pareto principle applied to go

[Kirby]

It’s not more noise in a blitz game - quick thinking and avoiding mistakes are important parts of the game. You can be strong at blitz, but weak at slow games, or vice versa.

I think your basis for a “strong player” is one that makes fewer mistakes in the abstract sense, ignoring the time dimension.

But time is one dimension that I believe players can get strong at.

Think of it this way - you have a player that can play optimally, but requires three weeks of thinking time per move. Is he strong? Maybe in some sense, but he sure wouldn’t do well in any tournaments.

I suppose the definition of strength depends on what you want to optimize. If you want to get strong at winning tournaments, you’d better be good at winning under those time conditions.

[moha]

It’s not more noise in a blitz game - quick thinking and avoiding mistakes are important parts of the game.

I think we won't agree on the "not more noise" part. And it's not about any kind of mistakes (some of which are OC part of one's strength or weakness), but the random-like things in particular. Which percentage of games are awarded to the loser in quick games, and in serious games? This is a factor winrate must be adjusted with, to get a reasonable idea about strength. While personal experiences may vary, I'd say these numbers are quite different in the two cases.

I think your basis for a “strong player” is one that makes fewer mistakes in the abstract sense, ignoring the time dimension.

Not really, I'd just like to simplify and avoid the need for the above adjustment step. Finding good moves in three weeks is not something I would consider strength in practice.

[Kirby]

Which percentage of games are awarded to the loser in quick games, and in serious games?

Let's stop distinguishing between "quick games" and "serious games" (we should actually use concrete examples - what time settings constitute a "quick game"?). Quick games can be serious - it's just a different type of game than a game played with longer time settings. In extreme circumstances, you probably exercise different parts of your brain.

Finding good moves in three weeks is not something I would consider strength in practice.

Why not? You said earlier that strength, to you, is the "rough expected winrate in slow, serious games". What is a "slow, serious game"?

---

Is there a particular time setting and/or range of times that you consider to be slow and serious?

[John Fairbairn]

Kirby: Given our past history of discussing fast/slow, I was tempted to follow the bush warbler: 歌は思えど　時にあらずと　声も立てず. But it's autumn not spring, so...

You are of course entitled to use "strength" in any way you like, and to choose to enjoy or respect fast games as much as you seem to. But I think you have to also acknowledge that others take a different stance, and even that you are in a small minority.

Chess has faced the same issue. They have decided officially (at FIDE level) to separate out classical, rapid and blitz as distinct modes, each with their own rating list. There is some correlation between each list (Carlsen tops all three) but there are significant differences in the make-up of each list and the point spreads.

Chess organisers and sponsors generally opt for classical time settings, and the classical rating is the one generally quoted as the mark of a good player. There is no fixed time setting but broadly speaking a classical setting is one where the players can only play one game a day. That definition can work in go, too. Behind this choice is a feeling that non-chess factors such as age, luck, agility (in pressing clocks), concentration, testosterone, psychology and sensitive bladders become much less important than in fast games. It is a feeling more than a proof, but, for example, Hikaru Nakamura, noted as the "Speed Demon" of chess, has just said that blitz games have become much less interesting to him now that he is older.

Go organisers, especially in Korea and mainland China, have experimented with fast speed settings, not for go-specific reasons but to meet the demands of television or the internet. But they seem to be climbing back down. The time settings in China are now rarely below 2 hours, and 2 hours 40 seems to be becoming very common. Korea is slower to change back, but they have edged byoyomi up from 30 seconds to 40 seconds (and of course many of their fast events have just disappeared). And of course, while the organisers/sponsors may have the last word, there are countless cases where pros themselves have complained of the plethora of fast time limits.

As to what "strength" means, it can't just mean win rate. Rank != rating. Obviously at some point a player considered strong has had to prove he can win lots of games against top opposition, but what it really means is that he understands the go-specific factors of the game better than most people. That is why top players still flocked to Go Seigen for study sessions even when he was doddery and would have had an abysmal win rate had he still been playing in tournaments. In a commentary, would you really take the view of a 14-year-old blitz winner over the different view of Cho Hun-hyeon?

There are competitive activities such as golf where we never say the best players are "strong" (in a win rate sense). They are either successful or not successful. This is probably one useful way to distinguish between time-sensitive physical sports and time-elastic mind sports. In games such as chess and go, the best players don't lose their game-specific prowess - they just get older and so factors not specific to the actual games occupy a bigger slice, the more so at fast settings.

[Uberdude]

Finding good moves in three weeks is not something I would consider strength in practice.

It beat Dinerstein and got me to 8 dan on old OGS!

I think part of the appeal of the idea that people have an inherent strength, and play at faster time limits reveals a weaker version via mistakes (but how much weaker varies from person to person) is that it makes slow players like myself feel better about themselves and gives us excuses for when we mess up and lose "Oh, woe is me! I would have won were it not for that pesky byo-yomi!" I do recall reading Lee Sedol's commented games book being struck by one point where he was talking about leading on the board versus leading on the clock: the impression I got was that he saw time management as an integral part of the game so you could do a positional judgement and say black is leading 5 points on the board after komi but white is half an hour ahead on the clock to it's even, or something like that. He saw it as an important part of training to be able to play good moves relatively quickly, and this is from the player who lobbied for longer time limits in his big match with Gu Li.

[Kirby]

Re: fast games vs. slow games

If you like “slow games” (which we never really defined), that’s fine. I like them, too. I just don’t see why it is so difficult for people to admit that playing well under fast time pressure is also a skill that can be considered, by some definition, “strong”.

Anyway, my main point to begin with was not about fast vs. slow. It was about win rate being the real indicator of strength.

As to what "strength" means, it can't just mean win rate. Rank != rating. Obviously at some point a player considered strong has had to prove he can win lots of games against top opposition, but what it really means is that he understands the go-specific factors of the game better than most people.

I would say that understanding go-specific factors of the game means that you are strong at go knowledge. If that doesn’t manifest in winning games, you are not strong at winning go.

For some, go knowledge may be more important than winning. But as go is a competitive game, I find it very reasonable to think of a strong player as a player that has a good win rate.

Analogies suck, but if I make one from the perspective of developing a machine learning algorithm: Let’s say I make a computer program with the objective of creating a STRONG go playing AI. I can have the most sophisticated model in the world, great results from training examples, and the best “knowledge about go” of any AI out there.

But the real test is if my AI can win games. If I have a fancy algorithm that “knows” a lot about go, it’s meaningless unless it has a strong win rate.

There is some benefit to go knowledge in regard to teaching, etc., but this is a different type of strength.

[Bill Spight]

One the fast vs. slow question and strength

Strength at go is not a single thing. There are a several components to strength at go, including, as JF says, testosterone level. There are different skills, as well. If I had to guess, I would say a couple of dozen. That is why strength at go is not transitive. Player A may regularly beat Player B, who regularly beats Player C, who regularly beats Player A, for instance.

Anecdotally, strength at bridge does not decline much with age. The oldest US National Championship winner, the oldest on his team, was 80, as I recall. I suspect that that is because bridge is a fast paced game. Played at normal bridge speed, a game of go would last 30 - 40 minutes or so. Seeing does not require much effort.

Years ago, Bruce Wilcox advised people to finish a game of go in 15 min. Now that's blitz! I'm a slow player, myself, so I have no experience to guide me, but for new players that is probably not such bad advice. The reason is that dwelling a long time on a bad line of play reinforces it in your brain. Even if you eventually discover that a certain play is bad, following a train of thought that assumes that it is a good play makes it easier to come up with it later. And coming up with bad plays, only to reject them, is a waste of mental energy. (Often this happens at an unconscious level.)

Much of what the brain does is inhibition, such as inhibiting the urge to make a certain bad play. And part of progress in learning is unlearning bad habits or the wrong things. In times of stress, such as time pressure, bad habits have a way of resurfacing as inhibition fails. "The return of the repressed," as Freud put it. Much better to learn good habits and the right things from the start.

Where do you find examples of good habits and right things in go? One obvious place is pro games. The study of pro games should be an important part of any study regime. For anybody.

[moha]

Although this narrowed down to fast vs. slow, I agree the differences about strength go beyond that.

In the example of a retired go legend, I would not say he is still strong as he used to be. His wealth of go knowledge is probably even richer now, and OC players want to learn from him, but aging reduced his reading ability thus his strength. Here I agree that strength has to do with current ability to win games. (Unlike go wisdom - this is almost the same as Alphago with search disabled: intact go knowledge but diminished playing strength.)

I actually prefer playing quick games over slow ones, in spite of the bigger random factor: risking a three hour game getting ruined (for both sides) by a glitch seems too much. I also recall the saying attributed to Paul Erdős, which translates like "it's ok to play go badly/poorly, but not slowly". One meaning behind this may be to avoid the game deteriorating into a pure reading contest (which is strength but not fun).

But even the best moves cannot do more than avoid defeat - winning a game needs the opponent's mistakes. But it does seem to matter which kind of mistakes those are. Accumulating an advantage by slowly outplaying the opponent seems to be an acknowledgement of one's strength. Winning a game by a huge swing in the theoretical value (a big blunder, or even timeout/connection loss) is often a luck factor (and may not increase future winrate either). To me, playing objectively good moves seems a key point of strength - even if that helps less in winning fast games.

I also recall the 4-player tournament some months ago, with one player from each of China-Japan-Korea, plus Zen as AI. Iyama's first game against the Korean player could test one's notion about what strength is, and how it is related to victory.

[djhbrown]

dwelling a long time on a bad line of play reinforces it in your brain.

True, but the best way to reinforce bad habits is to repeatedly play thoughtlessly, so Bruce would have been up a gum tree if that's what he meant, but i rather think that his "Instant Go" didn't put it quite so casually, and rather that he was looking for ways for people to be able to form a Gestalt of a position - an overview, a conceptual "big picture" - so they could tell the wood from the trees and thereby find meaningful moves instead of mindless stabs in the dark such as pulling out stones that are better left for dead because at least they create a bad taste (aji) in the opponent's mouth - as Alfie herself failed to do on one celebrated occasion. Bruce's work is a significant influence on Swim, even if it doesn't use his particular methods in quite the way he did.

Alfie has shown us that all the gestalts in the world are no substitute for reading until the cows come home, which is how she can upset even the world's best, as Michael Redmond is finding out the hard way through his extensive (and wonderful) analyses - in A0 v M game 3 i think it is, there's a moment when his sidekick observes that a weird-looking move turned out to be in just the right place 30 moves later, so unless you can read that far and more, you're never going to be able to comprehend her, let alone compete.

All this waffle about Alfie learning intuition is sheer nonsense - she's a reader par excellence of learned kneejerk reactions, and anyone like me with dodgy knees who can't be bothered to read should get a girlfriend instead of wasting life on a Sisyphean struggle suited only to autistic savants.

[gowan]

Only parts of Go can be learned by breaking things down into small units. We all know that in go it is important to think of the whole board so there is more than learning what happens in small areas. I would imagine that learning to think globally would be difficult to learn by looking at small parts. When we learn a human language, reaching fluency requires integrating all the small syntactical rules and vocabulary. We do it like the pros learn go, by doing it and observing what happens when we make choices.