Life In 19x19

Some niggle in my brain made me check something today. I wanted to check up on Yi Ch'ang-ho.

Although he is going backwards at an alarming pace, his career record (in terms of games I have) is STILL touching 70%. That's Go Seigen/genius level. Of 1,956 games he won 1,350.

But... This genius, this Korean national treasure, has only been able (by my count) to play 30 games this year. And of those, a high proportion are pair go, exhibition games and invitation games. In proper play almost every game is at time levels as ridiculous as 10 minutes each.

He's not alone in this. Indeed his once great domestic rival Yu Ch'ang-hyeok has almost been wiped off the playing scene for the past few years. His great international rival Chang Hao has had considerably fewer games this year, so it doesn't just happen in Korea.

But if you focus on Korea, many famous tournaments have recently disappeared and the number of events available for pros to play in has worsened, both quantitatively and qualitatively (and financially). If we try to assess Korean events in the round, we can see two patterns. One dominant one is that the time limits are overwhelmingly at the Mickey Mouse level. The other, less pronounced quantitatively but significant in terms of prestige, is that several Korean events are now hosted in Japan or China (even exhibition games!). Furthermore, several Korean stars now focus mainly on not just international events but on playing in China (mainly in the leagues). We have to wonder whether there is some cause and effect, and it seems legitimate also to wonder whether Korean go is being mismanaged.

To a large extent, Yi Ch'ang-ho, like many players of his generation (he is 43), is of course suffering from the "yesterday's man" syndrome. But a noticeable difference compared to veterans in Japan (especially) and China is that veterans there still play more in significant events with significant time limits and don't rely on scraps such as the trivial veteran's only games in Korea. The Japanese and Chinese veterans get to play regularly with young players at decent time limits and that seems to keep them fresh as well as afloat. In Korea, I do wonder whether the diet of murine games is actually eating away of Yi Ch'ang-ho's strength. It seems like asking a thoroughbred stallion to stop racing and to pull the milk-cart instead.

Yes, it's the way of the world. But it's still sad, and it doesn't even have to be quite like that. We still revere Go Seigen. Yi Ch'ang-ho's lifetime results match his.

Remember the gameless as well as the homeless this Christmas.

Here is Lee Changho's game record in 2018 according to go4go (so some games must be missing).
I don't know what the time settings are, but almost all his opponents were much younger than him.

Stopped reading at Mickey Mouse. You may think I'm egotistical, but I think the term is unnecessary, unkind, and reduces credibility in the post content.

Note here, I am attacking the TERM and not you, which is in contrasts to the comment about my being egotistical.

Yada, yada, you used the term 60 years ago, but there are a lot of other old terms that used to be the norm, which are offensive today.

Anyway, to the point, I find nothing shameful about Lee Changho's career, and he's still one of the most well know Korean pros in history. You can blame results on time settings, but it's pretty unscientific to jump to that conclusion given all of the other variables (I don't know, spending time with his daughter, perhaps? New styles of playing go? A generally stronger competition?)

Yeah, it's a shame that they have those time settings in Korea!

Whatever

Look. I calmed down a bit more, so maybe I can express myself more clearly.

We’ve done the whole Mickey Mouse discussion a million times. I’ve clearly expressed that I don’t like the term, and the nice thing to do would be to simply use a different term to express your intent. Same way as how you've been asked not to use the term "oriental" - I don't think it's unreasonable. There are many other ways to word things, but that probably wouldn’t be as fun for you on L19, right? Does it make me egotistical for thinking that? If there’s a particular phrase that you’d like for me to refrain from, I’d be happy to use a substitute.

Anyway, the argument as stated before is summarized well with this quote:

John Fairbairn wrote: With modern go, especially the Mickey Mouse events in Korea, you just want to be the first to start a successful fight. It's basically a crapshoot, but at least if you can control when and how the fighting starts you may have an edge. 

Yeah, got it. Crapshoots.

My argument is also simple: To jump to the “crapshoot” assumption on an entire class of professional games, based on your arbitrary decision that the time limits are too fast is both without data, and not necessarily accurate. As a result, we see a bunch of pseudo-persuasive ideas to try to reinforce the point:
* Kasparov agrees on Mickey Mouse
* Lee Changho must have bad results because of the time limits
The strongest evidence provided against the time limits has simply been the opinions of other pros. Everything else has been not based on any sort of real data. Simply the assumption that less time means more mistakes.


But it’s not an obvious assumption to make. Josh Waitzkin, child chess prodigy, and champion of the US Junior Chess Championship in the 90s, writes in his book, The Art of Learning:

Josh Waitzkin wrote: For a number of years, when notating my games, I had also written down how long I thought on each move. This had the purpose of helping me manage my time usage, but after my first session with Dave, it also led to the discovery of a very interesting pattern. Looking back over my games, I saw that when I had been playing well, I had two- to ten-minute, crisp thinks. When I was off my game, I would sometimes fall into a deep calculation that lasted over twenty minutes and this “long think” often led to an inaccuracy. What is more, if I had a number of long thinks in a row, the quality of my decisions tended to deteriorate.

And sure, this is chess: maybe the same rules don’t apply to go - I’m not saying that Kasparov agrees, either. Furthermore, go is a different game, so maybe the two- to ten-minute think that Waitzkin found valuable in chess would translate to different time limits in go.

I have only one point in saying this: the idea that increased thinking time necessarily results in increased quality is an assumption. A number of factors are at play, and the equation is much more complicated than that. To claim that 10-minutes isn’t enough time for a quality go game is unfounded and without real data.

—

So how do we get real data? And how do we define a “quality” game? These are undefined questions. The closest thing I can think of in today’s go world is to use our super-human AIs to evaluate games for us. But even then, how do we interpret the results? Is a game that maintains a close to even result of high quality? Or is it better to have less variation in the game?

As a small experiment, I compared a few games, below.

—

Game 1: 42nd Meijin Title Match, Round 1 <— This one’s OK to use as a non-blitz game, right? The definition of blitz is without data and arbitrary, so I’ll have to assume that it works.

Anyway, I ran through the first 150 moves, giving over 1000 payouts to each move with Elf, using Lizzie. Here’s a screenshot of the win rate graph, as perceived by Elf: How do we interpret the win rate graph? Surely, it’s not stable. Perhaps that’s partially a characteristic of Elf, but nonetheless, the volatility of the graph is pretty extreme. Around move 50 or so, it looks like black makes a huge mistake, and win rate drops below 25%. But then a short while later, win rate for black is approaching 70%. No, wait, a move later, it’s below 50 again, then below 25. Now up at nearly 75%. No wait, down below 50%, now up close to 75% again. Now down to 25%, no 50%, oh wait, now the game is lost.

Recall again, the view of modern go in “many” Korean events:

John Fairbairn wrote: With modern go, especially the Mickey Mouse events in Korea, you just want to be the first to start a successful fight. It's basically a crapshoot, but at least if you can control when and how the fighting starts you may have an edge. 

Now, personally, I find it rude to call a professional game a “crapshoot”. But if we trust Elf’s evaluation, certainly, it’s hard to say who’s gonna win this game!

—

So how about a recent “blitz” game? Again, we are choosing what is considered blitz somewhat arbitrarily, but the Crown Haitai Pro Go Tournament looks reasonable - it’s the most recent Korean tournament I see game records for. They use Fisher-style time with 10 minutes, plus 20 seconds per move.

Game 2: Crown Haiti Pro Go Tournament #1 How about this graph? Well, things start out somewhat evenly, and black makes a mistake, thereby dropping to a win rate of less than 25%. White doesn’t capitalize on its, and has a win rate exceeding 90%, and then a short while later, black makes another mistake. Certainly some ups and downs from Elf’s perspective. Compared to the Meijin game? Hard to say, but perhaps less volatile. The lead goes back and forth with larger swings, but fewer times.

Which game is higher quality? It’s not apparently clear to me.

How about this, another game from the same tournament?

Game 3: Crown Haiti Pro Go Tournament #2 This game is less obviously one sided - a majority of the game lies within the 25~50% area. There are fewer ups and downs than either of the other two games, and the swings are less extreme. Does it mean higher quality? I don’t know.

—

What’s the point?

I don’t like the term “Mickey Mouse” when referring to time limits:
1. It’s a gross over-generalization: we bucket a bunch of modern pro games into a single category, and classify them in our minds as blitz
2. It’s not apparently clear that there is even a significant difference in quality between these so-called blitz games and ones with longer settings. All examples indicating that the quality is lower have been anecdotal and taken from a handful of old pro opinions at best.
3. It’s not respectful to pros playing in the tournaments - the term is pejorative and insulting. You can claim that you’re only dissing the time limits, but for a number of pros, this is what they’ve chosen to do for their livelihoods.

And sure. The analysis that I’ve done here with Elf? I probably trust it less than most people might. But it’s the only way I could think of to do any sort of objective comparison of a few games. And maybe the time limits for these tournaments are too fast, in the sense that they result in more mistakes by the players - maybe not. I don’t know.

But I’m formally expressing here that I don’t appreciate the “Mickey Mouse” term, and I would personally appreciate it if other terms would be used (e.g. blitz). Better yet, I’d prefer more specifics about games or particular tournaments, rather than bunching a huge number of tournaments into a single bucket, and pretending that they’re all of the same nature. That’s not scientific.

Of course, nobody is under any obligation to honor my request, and have the freedom to post what they’d like on the forum. I’m also exercising this freedom by politely expressing my preference.

Kirby wrote:The strongest evidence provided against the time limits has simply been the opinions of other pros. Everything else has been not based on any sort of real data. Simply the assumption that less time means more mistakes.


But it’s not an obvious assumption to make. Josh Waitzkin, child chess prodigy, and champion of the US Junior Chess Championship in the 90s, writes in his book, The Art of Learning:

Josh Waitzkin wrote: For a number of years, when notating my games, I had also written down how long I thought on each move. This had the purpose of helping me manage my time usage, but after my first session with Dave, it also led to the discovery of a very interesting pattern. Looking back over my games, I saw that when I had been playing well, I had two- to ten-minute, crisp thinks. When I was off my game, I would sometimes fall into a deep calculation that lasted over twenty minutes and this “long think” often led to an inaccuracy. What is more, if I had a number of long thinks in a row, the quality of my decisions tended to deteriorate.

And sure, this is chess: maybe the same rules don’t apply to go - I’m not saying that Kasparov agrees, either. Furthermore, go is a different game, so maybe the two- to ten-minute think that Waitzkin found valuable in chess would translate to different time limits in go.

Waitzkin is not the first to realize that decisions taken after long thought are not necessarily better, but may even be worse than decisions made more quickly. In my own reading, Terence Reese in bridge and Botvinnik in chess made similar observations. Botvinnik even recommended taking note of time per play in order to identify problematic plays, by virtue of taking a long time.

But I question Waitzkin's suggestion that thinking a long time produces bad decisions. OC, it can. Reese speaks of "dithering", and Kotov lampoons similar indecision. For most people, most of the time, I think Botvinnik is right. Some decisions take a long time because they are difficult to start with, and difficult decisions are often wrong. The difficulty might arise from being off one's game, for one thing. The question of a series of "long thinks" in Waitzkin's case is probably a separate question.

What time limits are optimal for go (for humans)? Chess and go are different, and, IMO, go can use shorter time per move than chess, but for fun I tried applying Waitzkin's two to ten minute "crisp thinks" to go. The harmonic mean of 2 min. and 10 min. is 3 min. 20 sec. At 120 moves per person per game, that comes to 400 min., or 6 hrs. 40 min. That's compatible with time limits of 5 hrs. plus byoyomi. Certainly less than the 20th century Japanese time limits of 10 hrs., but fine for games that finish in one day. Certainly not Goofy time limits, to pick a different Disney character.

Hi Kirby,

A number of factors are at play, and the equation is much more complicated than that. To claim that 10-minutes isn’t enough time for a quality go game is unfounded and without real data.

Happy holidays.

I agree the curve is multi-dimensional and I have no idea of its actual shape.

However, my hunch is that if pros were forced to play at 1-second-per-move, the 'quality' ( debatable term ) would drop measurably ( compared to, say, 1-minute-per move ).

At the same time, I'm also not convinced 2-day games ( like the Honinbo ) are of much higher said quality ( again, debatable term ) than say that of 1-day games. My sense is that the main difference between a 2-day game vs. a 1-day game is the viewership ( the number of eyeballs x time, the original purpose in printed newspaper days ) and not the 'quality'.

( I'm very curious about the shape of said multi-dimensional function. )

Now, now, we all know blue hedgehog is the real name for speed games.

Faster games showcase professional perception. Viewers may migrate towards the depths of the ocean or the smell of roses on a grassy path, but those would require long term strategies and balances of the benefits and drawbacks of each type.

To that end, the only short term solution I can conjure is a tournament with increasing time limits each round.

I totally fail to understand the vociferous objection to the phrase "Mickey Mouse".

I don’t like the term “Mickey Mouse” when referring to time limits:
1. It’s a gross over-generalization: we bucket a bunch of modern pro games into a single category, and classify them in our minds as blitz
2. It’s not apparently clear that there is even a significant difference in quality between these so-called blitz games and ones with longer settings. All examples indicating that the quality is lower have been anecdotal and taken from a handful of old pro opinions at best.
3. It’s not respectful to pros playing in the tournaments - the term is pejorative and insulting. You can claim that you’re only dissing the time limits, but for a number of pros, this is what they’ve chosen to do for their livelihoods.

.....

Of course, nobody is under any obligation to honor my request, and have the freedom to post what they’d like on the forum. I’m also exercising this freedom by politely expressing my preference.

No, you are being incredibly rude.

First, you have instantly subverted a thread intended to be about a player I have met and admire into YOUR obsession about something completely different (doubly inexcusable as an admin guy).

Second, you are trying to badger me into changing my native language (which is not American). I had the same experience with 'oriental'. An American admin here asked me to stop using it because some Chinese people in America didn't like it. He preferred 'Asian'. I pointed out that 'Asian' here in Britain means Indian/Pakistani and that a local mall near where I live, which was a collection of Chinese, Japanese and Korean shops, called itself 'Oriental City' (and the term can be found in that usage all over London). I refused to change. This is an international forum.

Third, you don't like 'Mickey Mouse'. I don't like ultra-fast games. You are badgering me into watering down my dislike. And you have never explained why MM is unworthy of use, whereas I have tried to explain why I don't like ultra-fast games. Using MM rather than ultra-fast is because it adds the flavour of 'it disnae work' - which is my view in a nutshell, and that specific point is more important than just my preference.

Fourth, you are the one making wrong assumptions. Not only des MM not mean what you think it means, MM was never about you. It was also never about lack of respect for pros (although I admit I don't feel any special respect for them collectively compared to other human beings). If I am disappointed with anyone, it is with those who run go. They've gone for the quick fix and glitz of tv and the internet. I believe (but can't know, of course) that this will end in tears. If the wheel does fall off go, people like me who have spent 50 years on the game are likely to feel it's all been a waste. I think you should learn to respect other people's feelings apart from pros (who don't all even feel what you say they feel - many of them dislike ultra-fast games).

Fifth, it is self evident that the quality of games must suffer if played too fast. Given that a third of a game is the endgame and the quality of that depends on having time to count, a third of the game is at risk straightaway. I believe the middle game is also at grave risk.

Sixth, no-one else has ever expressed dislike for MM. So why should the world centre round you? (centre notice, not center - or do you object to that, too?).

Debates about the way a game is run are common. There was a major one recently in golf when commentators and fans enjoyed an orgy of debate over the Woods-Mickelson match. I don't know what the overall conclusions were in the States but over here it was apparently mostly seen as tacky and brought a grave risk of introducing excessive gambling into the game. We are entitled to have such debates also about the way go is run.

I think I know why you took an instant dislike to me, years ago, and it was nothing to do with Mickey. So do try not to "diss" him as well. It's not "cool," "dude".

In an ongoing fight between the common sense and raw data, only time can tell the victor. We can keep going from 'it makes sense' to 'there's no proof' and back, but it's pointless. Intuition helps us when there's no data to use as foundations for our decisions and it gets refined as we acquire more knowledge. I understand this is not The Rational Way™, but at the end of the day, one has to accept the possibility of agreeing to disagree for the sake of health of the discussion.
Also, don't assume malice, it's unhealthy. Irritation can be understandable, but there's no need to fire the cannons at mice (it's also counter-productive).

(I understand that, since I'm an anonymous nobody on the Internet, who am I to give lessons on this, but I think this is something we generally agree upon.)

As for the topic at hand, I also strongly dislike short time settings. For me, this, however, is just a specific example of a wide variety of methods of appealing to the general public in an entrepreneurial way, i.e. securing profit. I played video games competitively for twelve years of my life and observing the rise of e-sports and the shift in mentality of players and game producers alike has been an overall excruciatingly painful experience. It was unfathomable to me years back that people who barely indulge in a title can have such outstanding impact on how a game evolves, or that people who commit years of their life to mastering their passion are forced to adapt to absolute bullshit (not to mention those who choose the professional path and have to accept playing their part as Internet celebrities), yet here we are. Whether this is in any way relatable to Go, it's arguable; however, following various e-sport scenes over the years, one would have to work hard to convince me that reducing game time is not a really bad trend.

Quality hasn't been defined. Here is a possible definition, which may be questionable but which is operational.

The size of a mistake is the difference between the winrate after the move and before the move (it's generally a negative number if measured by a superhuman bot).
The quality of a game is the size of the biggest mistake.

We say that a given player is better at slow games if the median quality of his slow games is greater than the median quality of his blitz games.

One could refine this definition by just looking at the fuseki, or the middle game, or the endgame.

The endgame is known to be one of Lee Changho's strongest points. Since the endgame requires precise calculations, it wouldn't surprise me that he is much better at slow games than at blitz games.

jlt wrote:Quality hasn't been defined. Here is a possible definition, which may be questionable but which is operational.

The size of a mistake is the difference between the winrate after the move and before the move (it's generally a negative number if measured by a superhuman bot).

Two problems with that definition:

1) There is no single winrate for any play. All win rates depend upon the players making errors, but which and what kind of errors are unknown.

2) In theory the winrate before and after a perfect play should change, by some unknown amount and direction. For instance, suppose that at some point Black has a winrate of 90%. It is quite possible that with perfect play by both players White will win the game. In such a case, by this definition, Black's perfect play will appear to contain a number of mistakes.

As for the first problem, at the moment, aside from certain situations, such as the late endgame and ladders, the best estimate of the value of a position for a player is its winrate, as calculated by a top bot, or perhaps by the average of its winrates, as calculated by a number of top bots.

As for the second problem, the correct comparison is not between the winrates before and after a play, but between the resulting winrates produced by different plays. For instance, suppose that Black is to play, and play A produces a winrate estimate of 54% for Black, play B produces a winrate estimate of 52%, play C produces a winrate estimate of 57%, etc., and the top winrate estimate of all possible plays is 58%. We may then estimate play A as a 4% error, play B as a 6% error, play C as a 1% error, etc. (or we might use a different measure, such as the log of the odds ratio). It does not matter if the winrate estimate of the position before a play was 50%, 60%, or some other number.

Yes, I am aware that my definition has problems, but I just wanted to come up with a definition which is more precise than just looking at the winrate graphs, but which is not too difficult to use in practice.

John Fairbairn wrote:But... This genius, this Korean national treasure, has only been able (by my count) to play 30 games this year. And of those, a high proportion are pair go, exhibition games and invitation games. In proper play almost every game is at time levels as ridiculous as 10 minutes each.

I don't know about Lee Changho in particular, but I heard from multiple sources that playing with short time limits does very little to the quality of a pro's game in general.
I heard that pros estimate the difference to be only 1-2 points.

About the "Mickey Mouse" labeling of games with shorter time-limits: I notice that I also have a strong negative emotional reaction to it, somehow I associate it with a statement that reflects lack of respect for the games and for the pros that play them. Rationally I understand that you don't mean disrespect, but emotionally I dislike reading/hearing this association.

sorin wrote:I don't know about Lee Changho in particular, but I heard from multiple sources that playing with short time limits does very little to the quality of a pro's game in general.
I heard that pros estimate the difference to be only 1-2 points.

That sounds counter-intuitive to me. Amateurs play much better with longer thinking times. Leelazero plays much better with 10000 visits than with 100 visits. How can that be different for pros? What if they have to solve a complicated life-and-death problem? Does it mean that if they can't find the solution in 30 seconds, then they won't find it in 30 minutes either?