What went wrong with AlphaGo ?

[dfan]

To my eyes, chess is more popular now than it was 20 years ago, even in these click and swipe times. So I would not give up all hope just yet.

[BlindGroup]

What happened in 1975 to about double the membership?

[Uberdude]

What happened in 1975 to about double the membership?

The first London Open was held in 1975 at Imperial College. In March the same year a permanent home for this and other Go events in London became available in the shape of the London Go Centre. This centre was opened with generous sponsorship thanks to the efforts of the great Iwamoto (9 dan), and was run until October 1978 by Stuart Dowsey and David Mitchell. It was a seven day a week centre for playing and teaching Go, a focus for Go publicity, and for the distribution of Go material. In the end it proved to be over-extended and was unable to attract enough members to be financially independent in the expensive London environment. During this period BGA membership was over 1000, boosted by the Go Centre, the Open Door television program on Go and a feature on Radio 4.

[Kirby]

Pessimistic rant:

Nice rant, but that's a question of what went wrong with LifeIn19x19.

AlphaGo has as much to do with L19 as it does with UK membership.

Fewer people around, fewer people discussing the game. You're an admin, ignore the problem or call it out and pretend it is the fault of the users.

What problem? People are just interested in chatting about other stuff, now.

[Kirby]

That's a shame. Am I right in thinking this is largely because you have fallen out of love with go (in large part due to arrival of strong go bots) and not because you think, were you mentally in the same place as a few years ago, the congress will be worse now than before? Or does the fact Andy Liu and Ryan Li will play an AI pair go game, or during a lecture they might talk about AI inspired moves like early 3-3 invasions, or your opponents play such moves actively turn you off?

I don't have any reason to believe the congress will be worse now than before. The organizers seem to have been doing a lot of work. There's even an app, text message notification of pairings, and several other things that have never been done before at other congresses. I guess I'd be turned off by too much AI related discussion, but I didn't really assume that'd be the case. Just less interested in go, overall. Maybe I'll change my mind later.

I still like it enough to be checking L19, after all

[hyperpape]

You're an admin, ignore the problem or call it out and pretend it is the fault of the users.

Are you saying you think Kirby has some responsibility for how many people come to L19?

[Simba]

BGA membership is a poor metric to use for go popularity.

People become members of something because they feel it presents enough value to be worth it.

From the mid-2000s onwards, with places like KGS becoming accessible and popular, the draw of national go associations decreased if they didn't offer something more than they did previously (perhaps the Russian Go Association did this, based on looking at their numbers).

For people who mostly play online like me, I see very little tangible value in being a BGA member.

The only thing I'd consider becoming a member today for would be the national championship, but honestly I feel even that is pretty spurious in its own legitimacy given several of our strongest players don't even enter. The eligibility requirements are also strange and arbitrary at points. Our play-off tournament for the title match this year had a kyu player in it even (all credit to him for qualifying, but the fact a kyu player got anywhere close to qualifying given how many dan players we have in the country makes the tournament really lack credibility). Don't get me wrong; it'd be nice someday to be 'officially' the strongest player in my country, but my mind says "hollow victory" when people like Matthew aren't in it. Yet the BGA would baulk at the idea of it being held online where it's more accessible and hence has a better standard of players.

Anyway, on the original topic: I think the original new wave definitely came, but it largely faded because there isn't much to keep ~20-25k people interested and motivated. Some still play, for sure, but only a small proportion. It's certainly the case that more of my friends have now *heard* of go. They just don't want to play it. It's quite simply viewed as a crappy, old, slow board game in a world that moves at 1000mph and is filled with vibrant colour. It's hard for us to compete.

[Javaness2]

According to Javaness2, UK membership peaked in 2003. For some reason, the European Go Database shows a peak for UK in 2008. The graph below shows that Russia managed to develop go well after that.

Yes, there is probably a trend that membership numbers increasingly reflects more strongly the tournament playing population. This reflects what Simba wrote as well.
The numbers are not really the point in themselves; it is just the situation I wanted to discuss. Was the game interesting but too hard to be fully accessible - something Paul Smith warned us of years back.

[Uberdude]

On the BGA numbers side topic:

BGA membership is a poor metric to use for go popularity.

Agreed

People become members of something because they feel it presents enough value to be worth it.

I don't think that's the only/main reason for some. Some might see it as a worthy cause to support, or simply reason "I play Go in Britain so I should join the BGA". Perhaps there is a generational difference with "what's in it for me?" attitude more common among the young (who are also generally poorer). Technological dinosaurs probably value the printed journal more. I recall we discussed this on gotalk when a poor student (who helped organise a BGA online teaching event) wasn't a member and questioned the value (it's £24 per year now, was £18 back then I think). There is a student rate of £10 (used to be £5) that doesn't get you the printed journal introduced as a result but it hasn't been popular afaik.

From the mid-2000s onwards, with places like KGS becoming accessible and popular, the draw of national go associations decreased if they didn't offer something more than they did previously (perhaps the Russian Go Association did this, based on looking at their numbers).

Agreed. I think Russia is doing well promoting/teaching Go. I also get the impression they succeed in getting more support from local businesses and government/civic organisations than we do. Having paid employees as I think they do instead of just volunteers surely helps. Also maybe they do a better job of making people join the association: I heard that back in the 70s with the London Go Centre Stuart Dowsey or someone would strong-arm you into joining the BGA on your 2nd visit. These days in England I think many club players (particularly who don't go to tournaments) aren't BGA members or even members of their club (I don't even know if the Cambridge University Go Society still has a membership secretary, it's free to play unlike e.g. London City). I think there's a legitimate concern that if you require people to pay/join you can turn some away, but also if you then give them something for it maybe it can help retain casual players and get them more involved (e.g. CUGOS members can borrow books from the club library, but that's not happened for years, plus with so much information online these days could be less of an incentive).

For people who mostly play online like me, I see very little tangible value in being a BGA member.

The only thing I'd consider becoming a member today for would be the national championship, but honestly I feel even that is pretty spurious in its own legitimacy given several of our strongest players don't even enter. The eligibility requirements are also strange and arbitrary at points. Our play-off tournament for the title match this year had a kyu player in it even (all credit to him for qualifying, but the fact a kyu player got anywhere close to qualifying given how many dan players we have in the country makes the tournament really lack credibility). Don't get me wrong; it'd be nice someday to be 'officially' the strongest player in my country, but my mind says "hollow victory" when people like Matthew aren't in it. Yet the BGA would baulk at the idea of it being held online where it's more accessible and hence has a better standard of players.

Are you supposed to be a member to play in PGETC? I agree the British championship is sadly weak these days and it's a shame Matthew(s) don't play in it or at least one of the younger generation beat him in the final as a "passing the torch" moment. But when I won I didn't think it made me "'officially' the strongest player in my country", just the strongest who bothered to turn up and a nice achievement I'd been aiming for. I hope you do play someday. How to stop cheating if online?

[jlt]

I looked at the number of tournaments in the last few years that were registered in the European Go Database:

capture.png (21.26 KiB) Viewed 9416 times

Of course, not all tournaments are registered on the EGD, nevertheless this graph seems to indicate that maintaining a high level of activity is important to promote go.

[Uberdude]

Impressive tournament growth in Russia. I wonder are these mostly "proper" tournaments as in dozens of people go somewhere and play a bunch of games a day, or could there be a lot of "club/ladder" tournaments in which more casual club games played over an extended period are entered into the rating system as a tournament, as I think these are now allowed subject to certain conditions.

Edit: clicking through a few there seem to be loads of tournaments with most players from one club, lots of ddks, probably school/junior events?
http://www.europeangodatabase.eu/EGD/To ... y=T180520B
http://www.europeangodatabase.eu/EGD/To ... y=T180707C
http://www.europeangodatabase.eu/EGD/To ... y=T180410C
http://www.europeangodatabase.eu/EGD/To ... y=T180512N
Bigger with dans:
http://www.europeangodatabase.eu/EGD/To ... y=T180623D
Some Go school league played on KGS (more info on Central Go School from Iwamoto Awards entry: http://www.iwamoto-awards.com/Awards/Po ... hp?A=27083)
http://www.europeangodatabase.eu/EGD/To ... y=T180319I

So a few might be analogous to e.g. the British online / junior league which isn't in EGD, but mostly this is just reflecting a massive amount of Go activity, particularly amongst children, which is great to see. No wonder we keep seeing new Russian mid-high dan children.

Sergei Pavlov must be busy submitting all these results!

[John Fairbairn]

I suspect I'm in the same camp as Kirby. From everything he's said over a long period, even before AI hit the scene, something has demotivated him slightly, and it sounds like what afflicts most of us: the responsibilities of real life - family, job, the future in general... The same afflictions no doubt explain why more people don't take up go passionately. At any rate, I think the decline or stagnation in western amateur go interest has absolutely nothing to do with lack of advertising or publicity, and in the few places where amateur go does thrive it is almost entirely because of the efforts of a few energetic or charismatic individuals. Once these individuals disappear, stagnation and decline re-assert themselves.

Some of us find a reason even so to continue with the game. I know, for example, that a big chunk of people used to find the social aspects attractive via clubs and tournaments. In England it was possible to play in a reachable tournament almost every weekend, and in London you could attend a club most nights of the week. But even in America, where distances make tournaments few and far between, their very rarity made them a high point of go life.

But both clubs and tournaments have been clobbered by the internet, long before AI. It's not just that it disinclined many people physically to go to an event, but the instant and rapid spread of anonymity, foul language and cheating was the very antithesis of social go. The craving for rip-offs (plus the same bad behaviour on the servers being directed at publishers) has killed off go books and magazines. The magazines in particular used to have an important social function, as did the book stall at tournaments.

There are, on the other hand, quite a lot of players who may not be passionate about the social aspects but who value tournaments as a way of testing themselves. While there are a few idiots who keep popping up to tell us they are going to study go obsessively for a year to see whether they can become pro, the vast majority of people who want to improve just want to measure themselves against their own personal yardstick. People like that, in the recent past, found that books and live interaction with other players were invaluable in providing opportunities and motivation for improvement. Maybe it's a generational thing that will disappear with time, but the internet does not seem to have provided satisfactory replacements for such people.

In my own case, both as a chess player and a go player, I was never the slightest bit interested in becoming a pro, or even in playing in tournaments, though I used to do some of that. I was always more fascinated by the cultural and historical aspects, but even deeper than that I was mostly fascinated by how top players thought, what made them strong. Why were Emanuel Lasker or Capablance stronger than their contemporaries - was it genes, background, politics, luck? Was Dosaku a freak who could compete well even today, but even in his own time how could he be so much stronger than the rest? Do modern Mickey Mouse games damage go? I have no complete answers to these questions but have enjoyed trying to find out. One reason I stopped going to tournaments, incidentally, was I got bored with the interminable post-game discussions along the lines of: "if he played there I would have played there", or "this is not joseki, that is." I wanted to hear people step away from tactics and delve into the realms of creative ideas, by saying "I played this because..." The only player I ever heard do that in British events was Matthew Macfadyen (and the fact he was/is our strongest player may not be unconnected with that; ubderdude seems to be in he same mould but I haven't met him yet).

So what I am saying is that the pressures of real life and the real reasons people play amateur go generally leave them inclined to have some negative feelings about how long or how intensively they will stay in the game. When something like AplhaGo comes along, it just adds to the existing malaise.

It doesn't cause the malaise, but how much it affects an individual can vary substantially. In my own case it has two effects. I am certainly not opposed to computer go (I worked for quite some time on the first computer shogi machines with David Levy), but I do feel demotivated by the way computers have undercut the traditional way of talking about the games. What I mean is that, in chess, where once a top player could say "I played this combination, even though it is probably unsound, because...", nowadays some kid can come along between nappy changes and say, "Stockfish shows there's a forced mate in 45 moves." The top player tells me something valuable about how a human thinks, but the kiddie tells me nothing but trivia. We are getting the same situation with winrates in go. But there's a more insidious problem, long apparent in chess. There, writers no longer give their true thoughts on a game. Out of fear, they check everything on a computer first, and that in turn changes what they write. So we no longer really hear what a human thought, but what a human pretends he thought. It is even worse in chess because three-quarters of the game being commented on may have already been played by the computer even before the first move, as part of opening preparation.

People often say that chess is thriving "despite" or "because of" computers. I'm not sure that either case is true. For me, as a fan rather than as a player, the activities of the top players are the main interest (so long as they share their thoughts). I have already adumbrated reasons why I have fewer reasons to believe they are really sharing their own thoughts nowadays, but top players outside the top ten in general are having a tough time in chess. I was struck by the fact that Britain's current top player, Mickey Adams, has played just 3 events in the first half of this year. That's a similar pattern for all the other top players I looked at. They can look much busier if you count their games, because top chess events are typically round-robins. But in terms of prize-winning opportunities, Adams seems to have had just three this year. And he's a top player who can expect some of the invitations (for usually not very substantial prizes) the chess world depends on. That can't be good for chess, surely? I can't see how chess computers are going to change that - except for cheaters, of course.

For amateurs, I haven't seen anything from the chess world that suggests computers have taught us anything. They have provided many more opportunities to play and practise, online or offline, so to that extent they have helped people improve, but they haven't taught them anything. Go and chess have different balances of tactics and strategy, so maybe it will be different in go? So far I have seen nothing to suggest AI can teach us anything in go either, beyond perhaps raising a few questions that we can ponder on - but even there different versions of the same machine seem to contradict each other most of the time. Furthermore, all the academic opinion I've read seems to conclude that such teaching is either a long way off, or a chimera because AI works on entirely different principles from humans.

With so much potential negativity about AI and computers, I infer from some comments that some L19 readers even find the AI threads annoying. I for one do find the "look at my shiny new toy" comments tiresome, but I enjoy the sage comments from people like uberdude and Bill Spight: they both add spice to my "how humans think" interest. Nevertheless, that enjoyment doesn't alter my generally pessimistic view of AI in go for me. Others no doubt will find positives and will continue to follow the game, but I suspect what they will be enjoying is not go as I know it, but some new hybrid (as has happened in chess). Which means we have to think in terms of apples and oranges now when talking about go.

[jlt]

Impressive tournament growth in Russia. I wonder are these mostly "proper" tournaments as in dozens of people go somewhere and play a bunch of games a day, or could there be a lot of "club/ladder" tournaments in which more casual club games played over an extended period are entered into the rating system as a tournament

I don't know how many are "proper" tournaments, but I can see that in 2017, 85 tournaments in Russia gathered 30 players or more (compared to 13 tournaments in UK). A significant proportion of smaller tournaments are tournaments for children and/or young players. Admittedly, many children are in the 15k-20k range, but I don't think children tournaments should be discarded from the data. To build a high pyramid you need a large base, and it seems that among the best young players in Europe, many are Russian:

http://eygc2018.org.ua/?page_id=484

[Uberdude]

A significant proportion of smaller tournaments are tournaments for children and/or young players. Admittedly, many children are in the 15k-20k range, but I don't think children tournaments should be discarded from the data. To build a high pyramid you need a large base, and it seems that among the best young players in Europe, many are Russian:

Indeed; in fact what about the players weaker than 20k? I've heard some criticism of the EGF rating system for not allowing ranks ratings (Edit per Herman) weaker than 20k (presumably because such player have large variances so the idea is the ratings won't be high quality) for excluding weaker players and making them feel like they are not proper players involved in the community. A 10 year old 27k could still find making 26k motivating, even if there are big error bars.

[Javaness2]

Indeed; in fact what about the players weaker than 20k? I've heard some criticism of the EGF rating system for not allowing ranks weaker than 20k (presumably because such player have large variances so the idea is the ratings won't be high quality) for excluding weaker players and making them feel like they are not proper players involved in the community. A 10 year old 27k could still find making 26k motivating, even if there are big error bars.

I still struggle with the argument that giving a 30k a rating of a 18k is better for the accuracy of the system. Maybe there should have been more beginner events post AlphaGo.