📝 Make reviews great again! 💪

[shinuito]

I think as a proper teacher or instructor to regularly give structured lessons and reviews, I could imagine a sizeable difference in rank (a proxy for knowledge) being useful, because you might need more knowledge in multiple areas to review a full game and make a lesson from it or recommend some relevant puzzles etc.

I think someone at the exact same level as you however can also teach you a lot, anywhere from 25kyu to 1kyu, because there’s still so much to learn, and you can’t know it all (rank as a knowledge proxy).

I think regularly someone around my level would show me a tesuji I hadn’t seed before at 15kyu or a joseki (or joseki variation) I hadn’t seen before at 5kyu and so on.

It’s also what makes playing even games as valuable as teaching games.

[Groin]

I don’t agree, I think a 16k can teach a full beginner without problem and even propose a kind of handicap game, which he could win even with a high handicap.

Teaching is a very fruitful experience be for the student and be for the teacher himself.

Back on topic in the general way, I think I am very nostalgic about old times on kgs but everything changed a lot. Apps, social medias, streaming… We don’t use internet in the same way.

My assumption is that streaming could be the real future but it’s not yet really an improvement because of the lag (far too big for interactivity) and because nothing is really built yet for some horizontal exchanges between participants.

Now how social medias are evolving make me think that the internet go may still evolve a lot too.

[jonquil]

@Groin

I don’t agree, I think a 16k can teach a full beginner without problem and even propose a kind of handicap game, which he could win even with a high handicap.

There is some truth in what you’re saying; however, in practice, it’s much, much, much, much harder.

First of all, a 16 kyu needs to understand their own limits. Maybe they know the basic rules (which, again, isn’t even necessarily a given at that rank), and sure, they might be able to explain some fundamentals. But beyond that, there’s not much they can reliably teach. In my personal experience, even up to around 4 kyu, a lot of players don’t have a particularly solid understanding of the rules or how to apply them in practice.

Secondly, there’s the issue of actual playing skill. Most of the mechanical aspects of the game are simply not developed enough at 16 kyu. At best, their understanding is still very limited.

Sure, they might be able to teach someone something. But there is also a very real possibility that they will misguide them without even realizing it. And when you’re teaching someone below 15 kyu, it’s not just about explaining mechanics or pointing out individual mistakes. A huge part of teaching is actually getting the person interested in learning and knowing how to guide them through the learning process. There are a lot of teaching skills that have to be developed over time. There are even entire books in China and Korea dedicated to the art of teaching and to understanding how to guide a student at different stages of development.

Someone at 16 kyu obviously isn’t going to have a deep enough understanding of that to apply it properly. A dan player, on the other hand, has almost certainly gone through this process themselves. They have had teachers, received instruction, seen different approaches, and can gradually develop their own way of teaching based on that experience. A 16 kyu usually doesn’t have that foundation. At best, they can imitate how stronger players teach, which kind of defeats the point.

And this is where I think we’re talking about two different things when we say “teaching.” If you mean giving very specific pieces of advice - like explaining a particular joseki, tesuji, or pointing out one obvious mistake - then yes, a 16 kyu can absolutely do that. But that’s not what I’m talking about.

I’m talking about reviewing an entire game - say, a 250-move game - and actually providing useful guidance throughout the review. How do you expect a 16 kyu to do that effectively for a lower-ranked player when both players still have major gaps in their understanding and both of them should probably be seeking guidance from someone stronger?

Of course, 99% of the time, a 16 kyu is going to beat a 25 kyu. But that doesn’t automatically mean the 16 kyu is capable of teaching the 25 kyu how to improve. If the 16 kyu is making many of the same mistakes, just fewer of them, then what exactly can they teach beyond pointing out isolated errors?

That’s the distinction I was making. I’m talking specifically about the ability to review an entire game, understand why the mistakes happened, explain them correctly, and then give the player useful advice that they can actually apply in future games. That is a very different skill from simply being stronger than the person you’re teaching

No matter how harsh it sounds, I think that before reaching a certain level, people simply need to focus on learning rather than teaching others

[PJTraill]

I am not yet convinced. I think your most interesting claim is that a dan player has learnt how to teach,from their own learning experience; I think you mean because they themselves will have had the right sort of teaching.

I would have expected there to be good and bad teachers at a wide range of levels, and that a 16 kyu with sufficient humility and sensitivity could be more useful to a raw beginner than a poor teacher of dan strength. Obviously there is a risk of teaching wrong ideas, but those wrong ideas are still likely to be better than the ideas the beginner would start with, and I suspect that this process of moving from one set to a better set of wrong ideas characterises everyone’s learning, up to and including the AI revolution at professional level.

I think that an important part of teaching is understanding what the pupil was thinking (which they may often not be able to formulate well). While experience of many players of all levels plays a big role, empathy and having a similar mindset or level can also help a lot. It is also important not to teach more than the pupil is ready for, and there again a good teacher can choose the most helpful ideas.

[tonybe]

@PJTraill

empathy and having a similar mindset or level can also help a lot

Unfortunately, those personality traits don’t automatically show up as a given Go player achieves a higher rank

I absolutely agree with you that some people make better educators than others based on very complex things like empathy and being able to understand where the beginner is coming from, etc, but yes, those things don’t necessarily correlate with being a higher rank

[Jon_Ko]

According to insights from educational studies (i.e. Hattie’s Visible Learning, one of the most influential works in the field) you seem to be just plainly wrong.

Peer tutoring has an effect size of 0.53, while Teacher subject matter knowledge only has an effect size of 0.11.

Hattie effect size list - 256 Influences Related To Achievement 

[jonquil]

According to educational studies, sure, but I think you’re stretching what those numbers actually show. I’m not saying peer tutoring is ineffective. Obviously, having someone who understands a subject better than you explain it can be extremely useful. But “peer” doesn’t mean “someone with the same level of knowledge.” It just means another learner. In a school setting, having a smart friend who understands math better than you can be very helpful. But if neither of you understands the subject, being peers doesn’t magically make one of you capable of teaching the other.

And that’s especially relevant here. Peer tutoring is most useful when both people already have some basic understanding of the subject. If two people barely understand Go, there’s a much higher chance that they’ll just reinforce each other’s mistakes. Also, you’re comparing peer tutoring (0.53) to teacher subject matter knowledge (0.11), as if those were the only relevant teacher-related factors. They aren’t. The same Hattie data has much stronger effects for things like teacher credibility, teacher clarity, feedback, direct instruction, and teacher estimates of achievement. So you can’t really conclude from those two numbers alone that learning from a peer is better than learning from a teacher.

And in Go specifically, a dan player who has spent years being taught by stronger players has more than just stronger playing ability. They have accumulated experience of being taught: how stronger players explain positions, how they identify important mistakes, how they structure reviews, and what kind of advice actually helps someone improve. If they have teaching experience themselves, that’s another advantage.

Could there be exceptions? Obviously. If you have some genuinely excellent teacher with years of experience teaching other strategy games and a strong understanding of pedagogy, they could absolutely be better at teaching than some random dan player. I’m not arguing that rank automatically makes someone a good teacher. I’m talking about the general baseline: a dan player who has reached that level through years of playing and receiving instruction will generally have a much stronger foundation in the game, and potentially in teaching as well.

And my original point was never “a 16 kyu can’t tell a 25 kyu anything useful.” Of course they can. They can explain a joseki, point out a mistake, or give some specific advice. I’m talking about reviewing an entire game and giving reliable guidance on how the player should improve in future games. That’s a completely different skill, and I don’t think a 0.53 effect size for peer tutoring demonstrates otherwise.

[jonquil]

Yeah, I actually agree with a lot of this. A humble 16 kyu who understands the beginner’s mindset can absolutely be more useful than a bad dan-level teacher in some situations.
My point is more about the baseline, not exceptions. Empathy and understanding what a beginner is thinking are important, but you still need enough knowledge to correctly diagnose their problems and know which ideas are worth teaching. Otherwise you can be very good at understanding the student while still giving them bad advice. And that’s why I specifically mentioned full-game reviews. I’m not saying a 16 kyu can’t teach a beginner anything useful. I’m saying that there’s a pretty big difference between giving a beginner some useful ideas and reliably reviewing their whole game, understanding the underlying problems, and guiding their future improvement.

So I think we mostly disagree on how much you can reasonably expect from a 16 kyu, rather than on whether peer teaching can be useful at all.

[jlt]

Perhaps the kyu/dan range is misleading since gaining 1 stone becomes harder and harder. Perhaps the Elo difference is a better measurement of skill difference. On OGS, a difference of 250 Elo represents roughly 6 ranks around 15k, and 3 ranks around 1d. So perhaps a DDK teaching to someone 6 ranks weaker is like a dan player teaching someone 3 ranks weaker. That’s perhaps not ideal, everyone would like a perfect teacher with lots of knowledge and teaching experience, but it’s not possible in practice, not many such teachers are available. That’s why a pyramidal system is more viable: high dans teaching low dans, low dans teaching SDK, SDK teaching DDK, DDK teaching beginners.

By analogy, it’s like an elementary school teacher who only has 8th grade knowledge, so can make a few mistakes and sometimes communicates wrong ideas, but still does the job.

Similarly, when reviewing the game with your partner, you may exchange wrong ideas. Like “you played A because you thought of variation X, but you should play B because of variation Y”. A high dan teacher might think that both players are wrong, the correct move being C because of variation Z. But talking about variation Y to someone who only saw X is already an information, so may have some value.

In any case, it is important that the reviewer does not consider himself as an expert saying definitive truths, and that the student does not trust the reviewer blindly.

[Jon_Ko]

@jonquil

my original point

Maybe you should re-read your original point. Because to me it sounded like you were telling a 16k them teaching a 19k is doing more harm than good.

And to me you were saying I cannot give good advice to anyone stronger than 12k. Based on my personal experience that is simply wrong.

[jonquil]

@jonquil

And this is where I think we’re talking about two different things when we say “teaching.” If you mean giving very specific pieces of advice - like explaining a particular joseki, tesuji, or pointing out one obvious mistake - then yes, a 16 kyu can absolutely do that. But that’s not what I’m talking about.

I’m talking about reviewing an entire game - say, a 250-move game - and actually providing useful guidance throughout the review. How do you expect a 16 kyu to do that effectively for a lower-ranked player when both players still have major gaps in their understanding and both of them should probably be seeking guidance from someone stronger?

I doubled down on my opinion; I don’t see much difference, to be honest

@jonquil

And my original point was never “a 16 kyu can’t tell a 25 kyu anything useful.” Of course they can. They can explain a joseki, point out a mistake, or give some specific advice. I’m talking about reviewing an entire game and giving reliable guidance on how the player should improve in future games. That’s a completely different skill, and I don’t think a 0.53 effect size for peer tutoring demonstrates otherwise.

Anyway, I should have mentioned that anything that really helps a person grow their skills is fantastic, no matter who explained it or how. I was talking about probabilities and general perception

[Groin]

It matters if you discourages 16k to teach beginners. Most at this stage are afraid to do it. They are afraid to give handicap stones too.

that’s a shame, we don’t need always a full game review, we don’t need always what a dan player can see.

It the 16k is cautious enough to not pretend to be able to answer to any question he will make a great teacher. He will get a very nice experience as a teacher and discover that he knows more as what he thinks, but with quite some limits of course.

[jonquil]

You are right, sorry if my words been wrongful then

[Jon_Ko]

Sorry, I should have quoted the post I was actually referring to:

@jonquil

I don’t mean to be rude, but I am saying this from the bottom of my heart: in the kyu range, good advice can really only come from someone with a 10+ kyu difference. In many cases, you might accidentally mislead someone, so it’s better to leave teaching to the stronger players, but its exelent that you have such intentions!

Anyway, I think we can agree that a random dan player would give a better review to anyone than a random kyu player would. Also, that if OGS will add a real queue, we don’t want DDK’s to review an SDK’s game and make them get removed from the queue by doing so.

But I disagree with your 10+ rank difference rule (or guideline)… (and with telling kyu players they shouldn’t review games of slightly weaker players because it might do more harm than good).

I guess it would be important to figure out which rank difference is actually appropriate when creating such a review system. Personally I’m confident that I can review games of players that are 5 ranks weaker than me and come up with good advice for them (less confident when the rank gap becomes smaller than that).

[jonquil]

I think it should be somewhat vague: the stronger the player providing the review is, the smaller the required rank difference should be.

For dan-level players, a difference of 2 - 3 ranks should be enough; for SDK players, around 4 ranks; and for DDK players, around 5 ranks sounds reasonable. However, I would still like to set a minimum rank from which someone can provide such a review - perhaps 15 kyu, or whatever you think would be the most appropriate threshold.

Also, when I mentioned a 10-kyu difference, I was specifically talking about the difference between two DDK players. I wasn’t clear about that at all.

That said, honestly, I don’t really see how a review by a DDK player would be particularly useful without a significant rank difference between the reviewer and the player being reviewed. At that level, it might be more appropriate to treat it as advice or an opinion rather than a reliable review

[marinakai]

I suspect formal classroom pedagogy may not be the right field to reach for here — I’d be more interested in comparing to skills like e.g. music training — but I want to emphasize that a 16k getting a review from a 6k (or whatever) is still peer learning. The difference is generally “are they a fellow student” vs “are they a formal educator”, not “are these two students at exactly the same level”.

[Jon_Ko]

Again, I think we agree that the reviewer should have a higher level than the player. The question is how big the difference has to be.

Maybe looking at music training is interesting too, but lots of countries are putting lots of effort into education sciences to examine how they can improve their educational systems (which are mostly based on formal classroom pedagogy). I think building on that would be valuable. That’s a different conversation though.

[ronin3]

I think it’s also important to note that most amateurs might have a particular level (let’s say 10k) not because they are consistently 10k at everything, but maybe because they know a thing or two but consciously avoid certain areas. At least, maybe it is different for other people, but I know I do. So I think a 16k who plays in a style that the 10k tries to avoid might get zero benefit from that particular reviewer even if much stronger player on paper, and in some cases may even be misled.

[tonybe]

I’m enjoying this discussion - however, I’m a bit concerned that we are focusing so much on the TREES that we are losing track of the FOREST - i.e. the possibility of getting a functional game review queue up and running

Is there a way to GUARANTEE that every single peer-to-peer review will be of the finest possible quality?

NO - and IMHO - this is one of those situations where “the perfect is the enemy of the good” - in other words, I believe that it might be more valuable to just get a peer-to-peer review system up and running with A FEW functional guidelines / suggestions, allow it to hit the ground running, and then adjust as friction points come up - rather than discuss this to the Nth degree and never actually decide on what to build.

As such, I suggest ideas that will SUMMARIZE the suggestions posted above with the aim of finding a middle ground that most can agree on, and then do a live test with actual humans, see how well that works, and adjust from there.

Examples: for SDK / DDK games, volunteer reviewers can only review games from players that are 8 kyu weaker than them. Dan reviewers need only be 1 dan stronger than the player they review (i.e. a 4 dan player can only get reviews from 5 dan and up, etc).

A given review should (in aggregate) be no shorter than 800-1000 words (about 1.5 to 2 pages of printed text in 12 point font on a blank page - with the understanding that these will be spread out over many comments across various important points throughout the game. Conversely, try to limit your reviews to under 5000 words, as this might overwhelm the reader.

My 2 cents.

[jlt]

The number of words is not a good way to measure the length of a review, the number of variations is more relevant. I’d say a review should contain between 5 and 20 variations.