably high. If you’ve played at least 10-20 ranked 19x19 games on this server though, you’re definitely in the clear.
  2. For all intents and purposes, the allotted time per day will include everything. More about this at a later date.

i’m down for this, hopefully it’ll motivate me to play more!

OGS Nick: gwqi
rating: 1587±113 [11k] (i probably haven’t played enough for this to be super accurate)
hours: at least 1
age: sure man

1 Like

I’m in :slight_smile:

  • OGS nickname: sulcino
  • 1386 ± 100
  • 14.7k ± 2.2
  • <1 hour a day
  • I’m willing to tell you my age, yes
1 Like

I would like to be in. My only issue is available time. I almost only play correspondence game because of that.
So if it’s needed to play live, I could not be able to do it.

  • nickname: lysnew
  • today I’m 1605 ± 86; 10.1 ± 1.6 but you know… rank instability… :wink:
  • hours: <1 hr
  • age: I will

Really? :slight_smile:

I’m totally down.

  • Nickname: Zurtar
  • As of today I’m 1194 ± 69, 19.4 ± 1.8
    -Hours: 2-3 (I already spend around this but I’m willing to do more)
    -Age: For sure
1 Like

In my (unrequested) opinion, the best way to improve is actually to review your games, especially ones that were lost and / or spent a longer than average time on. If you can get a stronger player to help you, more the better. I remember someone once quoting Lee Sedol as saying that the fastest way to improve was to play many fairly fast games (“fast” is ofc a very vague term) and then to review them.

This is kind of a silly question (if not stupid), but I’m assuming that this study is restricted to 19x19? Just sort of crossed my mind that some players only play certain board sizes.

19x19 is the targeted board size, yes. Even though 9x9 makes for excellent reading exercises, it is severely limited in terms of flow. 19x is a dance, to rhythm and melody. 9x is more like a wrestling match and 13x is… uh… chessboxing? It attempts to combine two aspects that are decent when considered individually, but really it’s just sad to watch.


I, too, almost always play corr games due to limited single-session time…yet I do Go actitivties (Youtube vids, playing, some books) culumatively for 60-90 minutes daily. Love to help you, smurph…please tell me/lysnew: Will corr games work with your study, or is it for single-session live games?

1 Like

I think 9x9 and 13x13 are useful in the very beginning to learn the mechanics of the game. You can play more games in a shorter period of time, so it speeds up your learning of a lot of basics concepts. I usually recommend starting with 9x9 against a computer (start with 5 in handi or whatever), after a week or two (depending on number of games per day), you learn to connect your groups, and playing solidly and how to kill unreasonable invasions. Once you win at least some games with 2 in handi (might be more difficult these days with much stronger computers, not sure), you can start playing a bunch of 13x13 to learn more of the basics on a bit larger board. Once the 13x13 starts looking like a normal game, move to 19x19 (I played maybe 10-15 games on 13x13).

But isn’t this something that should be part of your study? Will DDKs improve more if they play 9x9/13x13 than 19x19? :slight_smile:

Thanks for your question, Tongue. Due to the nature of correspondence games, they never represent a snapshot of someone’s abilities and for that reason they are not very useful in this case. Since even a month (short for a longitudinal study) is a long time for people to be recording all their study activities, we have to get all the data we can. Having only a handful of games to affect someone’s rating would be fairly pointless as it wouldn’t allow for much variance (n games = n results). In addition, your other games would have to be removed from the calculation… it’s just not feasible.

As for Bitt’s question - the advantage of smaller boards is that mistakes have a larger impact on the outcome and that games finish quicker. However, in order to add board size as a factor (3 levels) i would also need three times as many participants.

Currently the bare minimum I’m looking at is a simple 3-group repeated measures design with at least 5 people in each group. If you manage to recruit 30 more ddk, we can talk about one more factor. :wink:


Good point :slight_smile:

I would just like to say that I really like the idea, and am very curious about the results. I wish I had time to participate, but I only play correspondances, and could possibly spend up to 45min a day later in the summer =/.

But, as a mathematician (really not specifically within mathematical statistics however) I’m also a bit interested in the methods youre thinking about using in terms of regression, and so forth. Whats your plan, or will you decide most of that based on the amount of data you get? Maybe its a bit much to go through here but what are at least some of the regressors you are planning on using, and e.g. what response? Will you have an open github project for this or similar?

My thoughts are that I think it would have to vary vastly from individual to individual, and that the best way to improve is going to be the one (among a set of reasonable choices) that you actually end up spending most time with (i.e. probably the one you think is most fun). One hard part (I think personally from experience anyway) is to getting used to losing alot and seeing it as practice, and learning from it rather than just thinking that you suck =). So I would bet my money on the ‘playing lots of losing games and reviewing them, while winning enough to keep you playing’-card =), If I had to choose.

I really hope it goes well and that you come to some significant concluding results. However, just 3 groups of 5 sounds like a very small populatuion for such a, possibly high dimensional (even though shrinkage, pcr or likewise can ofcourse be sucessful even when p>n), complicated regression, I guess it depends a little on your approach though. But also, participants dont grow on trees, and more might be too expensive.

Thank you for doing this anyhow, and good luck!

Thanks for the questions, Chris.

As mentioned earlier, I’ll have to decide on the method based on how many participants I can gather. Right now that’s merely 5, so it wouldn’t even be enough for a regular 2-3 condition experiment. I would have to resort to a crossover design and analyze the data via 1-way RM ANOVA, though I wouldn’t expect much of a finding at all.

If I can get 10 more people, it would be a simple factorial RCT, analysed by means of 1-way independent ANOVA.

If 45 or more people were to join, I could actually do some interesting preliminary and post-hoc analyses, plus I’d expect an actual result worth reporting. My special interest lies in effect sizes, and I know it takes either a superb treatment difference, large group of participants or selection bias to find a large effect.

Your intuition about [time] as a confound is reasonable, that’s why I will do my best to factor out time spent. After all, I’m looking for the most efficient way to study, not the mere confirmation that spending time studying will improve your Go.

1 Like

OGS name Quinlan, 30 years old, currently ranked 11k, 1597 rating. I have the time, currently 2-4 hours a day. Picked up stones for the first time 9 months ago and have been studying hard. Biggest impediment for me atm, is the breadth of things one needs to know to play consistently well. I think I’m at the point where there are no more quick wins in terms of learning, and just hard work left all the way to 1 dan. Would be nice if I stopped making silly mistakes in games, ah and if someone could author the book “the hurried man’s guide to reading 30 moves per minute”.

I’m not stopping my crazy study schedule until I reach my plateau, so a longer period is fine with me. I’m happy to try new things for weeks to test hypothesis.


Master Vos?

Im in

OGS Nick: JoseCastanys
rating: 1535±83 12k
time < 1h
age: 38