How Long Does It Take Ordinary People To "Get Good" At Go?

I recently read this post which discusses aggregate statistics over time for how fast people improve at Chess:

How Long Does It Take Ordinary People To “Get Good” At Chess?

TL;DR: According to 5.5 years of data from 2.3 million players and 450 million games, most beginners will improve their rating by 100 lichess rating points in 3-6 months. Most “experienced” chess players in the 1400-1800 rating range will take 3-4 years to improve their rating by 100 lichess rating points. There’s no strong evidence that playing more games makes you improve quicker.

Link to Full Article:

lichess_database_ETL/README.md at main · jcw024/lichess_database_ETL · GitHub

I would love to see a similar analysis (or at least some quick aggregate statistics) to the posted article for online-go . Surely the data exists!

Also, I wonder if the above fact found for lichess also holds for online-go?

2 Likes

OGS updates rank system too often.

3 Likes

Ordinary people don’t play Go!

15 Likes

I highly doubt that. Here are some issues off the top of my head.
Let’s suppose that we use DGS rankings that are not very fluid, then there is the problem of DGS being a correspondence server, so speed is different there, just from that.
Let’s say we pick only non-correspondence games, what time-frame is considered as a “game to improve”? Definitely not some uber fast rapid games (but some people will disagree on that, for sure) ? You shift those out as well.

Now what remains has a nice array of issues, even supposing that you get data from a server with a more “stable” ranking system.
a) Sandbaggers polluting the data.
b) non-sandbaggers coming from other server, declaring a normal provisional rank and rising fast through it, till they reach their real rank.
c) not everyone is aiming in actively improving. I haven’t had time to study and improve at Go since Aug 2018. My flat rank line would pollute the data

I think those - among other issues I obviously cannot fathom - were the reasons they had to go for such a huge dataset (5.5 years of data from 2.3 million players and 450 million games). We do not have that though.

2 Likes

I think these statistics would depend heavily on which players you include in the data. At any level, there are players who quit. And what about players with a long hiatus in their go “career”. And how to account or players who reached a plateau (which all of us seem to do at some point)?

IME there is a positive correlation between frequency of playing and improvement rate.
But at very high playing frequencies (which is nowadays possible online), the benefits seem to diminish. Or maybe playing that much just causes them to reach their plateau faster?

And what is an “ordinary” person? Most go players in the Netherlands have had some form of higher education and have an IT related job. So the population of go players has some different characterics than the general population. Go players are not really “average” people.

Those things being said, quickly improving players (not at their plateau yet) gain about 1 rank per month at DDK level and about 1 rank per year at dan level.
So I would say that “ordinary” people can get “good” (SDK range 9k-1k) in a year or two.

But there is a wide variety. Some quite intelligent active go players for decades, never get stronger than 15k (plateau). So there seems to be no limit to the slowness of improvement.
But there does seem to be a limit to the quickness of improvement. I have heard of several people who reached 1d in a year, but I have never heard of someone reaching 1d in a few months.

3 Likes

One problem is that people on OGS may have multiple accounts. Maybe it would be better to work on data from the EGF or the AGA.

1 Like

There has been a study on some EGF data. IIRC it was published in the Finnish go magazine years ago, but I cannot find it now.

They found that this model fits improvement of individual players quite well:
image

level = a - exp(b * -t)

(Edit: I’m missing some scaling correction in that formula, but I hope you get the idea)
Where a is the individual player’s max level (plateau), b is some constant to characterize the individual player’s improvement speed and t is time passed or games played (a unit of effort).

So if one would analyze historical data from OGS/EGF/AGA, one could find best fits for a and b for each individual player and then extract the typical values of those parameters within the selected population.

(Note: AI improvement graphs are quite similar, though those don’t go fully flat as human graphs tend to do, but maybe that is because AI don’t grow old or tired and their owners can work around a plateau by multiplying the network size, which humans brains obviously can’t do).

3 Likes

It depends on what you mean with “good at Go”.

Is it OGS or other website ranks?
Is it playing a decent game with a good but not pro opponent?
Is it playing a decent game with friends or at a local medium-skill club?
Is it becoming pro?

If you just want to be “good enough” for a casual game, I suppose it depends on how much time you dedicate to learrning and a little bit on your learning style (visual or studious).

It probably takes different amounts of time to different people.

1 Like

What exactly means to be “good at Go”?

I think it has a lot to do with the age when people start to play.

My experiences from my Go School (for kids) years would say about 2 to 3 years (up to 5 years, but it’s an ariticial boundary) to dan level (good enough?). Go schools generally divide kyu rank elementary school students into several levels (甲 乙 丙 丁 戊, similar to A B C D E alphabets, and toddlers class). And if kids are talented enough and practice hard enough, they can advance one level per semester (half a year), and participate the local dan-advancement tournaments when they reach highest kyu level class (so about 2 to 3 years). But kids normally don’t stick around after they graduted into junior high school (after 6th grade, the artifical year limit). I don’t know the dropped out rate, but my feeling is probably at least half wouldn’t pass and advance to the dan level (Taiwanese amateur low dan level I think is probably mid to high SDK here on OGS, different local dan tournament strength varied a lot). Some talneted kids would jump level, and quickly advanced into high dan level tournaments when they are still in elementary schools (Taiwanese amateur high dan level strength varied even more, some are close to pro, some are far off)

Another exprience is from my high school Go club (and some from my university club). People are more “general”, and are geniuenly interested in Go. A lot have exprience as kids attending Go schools, but some don’t. Those who are interested but lack the finance or family support and join Go club as their starting points, usually study the hardest. But they tend not to advance fast (and Go clubs are also less intense in training). But they normally can reach the high sdk level. However, a few people in my university Go club are true beginners when they join (pulled in by friends, bf/gf, etc.). They generally just stick around as audiences, and they just don’t “get it” about big and small, life and death, and repeat the same blunders over and over.

2 Likes

Ordinary people just doesn’t exist. I have no idea how to determine who will be it. So I can give from my experience the time it takes for let say a majority of players from the beginning to. …

To what? What is being good at go? A 20k can beat a full beginner easely. If these players don’t meet stronger players, the 20k will be honoured to be strong at go.
Anyone being 6 stones stronger as his opponent is very good at go in the eyes of him.
If you want a more general assessment of being good at go, i have no idea. Is it the top pros winning major tournaments? Is it to be in the top tier of amateur? Is it to become dan or sdk?

So for the children i share the views of @claire_yang.

for adults beginners it varies a lot, especially with what an adult (can or want to ) invest in time and effort. and varies a lot to what you think is good at go.

in my own view being good at go start around 13k and i feel to be a bit elitist already saying this. It will take between a few years to a few weeks, hard to say.
after that you are like a student trying to get a kind of mastery of the game which is said to happen when you breakthrough to shodan. all those students are ofc good at go. few climb the dan levels statistically then, but i think to master is better as being good.

2 Likes

Ordinary people will get good at Go at exactly 10.000 hours of practice, because they’ll never understand what Go is about.

1 Like

:woman_teacher: Have you been practicing today?

圖片

2 Likes

This answers the OP’s question directly, in the best way! :joy:

(As for me, I haven’t been practicing and it shows. I’m not putting the stones inside the boxes anymore :stuck_out_tongue: and that’s my best improvement. Shame, I know)

3 Likes

Let’s say that a person has stable rank r if that person

  • has rank r
  • has hit rank r at least three years ago
  • has played at least one tournament a year.

We could ask questions like:
A. For people with stable rank 1d, how long did it typically take to go from 10k to 5k, and from 5k to 1d?
B. For people with stable rank 5k, how long did it typically take to go from 10k to 5k?

Looking manually at a small sample on the EGF database, it seems that a player with stable rank 1d takes about 1 year or a bit more to go from 10k to 5k, and then several years (say 1…5 years) to reach 1d. A player with stable rank 5k needs between 2 and 5 years to go from 10k to 5k.

Disclaimer: I only looked at less than 10 players so the numbers above may be wrong. In addition there is a lot of variability, so these figures shouldn’t be taken too seriously. One reason is that there is no linear relation between time and effort.

1 Like

Personally, I can relate to the question the OP poses. I remember when I first started wrapping my mind around Go, looking up at a learning curve that seemed like climbing a vertical cliff, and wondering “what kind of commitment am I taking on here?”

At times like these, I find it useful to think of these things using a martial arts metaphor. So let’s say you know nothing - and you want to learn karate or tae kwon do, etc. If you were to join a dojo as a beginner you would be assigned a white belt to let people know that you’re just starting. At this point - you are no different from the regular people out on the street in terms of your knowledge / experience - the only difference is your willingness to learn something new.

In martial arts, a beginner / white belt is rated at a 10 kyu, and an expert / black belt is rated as a 1 dan (shodan). In between are various levels of basic skills (orange / yellow), intermediate (gold/green), and advanced (red / purple / brown)

The skill levels in Go are more granular, so they are spread out between 30 kyu to 1 kyu (rather than 10-to-1), but the same metaphor applies. From 30-20kyu you are learning basic skills. Once you rank up to 19-10kyu you are learning intermediate skills. From 9 kyu to 1 kyu you are becoming more advanced, and then you rank up to 1dan and become a “blackbelt” master. But it’s good to remember that even as an intermediate beginner (let’s say 18 kyu) - you already possess a much deeper understanding of the game than a “regular person” who has never played Go before - just like someone with a green belt in karate is going to have more skill in hand-to-hand fighting than someone who is starting from scratch.

So what would it take for you to feel that you are “getting good” at Go? Do you need to be a black belt master before you can say that of yourself? Or do you need to play well enough to be entertained by the game, and feel that you have an intermediate understanding of what’s happening, even if you don’t have all the skills yet? Because you can get that feeling at many steps along the way - at 20 kyu, 15 kyu, 10kyu, etc…

The question of “how long” is another big variable, because so much depends on what type of effort and discipline you put towards your learning. If you just keep playing casual games with people, but you

  • never play teaching games
  • never have your teacher / mentor review your games with you
  • never watch YouTube tutorials for beginners
  • never read Go guides aimed at your current level, etc
    Then you could play for 12 months and barely see any improvement because you just keep repeating the same tactics (and mistakes) over and over.

However, if you commit yourself to a system of learning that works for you, it’s perfectly reasonable for someone to rank up from 25kyu to somewhere between 15-12 kyu within a year. At that rank you might still be far from a master / black belt - but you already have the skills and experience to enjoy complex activities like watching pro games and really being able to appreciate how elegant and complex their play is compared to your own. You can watch live games on OGS of more advanced players, and be able to follow the flow and story of the game, and appreciate various tactics and strategies - even though they might be more advanced than you’re ready to try yourself.

If you change your perspective, and see the process of learning Go as a journey - rather than a destination - you can relax, and enjoy each step of the process. Good luck!

2 Likes

The chess article I linked isn’t immune to any of the effects you mentioned here for Go. Lichess may have more players, but I don’t see there being a difference in the array of issues you mentioned in online-go vs lichess.

Again, these same issues exist for chess as does for go. The linked study attempts to control for some variables and admits there are confounders, yet is able to produce some meaningful results regardless.

People surely have multiple accounts on lichess as on OGS. I don’t see that having a significant overall effect when properly aggregrated.

You’re right that taken literally “good at Go” is ill-defined. The linked article concedes this and rather attempts to characterize:

the purpose of this project: to figure out the typical improvement rates for typical online chess players on lichess.org

It proceeds to characterize the improvement rates for players that begin at different levels over time and as a function of how many games they play.

1 Like