My Journey to 1Dan... Will I ever get there

[lowiq60]

I was under the impression that most people end up at around 1d

by definition, 1d is where most people don't get, which is why Nihon Ki-in chose that name for that level, similar to 1d black belt in Judo.

Well for the local go clubs that I have participated in, it seems like 1k - 1d is pretty close to a 50th percentile mark - roughly half of the people are above it, and half of the people are below it. The kyu players in these clubs also tend to have less playing experience compared to the dan level players, so it is possible that some of them will improve to dan level in the future. Granted those who make the effort to go to a club meeting are more serious about the game than your casual amateur, but I am a club participant myself so I take the game seriously too. I don't think it is unreasonable to think as an average amateur who is intense enough to study the game and go to club meetings to play, I should be able to get close to 1k to 1d level. Again I am perfectly happy being a 10k forever, which is probably my current skill level.

my rank or ability at 10kyu will not stop me from having fun playing opponents of similar level.

that's a very healthy attitude; recognising and accepting one's own limitations is the first step on the journey of 1000 miles to maturity (or enlightenment, whichever you prefer). i look forward in unjustified optimism to being able to take that step myself one day

If you want to develop this attitude, just think about alphago and alphago zero. You probably have no more natural ability in Go than Ke Jie or Park Junghwan, and they will most likely never be able to be as strong as Alphago Zero even though they work really hard. Then every human player will have a ceiling rank that they can't surpass. For me, that ceiling is 10k, which is lower than most people, but every human player has that ceiling.

[Bill Spight]

Well for the local go clubs that I have participated in, it seems like 1k - 1d is pretty close to a 50th percentile mark - roughly half of the people are above it, and half of the people are below it.level.

In the US in the 1970s, aside from Japanese and Korean clubs, the median rank of club players was mid-SDK, it seemed to me. I have also observed that now the median rank in local clubs is around shodan. I think that US ranks have remained fairly steady over time, so on the one hand I am happy that the level of play among Westerners has risen, but I also worry that the gap between new players and average clubs players is so great that beginners can become discouraged. Where is the new blood? Online, I suppose, but I think that local clubs are healthy for the game.

[djhbrown]

yup, https://senseis.xmp.net/?AGARatingHistogram shows that more people who go in for its official competitions are 1d than any other rank.

as for new blood draining away, the Hikaru rush may be past it's prime, and the alpha wonder has demonstrated clearly to anyone with half a brain that it's not worth the bother unless you're as savant an autist as she is, just as no-one bothers any more to try to fly by flapping their arms and is content to let a machine do the work, so soon enough Go will return to being an old man's game, as they (used to) call it in Japan before Hikaru.

[lowiq60]

yup, https://senseis.xmp.net/?AGARatingHistogram shows that more people who go in for its official competitions are 1d than any other rank.

as for new blood draining away, the Hikaru rush may be past it's prime, and the alpha wonder has demonstrated clearly to anyone with half a brain that it's not worth the bother unless you're as savant an autist as she is, just as no-one bothers any more to try to fly by flapping their arms and is content to let a machine do the work, so soon enough Go will return to being an old man's game, as they (used to) call it in Japan before Hikaru.

I think go is still fun to play on a human level, so the number of casual amateurs shouldn't decrease. I think it is hard to take the game seriously post-alphago though, and I don't know if there will be people who take it seriously enough to achieve professional status in the future. If a neural network with millions of games of self-training and monte-carlo tree search can become so good at the game, then the game might not be as elegant as it was previously perceived (philosophically speaking).

[dfan]

I think chess is more popular now than it was 20 years ago when computers surpassed the best human performance, and it's not obvious to me that Go should be any different.

[djhbrown]

the game might not be as elegant as it was previously perceived

Monte-Carlo certainly did rumple the elegance of Go, but by not using Monte's random rollouts, Alfie0 has cleaned and restored it. There is an elegance to Alfie0 that Alfie Master does not share, both in their fundamental natures and in their behaviours. For example, Alfie0 doesn't make any of those weird "new" moves that people made so much of a fuss about when they first showed up.

Alfie0 would be the first to admit that she doesn't know what she's doing, so has to rely on probability statistics to choose for her - but that such a mindless knucklehead can run rings around the 10,000-hour club should perhaps give us pause to reconsider whether H.Sapiens is actually as smart as some of them fondly like to think they are. "The most complex thing in the Universe" - pah! That's utter nonsense hubris - the brain is only one organ of the body (to be more precise, it's 3+3 + a couple of satellites), so it's not even the most complex thing in a single individual.

A visiting Martian would smile at just how daft southerners are, celebrating the winter solstice in the middle of summer - and even in the north, they still get the date wrong just because Constantine did (not to mention muddling up the reason for its existence by spicing it up with a magical fairy story based on Gilgamesh). The solstice is about the Sun, but this year we celebrate it on Moon Day.

It's not Go that's lost its lustre, it's us. Happy Christmas.

[lowiq60]

Over the past two days I played a bunch of games with LeelaZSlow on KGS under the username lowiq50. My rank of 11k? comes from one win over GnuGo 12k - so don't take it too seriously. I have had both wins and losses against this program. ( but more wins than losses, so dare I say that maybe I am still a tiny bit stronger than Leela Zero at the moment?) Based on its current rate of progress, I am not going to have a chance against it in a week!

LeelaZSlow is one of the KGS accounts that is testing the strength of the Leela Zero self-learning program.(where its neural networks learns everything with zero human input!) I believe it runs about 40-50 thousand playouts per move, which makes it the strongest LeelaZero account as far as I know. Currently it plays a very center oriented strategy.

Let me share some of my losses against this program (I will follow up with some wins). This program has learned a very center-oriented strategy. It's fairly strong at reading and local fights, and has a pretty good global strategic sense. It is also very good with cutting and connecting tesujis and small-scale capturing races.

Loss #1

https://lifein19x19.com/download/file.php?id=9141

Loss #2 ( I made an undo to a misclick, and typed a message in the game chat about it, but for reason the message didn't show. It is also fun to read the live comments that other people made as I was playing horribly. It seemed like they were enjoying me getting killed and laughing at my horrible moves... They even made an estimate that leela is 7-10k based my play, but I personally think this network is stronger than that)

https://lifein19x19.com/download/file.php?id=9142

[lowiq60]

Now here are 3 of my wins against LeelaZSlow (I think I beat it 6-7 times), so I'm currently having like a 70% to 80% win-rate over the latest network c99f1ae0 running LeelaZSlow's very powerful computer. Some of these wins will show up as losses on my account because I will either timeout or resign if I'm think I'm clearly going to win, as LeelaZSlow can be very stubborn about resigning.


These wins will highlight various weaknesses the program is having. The program hasn't learned ladders yet, but its center-oriented strategy sort of mitigates the damage since Leela's opponent is unlikely to get a working ladder that runs to the center. Sometimes it has trouble with large scale capturing races (I killed the program in a no-eye with lots of liberties vs one-eye with few liberties capturing race in win #2). Lastly, the program "gives up" too easily when it's behind and just flat-out refuses to play good endgame. In win #3, I had a lead going into the endgame, but Leela Zero knew it was behind, and just started throwing stones at me. Seriously my endgame is HORRIFIC, and against a 4d human player I could easily lose that advantage that I had in engame.

Win #1 (I resigned this one but I hope it's clear that I won)

https://lifein19x19.com/download/file.php?id=9143

Win #2 (I forced Leela Zero to resign)

https://lifein19x19.com/download/file.php?id=9144

Win #3 (I intentionally timed-out because I thought I clearly won - please let me know if I misjudged)

https://lifein19x19.com/download/file.php?id=9145