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


One human game, 1 bot game. If you’re strapped for time, play only the human game and review with a strong bot (Leela11, Lizzie,…). I might merge this with The Grind, judging by the current group imbalance.


I’d also like to know if the activities are intended to be exclusive or not @smurph


Yes, there is no good way around the 1-game-a-day requirement.

And activities are exclusive to each method, as in: if method A1 has memorization and tsumego, neither A2 nor A3 group members should do any of these. A1 members in turn are not allowed to review their games with a bot, etc.


OK. Let’s see if I understand.

At present I am learning Go in the following ways:

  • I attend Go club IRL twice a week
  • I participate in the Open Study Room as a member and as an admin
  • I play 9 x 9 games online
  • I play 19 x 19 games online
  • I review my games and get them reviewed by other people where possible
  • I help other people review their games
  • occasionally I use Crazy Stone to review my games
  • I do tsumego
  • I participate in tournaments and leagues

To participate in the experiment, I’d have to stop all of that and limit myself to the prescribed activities, plus recording how long I spend on each of them. Is that right?

It seems very restrictive, in particular playing a lot without reviewing my games seems like it will be ineffective from the start. I feel I have learned more through the combination of playing and reviewing each game than anything else. I’m not keen on breaking a good habit.


I wholeheartedly agree. If you’re already putting a lot of time and effort into all these things, if it’s going well for you and you’re having fun doing it, go for it, man.

This study is aimed at people who would like to improve in the ddk ranks, who want to know how to best invest their time - in playing lots of games or doing deliberate practice. It seems you are making ample room for Go in your life, and that’s great.

But not everyone has the luxury of spending this much time at all, a go club remotely near them, a strong player ready to review games, a powerful computer, … this is about challenging yourself to follow a relatively well-defined practice schedule for a month and to see if any method shows more promise than others. Most participants aren’t in it for the service to the cause but to establish a habit for themselves, to try something they haven’t tried previously, for the experience.

And in order to be able to even remotely compare the usefulness of the individual components, it would be unwise to mix methods any further.


Make alemitrani a part of control group who don’t follow specific schedule.


That makes sense, thanks for your reply. I don’t think this is for me, and I’m only borderline DDK anyway. I hope it goes well and will be interested in hearing about the results in due course.


I think the “grind” people are supposed to be the control group.


He follows a very specific schedule that’s a mix of all different activities and more.

Imagine you wanted to run a drug trial and investigate the difference between the effects of drug A and drug B. It would not make sense to have someone take both drugs, certainly not in addition to his personal choice of medication in form of drug C.


@smurph Unfortunately, I will have to withdraw from the study for personal reasons (IRL). Good luck and I’ll keep an eye out for the progress.


@smurph, for those of us that will be reviewing pro games, do you have a preference or view on games played against AI opponents; specifically, should they be avoided?




If you like, you can do this with any pro+ level game, but AI games are more disjunct and the specific moves are heavily MCTS / reading dependent. My perspective is that human-human games are more heuristic/intuition inspired, which should be a little easier to follow or make sense of.

I will shortly be releasing a guide on how to approach each technique, though you all have reasonable freedom in what materials to choose and how much of it to tackle.

Most other info should be available here:



Thanks for the clarification. That’s exactly why I asked since, as awe-inspiring as our new AI overlords are, they do occasionally exhibit some quirky play. I’ll stick to reviewing human v. human games, at least during the experiment.



I tried to read everything and didn’t find it.
How will you measure the gap between people doing the study and a average ddk player whom play and progress by his own ?


The short answer is that it doesn’t really make sense to try that.


i know that what will happen here isn’t really a scientific experience but usually to know if something work, it’s better to know what happen when we do nothing.
If the participant of this exp progress the same way ddk who aren’t in the exp, during this time laps. Maybe that’s something we want to know.


If you want to know what happens when you don’t do anything, look at people’s ratings 1 month before the study and at the start date.

Then you find out most people didn’t play at all to establish a reliable rating. Then you look to others who did play and established a rating and how their rating changed in a month. Then you realize you don’t know what these people did for their rating to (not?) change. You realize you can’t compare that even if you wanted to, because you didn’t (and couldn’t) assess and compare other relevant individual characteristics. Who is the typical ddk anyway? How do you sample that? You figure that if people improved, it wasn’t because they went to the disco. So they either played games, did tsumego, got teaching games, played bots, watched videos, read books,… and you remember that the question to be tackled here is if there’s a difference between approaches and if so, if it’s measurable. You realize people have different schedules and different amounts of time available to them. So you figure that maybe the best way to go about this isn’t to ask “how do people doing stuff compare to people not doing stuff” but rather “assuming people do stuff, what stuff should they do?” which brings you back to the point where you compare different ways of doing stuff and comparing the time people take to do stuff so you can find out if the (ostensible) progress from doing stuff A is comparable to progress from doing stuff B but that stuff A requires less time to do.

If you still want to know, you can lobby some ddks to join the option B - Autodidact group. :slight_smile:


I understand what you mean.
But with a large amount of people, you could have enough data to go into the field of statistic.
I think if you look at all the ddk of OGS, you could see in a month how much they progress. And assume that they don’t have a discipline of learning the game that much.
Don’t you think ?
Even, maybe the OGS team can give you data. To see each month of the last year how much a DDK progress in general.

And did you try to contact all the DDK of OGS to ask them what do ?

I may try to :slight_smile:


That’s a great idea and I say go for it. Ask for the data, get it, format it, separate reliable from unreliable accounts, sample them, analyze the data, sample ddks, interview them, be wary of self-report bias, write it up.



I understand you.
I know that there what we should/can do and what we will do.
That was kind of my job…

I may ask the team, the access to some data to see what i can do this it.