Examining what it took me to get to shodan

[Numsgil]

After 10 years (nearly to the day) I have finally managed to get promoted to 1 dan on KGS. This might be part of KGS's ongoing rank rebalancing, so we'll see if it sticks, but still, this is the first time I've gotten a legitimate dan rank on KGS (a few 1dan?s not withstanding). Getting stronger at go is one of the few long term personal projects I've followed through on so I thought I'd examine what it took using data!

Over that 10 years I've played approximately 4700 games. 4250 on KGS, 250 on Tygem, 150 on Fox and 50 or so in person. Between playing the games and doing tsumego, conservatively I've spent 5000 hours on go. The actual number might be closer to 7500 hours, but probably not as high as 10,000 hours.

I never really had a teacher, but I do own several dozen US language go books and another two dozen virtual go books, most of which I've even read. I went to a few tournaments and workshops in the bay area when I lived there but it was a relatively rare occurrence. Also, shout out to those from the Google Go club when I was playing there in 2015 and 2016.

Now for some graphs! The data is just pulled from KGS analytics so it does not contain data from Tygem, Fox or in-person games. But it still has data on 90% of the games I've played.

Here's how many games I played at each rank on KGS. I don't know why I spent so many games at 2k and relatively fewer on 1k.

games_per_rank.png (5.17 KiB) Viewed 16250 times

Here's my rank distribution across time. It was smooth sailing up to 3kyu, but rather rocky after that. You can also see a few gaps in my play while I was doing other things.

rank_vs_time.png (10.61 KiB) Viewed 16250 times

Here are the total number of games I'd played by a given date. I played over half my games in the first 4 years.

game_count_vs_time.png (10.93 KiB) Viewed 16250 times

Hope that was interesting!

[Kirby]

Congrats!

[TelegraphGo]

Congrats!

Are there any particular moments from your journey that stand out to you right now? I've forgotten a lot of the flashes of inspiration that I had as a kyu, and I really wish I wrote them down somewhere. All your future inspirations will be dan-level!

[SoDesuNe]

Nice one, congratulations!

[xela]

Congratulations, both for the level up and the cool graphs!

4700 games plus tsumego study in 5000 hours means you were playing mostly blitz? Did you spend much time reviewing your games, or did you improve mainly from playing lots?

[Knotwilg]

most of which I've ever read

I assume you mean "never".

Thanks for posting this. I'm not so surprised by the 2k plateau - I was on a similar one myself (>20 years ago )

In my case, rivalries with serious games and reviewing them to do better in the rivalry, made the difference.

xela observed you must have played mostly blitz. Have you been playing more slow games lately?

[Numsgil]

Are there any particular moments from your journey that stand out to you right now?

Around October 2018 I was sure I had this whole shodan nut cracked. I had by accident stumbled on to a play style that was winning me games at a rather high rate. Basically I'd play 3-4 stones and try to force every corner to have a simple, low, territorial exchange for both players. Play the most boring joseki imaginable, and try to force the board to alternate black-white-black groups all across the board, so no moyos could easily develop. Then I'd play a slightly better end game and win by a handful of points. By rank I'm not a very good fighter, but I have pretty good endgame technique, and one or two endgame tesujis that netted me an extra forcing move and a point or two is all it takes to win a game where each side only has 40 points or so total. It also made counting the score super easy since all territory was definite. These games were terminally boring but I was able to get to 4dan on Fox with this playstyle. However I couldn't quite manage the winrate to get to shodan on KGS. To get promoted you need a winrate around 66-70% and I was averaging more like 64%.

Anymore I think the secret was not that this playstyle is particularly good or bad, but it was forcing me to not get involved in do-or-die fighting, where I would try to do too much with a move and get counter-attacked hard and lose a handful of stones without much benefit. So I've tried to bring the calmness of that playstyle to my current play, so I'm not afraid to complete a table shape even though it neither attacks another group nor makes territory. I think that's at least part of what's allowed me to reach shodan.

For a second anecdote, this is less to do with my studying go and more an ancillary benefit. I joined DeepMind in October 2016, after AlphaGo had beaten Lee Sedol. Not for my go playing but because I've worked in the video game industry as an engineer and they have a team that builds custom games for AI agents to play and be tested by. But in May 2017 AlphaGo had a match against Ke Jie. Most of the AlphaGo team were in China, but it'd be difficult for them to fix problems remotely, so they had a skeleton crew in London in case a server needed kicking. Unfortunately the skeleton crew weren't very knowledgeable about go at all, so in the event they had to restart the agent they wanted someone on hand to ensure that the moves coming from it were sensible. I was one of the strongest go players in the company not in China, so I got to "help". In the end nothing went wrong so I had nothing to do but I got to watch game 2 of AlphaGo vs. Ke Jie live from London in the middle of the night with a few other DeepMind engineers while I desperately tried explaining why that game is insane. If someone hasn't seen that game I'd definitely recommend checking it out. As a thank you for working the graveyard shift DeepMind gave me a rather high end goban and set of stones.

4700 games plus tsumego study in 5000 hours means you were playing mostly blitz? Did you spend much time reviewing your games, or did you improve mainly from playing lots?

Most games were 25 minutes main time with 30 second byoyomi. You'd think that would make each game an hour, but in practice it's not uncommon to end a game with 15 minutes on both players clocks. Either someone messes up a joseki and resigns at move 50 or the game was never really complicated and neither side tried to make it so. 30 minutes/game average is rather conservative. 40 or 45 minutes/game might be more realistic. Likewise it feels like I've spent more time on tsumego than playing, maybe by 2:1, but I can lose hours to playing games on a weekend and I've never lost hours to doing tsumego, so I split the difference and said 50/50 time on games and tsumego as a guess. It's probably not too far off, though.

I usually review games I've won, and almost never review games I've lost. Which is not a good habit!
I'm blase about wins but losses can tilt me pretty hard and I've found it difficult to review them, even several days later. In the new year I've been trying to be better about that.

most of which I've ever read

I assume you mean "never".

Oops sorry I actually meant "even"! I've spent a lot of time on my various books. A few of the drier ones have been harder to get through but I've gone through most endgame and tesuji books at least once and any book with actual paragraphs I've read probably 3-4 times minimum.

xela observed you must have played mostly blitz. Have you been playing more slow games lately?

Actually the opposite. Before I'd play 25 minutes + 5x30 second byoyomi, but I picked up chess around May 2019 where I play 5 minutes main time + 5 seconds per move. Playing that fast sort of forced me not to spend too much time thinking through lines that don't go anywhere. That leaked in to my go style as well and I've been playing faster time controls on occasion. Like no main time plus 3x20 second byoyomi. Still not blitz but much faster than I could have managed before, certainly.

[xela]

Most games were 25 minutes main time with 30 second byoyomi. You'd think that would make each game an hour, but in practice it's not uncommon to end a game with 15 minutes on both players clocks. Either someone messes up a joseki and resigns at move 50 or the game was never really complicated and neither side tried to make it so. 30 minutes/game average is rather conservative. 40 or 45 minutes/game might be more realistic.

To me that's close to blitz! At those time settings, I struggle to finish a game within main time. Even finishing a game at move 50 with 15 minutes on your clock means that you were playing at 24 seconds a move, which isn't time for deep reading or counting.

[Kirby]

I joined DeepMind in October 2016, after AlphaGo had beaten Lee Sedol.

Oh, no - You joined the dark side!

Just kidding It's an interesting story, and actually sounds like it was an awesome opportunity. Very fascinating

[Bill Spight]

Most games were 25 minutes main time with 30 second byoyomi. You'd think that would make each game an hour, but in practice it's not uncommon to end a game with 15 minutes on both players clocks. Either someone messes up a joseki and resigns at move 50 or the game was never really complicated and neither side tried to make it so. 30 minutes/game average is rather conservative. 40 or 45 minutes/game might be more realistic.

To me that's close to blitz! At those time settings, I struggle to finish a game within main time. Even finishing a game at move 50 with 15 minutes on your clock means that you were playing at 24 seconds a move, which isn't time for deep reading or counting.

FWIW, in his book, Golden Opportunities, i.e., to win the game, Rin Kaiho (Lin Haifeng) says that amateur games typically last an hour or less. That was my experience learning go. I would usually play three games in an afternoon or evening. Lin goes on to recommend spending up to 10 minutes on one of these crucial plays. If you do that, that leaves about 10 sec. per play for the other moves.

[Numsgil]

To me that's close to blitz! At those time settings, I struggle to finish a game within main time.

Ah I find this very interesting. I was always under the impression I was a relatively slow player, both from my interactions with other players in live games and on servers. Where do you normally play? Over the board or on a specific server? The slowest KGS automatch time is 25+30, which it describes as a "medium" time control.

[xela]

Mostly on KGS. I'm fine with 25+5x30 as "medium speed", on the assumption that you're actually going to use most of that time! When we used to have an active local club here, people tended to play a little slower, maybe an hour to 90 minutes to finish a game, so you'd generally get two games in an evening. And then I've played a bunch of tournaments at 75 minutes per person plus byo-yomi (six games in a weekend used to be standard for Australian tournaments).

[EdLee]

I was always under the impression I was a relatively slow player

Even 45 min. + (5x 60s) is blitz for some.

[Yakago]

And personally I find 25 minutes to be quite a 'relaxed' pace for a casual game (not to say slow ^^)

To me, 10 minutes 30x/3 is normal pace online, perhaps slightly quick. 5 minutes /20 sec is fast. Only faster I would consider blitz

For a serious game I would say 25 minutes is fast. Still not quite blitz though

[jlt]

Lin goes on to recommend spending up to 10 minutes on one of these crucial plays. If you do that, that leaves about 10 sec. per play for the other moves.

It's not so easy to blitz through 50 moves, then suddenly read deeply for 10 minutes, then be back to blitz again. Some mental discipline is needed.