pareto principle applied to go

[moha]

Pareto's rule is known to apply only 73% of the cases.

...that will make the biggest difference on your winning percentage.

I think the choice of objective is crucial here. Things that make you stronger not necessarily increase your winning percentage that much (and vice versa). The principle seems to apply well for winrates, but doubtful for getting stronger.

Winrate, as you noted yourself, depends heavily on a few simple things. I would add one thing to your list: overplays/tricks. In my experience this is the single most important thing for winning casual online games with normal time controls. With some practice, an overplay that doesn't lose too much even with correct response (but gains advantage otherwise) is much easier to play than to refute, so such moves are strongly +EV. One may even confuse them with complications and "severe" moves.

Getting stronger (or winning slow games) seems to be a different thing. Although I think there is one thing that matters the most here: reading. But if we exclude that and only consider go knowledge (which is how I interpret earlier posts), then see Alphago's network. It captures a great part of possible go wisdom, yet it's strength is limited. (It was shown that a pure network player has half of the elo of the full version - around high dan amateur level.) So there are limits for how far the remaining 99% can take you.

[oren]

Things that make you stronger not necessarily increase your winning percentage that much (and vice versa)

This confused me. While maybe not one for one, I would expect a very high correlation between learning things that make you stronger and your winning percentage going up. Granted you have to adjust for rank changes, but I don't think that's what you're saying.

[moha]

Things that make you stronger not necessarily increase your winning percentage that much (and vice versa)

This confused me. While maybe not one for one, I would expect a very high correlation between learning things that make you stronger and your winning percentage going up.

I meant in the case of casual quick-ish online games, and also not for really strong players. There, getting stronger lets you "win" more games (like outplaying the opponent, getting a decisive advantage). But the total correlation is less than one would think, I actually doubt that winning a quick game gives you even a 67% chance for having it counted in your favour ("winning a won game"). On the other hand, getting better at overplays and tricks will not really make you stronger (in slow games), but your winrates will go up (as you will "win" a percentage of your losses as well). This probably also depends on playstyles - some styles benefit better from more go knowledge, some styles less.

[djhbrown]

I've been reading about learning, and one method (listed partially below) is to:
1.) deconstruct a learning topic into small learnable units
2.) identify the ~20% of these units which can be learned to achieve ~80% benefit (i.e. the pareto principle)

the "method" sounds like it comes straight out of a 18th century conception of schooling - it's not so much a method as a timetabling convenience. Before gestalt psychology or DNA were discovered, grammaring was all the rage - imagining that everything in the world could be diced and rediced into neat unitary chunks. It was prompted by Linnaeus, who sent the field of taxonomy down countless blind alleys, and by some Bishop whose name i forget who tried to circumscribe English by a context-free grammar based on what at the time were thought to be the rules of Latin.

So it sounds like the source you read about learning hasn't learned anything about what has been learned about biology or language or neuroscience since then.

Be that as it may, it's still almost certainly the case that x% of effort produces y% of value, where x<y. In the case of Go, i would hazard a guess that the numbers are closer to 1:50 than 20:80.

But that doesn't mean there is a magic sauce than can make you lose 12 pounds of fat in 2 weeks - all the evidence from the biographies of the best players in the world is that the only way to get stronger quicker is to
1. kill yourself and be reincarnated so you can start learning earlier in life
2. have brain surgery to make yourself more like an autistic savant with a prodigious photographic memory, so you can emulate the feat of the 14 year old boy who strolled past our game at the London Go Centre in 1978 or 9 (i forget which), glanced at the board for half a second, and casually remarked "the corner is ko". My opponent and i looked at each other in astonishment as he wandered off. He was 4 dan at the time, but gave the game up because it was too easy for him.

PS Learning Go and learning language have only one thing in common: learning. Everyone can become strong at any language, but few ever become strong at Go, just as few become strong at golf, or music, or mathematics, or anything artificial. The easiest thing in the world to learn is Chinese - we know this because so many people speak it with ease.
PPS Pareto had studied the ratio of rich to poor, but i think his numbers were way off in those days, just as they are today; the inbalance is much greater. Eg, how much land do you own?...

[EdLee]

Hi Kirby,

Speaking of learning:
Carlsen v. novice

[Knotwilg]

Things that make you stronger not necessarily increase your winning percentage that much (and vice versa)

This confused me. While maybe not one for one, I would expect a very high correlation between learning things that make you stronger and your winning percentage going up. Granted you have to adjust for rank changes, but I don't think that's what you're saying.

This is due to the dogma "lose x games as fast as you can". It's a false dogma which somehow soothes those who like to study Y (which is a fine objective), then deep down get frustrated their hard study of Y doesn't pay off but excuse themselves with "studying and winning have nothing to do with each other". And then winning is equated with trickery.

No. Winning is important: it gives you the positive feedback on what you've learned.

[Kirby]

Hi Kirby,

Speaking of learning:
Carlsen v. novice

Nice, Ed. I saw this posted on Facebook a little while ago. His goal seemed a little ambitious

That being said, I’d love to be able to do a backflip.

[Kirby]

Things that make you stronger not necessarily increase your winning percentage that much (and vice versa)

This confused me. While maybe not one for one, I would expect a very high correlation between learning things that make you stronger and your winning percentage going up. Granted you have to adjust for rank changes, but I don't think that's what you're saying.

This is due to the dogma "lose x games as fast as you can". It's a false dogma which somehow soothes those who like to study Y (which is a fine objective), then deep down get frustrated their hard study of Y doesn't pay off but excuse themselves with "studying and winning have nothing to do with each other". And then winning is equated with trickery.

No. Winning is important: it gives you the positive feedback on what you've learned.

In my opinion, greater win rate is equivalent to greater strength. Win rate is an objective measure of real results.

One can argue that trick plays, etc., are weak, but if they work for against a given level, a player that knows the trick is strong at that level.

Saying someone is strong or weak independent of win rate is hypothetical nonsense.

[Bill Spight]

One can argue that trick plays, etc., are weak, but if they work for against a given level, a player that knows the trick is strong at that level.

Saying someone is strong or weak independent of win rate is hypothetical nonsense.

Well, if you want to improve, don't play against opponents who fall for your tricks. Play against stronger opponents.

Janice Kim went to study in Korea for a summer when she was around 2 kyu, IIRC. She came back around 2 dan. Since she had decided to become a pro, her teacher forbade her to play against amateurs. Even against stronger ones. No point in picking up bad habits from a 4 dan.

[Kirby]

One can argue that trick plays, etc., are weak, but if they work for against a given level, a player that knows the trick is strong at that level.

Saying someone is strong or weak independent of win rate is hypothetical nonsense.

Well, if you want to improve, don't play against opponents who fall for your tricks. Play against stronger opponents.

Janice Kim went to study in Korea for a summer when she was around 2 kyu, IIRC. She came back around 2 dan. Since she had decided to become a pro, her teacher forbade her to play against amateurs. Even against stronger ones. No point in picking up bad habits from a 4 dan.

Agreed. Improvement is a different story than strength. For improvement, I believe it’s good to optimize your internal model of how go works as much as possible.

Just that to actually measure how well you are doing, game results over time are the only real indicator of strength.

[moha]

In my opinion, greater win rate is equivalent to greater strength. Win rate is an objective measure of real results.

One can argue that trick plays, etc., are weak, but if they work for against a given level, a player that knows the trick is strong at that level.

Saying someone is strong or weak independent of win rate is hypothetical nonsense.

I didn't say independent, I said the correlation is less than one may expect, for casual and fast games. I also didn't say overplays are weak, on the contrary, I said they are strong since refuting them is harder than playing them, hence +EV (still about fast games).

But the situation is different for slow games, since it is less likely that the opponent makes a mistake, or don't find a strong enough response for an overplay. Things that help you winning more fast games (which is what winrate is usually) are not exactly the same things that would help winning slow games. Things mentioned earlier (no resign, paying attention to remaining time - and overplays) belong to the former class. The various directions for collecting go wisdom, trying to find the objectively best moves belong to the latter.

OC, the two are not completely distinct, the error-inducing style that is well suited for fast games can still be useful in slow games, when one tries to turn a disadvantageous game around. Also the mindset when choosing moves - you may target 101% efficiency in slow games, and 130% in fast ones. Still the differences remain, also in the correlation rate mentioned (the effects of timeouts, misclicks, yose blunders etc. - the strength-independent factors - in percentage terms).

Improvement is a different story than strength.

I feel some inconsistency here.

[Kirby]

I wasn’t particularly responding to you, moha, so I don’t know why you think I was putting words in your mouth. I agree that playing under different time settings could require different skill sets.

What do you find inconsistent in my viewpoint?

[moha]

I wasn’t particularly responding to you, moha, so I don’t know why you think I was putting words in your mouth. I agree that playing under different time settings could require different skill sets.

What do you find inconsistent in my viewpoint?

Sorry, I surely misunderstood you then.

What seemed inconsistent is first rejecting the idea of any discrepancy or weak correlation between strength and winrate ("Win rate is an objective measure of real results ... Saying someone is strong or weak independent of win rate is hypothetical nonsense"), basically denying the random factors in quick casual games. At the same time hinting at significant discrepancies between strength and improvement.

May be a wording issue. For me, "strength" is the rough expected winrate in slow, serious games (though I'd be tempted to even here exclude games that are awarded to the loser). I'm not sure how you would define "improvement", but I'd guess improvement of strength. So I'm not sure how to interpret any discrepancy between these two.

[Kirby]

What seemed inconsistent is first rejecting the idea of any discrepancy or weak correlation between strength and winrate ("Win rate is an objective measure of real results ... Saying someone is strong or weak independent of win rate is hypothetical nonsense"), basically denying the random factors in quick casual games. At the same time hinting at significant discrepancies between strength and improvement.

I believe that random factors may play a role in not only quick casual games, but also for serious games. But to me, the definition of "strength" is specifically defined as how good you are at winning games.

May be a wording issue. For me, "strength" is the rough expected winrate in slow, serious games (though I'd be tempted to even here exclude games that are awarded to the loser). I'm not sure how you would define "improvement", but I'd guess improvement of strength. So I'm not sure how to interpret any discrepancy between these two.

I don't know why the definition of "strength" should be limited to a particular time setting - random mistakes and hiccups in thought happen under all time settings. I could, however, see someone defining strength by saying, "I'm pretty strong at games with 30 minute time setting" or conversely, "I'm pretty strong at games with 30 second time setting".

Anyway, the distinction I am making with "improvement" is simply that strength, to me, is a measure of how good you are at winning games at this moment in time. Trick plays, etc., may hurt in the long run and not be good for your improvement.

[moha]

I believe that random factors may play a role in not only quick casual games, but also for serious games.

I agree, but the extent of this is vastly different in those two cases. Which is why I suggested that the 20/80 principle also applies completely differently: no resign, watch time, overplays for quick games, reading for serious games.

But to me, the definition of "strength" is specifically defined as how good you are at winning games.

Hmm. Here is an example of a particular player. When losing a game, he would play to the passing phase, maybe even requesting a resumption once or twice, then play some meaningless stones inside opponent's territory. Then he would start to capture dead stones (while the opponent kept on passing - this server forces Chinese rule so free for him). Then a few moves before all captures were done, he would play an atari or some threat in territory again. This technique did get him quite a few games, so his winrate increased, but I wouldn't say this increased his strength as well. And while this worked in quick/casual games, would be pointless to try in slow/serious games.

I don't know why the definition of "strength" should be limited to a particular time setting - random mistakes and hiccups in thought happen under all time settings.

Sure, this is the noise on the signal. It's simply easier to focus on parts with less noise (hopefully small enough to be ignored safely).