|6||TheKid (GnuGo lvl1)||194826|
BadukEllington is an absolutely hilarious name.
Very interesting data. Thanks for sharing.
There are some things that I wonder about:
- Fuego had a huge spike in popularity in early 2016, but that quickly died down. I guess it might be related to the PR produced by Lee Sedol vs AlphaGo match? Was Fuego one of the top bots on OGS around that time?
- Interesting that the total number of bots games also saw a huge decline from early to mid 2016. I wonder why that happened, since 2015 seemed to have a consistently high amount of bot activity.
- Does the second plot show the number of games across all bots (not just the top 10)?
- Are there any bots not in the top-10 that are significantly more active only more recently? That is, are there any bots that would dominate parts of the first plot (in the more recent parts), but simply have not been around enough to make it to the all-time top 10?
Wild theory: maybe the huge influx of players joining in early 2016 (as generated by the AlphaGo publicity) made for such a large active player base that people could more easily find games against other humans, and hence less people played games against bots?
Is there wrong with those bots?
I went looking at the profile page for the first four of them and they seem to have all paused games
Paused games may be due to players abusing the bots:
Although some of them also appear to be obsolete and offline. Out of the top 4, I think only amybot-beginner is still active.
I guess Fuego was the only bot at the time. It looks like GnuGo and and TheKid went offline at the beginning of the same month (both lines drop to 0, GnuGo comes back 6 months later). Fuego is the only bot in the sample with games in the following 5 months.
@S_Alexander are this all relevant bots? The API lists 193 accounts as marked as bot. Most of them will be irrelevant by game count.
I remember there was an OGS Forum thread where, user macfergus - the bot administrator for BadukEllington - came to the forum to ask for help, and we started riffing on other possible funny bot names.
Say what you say, but randombot will stay as our favorite forever <3
March 2016 is definitely Lee Sedol vs AlphaGo peak.
Easiest explanation for steep decline in total bot games in mid 2016 is that bots were janky and broken.
Yes, second plot should show games from all bots (it’s a simple sum so games between bots are counted twice, unfortunately).
Top-10 pretty much contains all curves that were on top of everyone else at any moment. Amybot dominates last year. All hail amybot.
More like there’s something wrong with the people who abandon games like that. One of the reason meta left, wasn’t it?
I’m not sure what the question means. But I diligently went through 193 bots month by month. Top-10 is simply a nice round number.
I was very happy Random Bot made it. I played my very first game against it.
I’m astounded LeeLa isn’t in the top 10. Would have thought she’d be in the top 3 for sure… but then I never play bots.
I take this as a hint. I will go over the paused games and remove the forever paused games.
2020-01-15T11:48:00Z cleaned up most games. I left some recent games out.
Fuego was taken down because it had several flaws (ladders etc), and it was being seriously exploited.
Gnugo was also turned off. Combination of bugs and abundance of other options.
I’m surprised spectral isn’t in the top 10… Maybe if you combined all her different ranks together
By the way, I didn’t do any filtering, games include annulled ones.
Full popularity list:
Could we investigate correlation between rank and popularity?
This isn’t exactly the proper thread, but it’s probably the closest one to it.
I found these observations on GnuGo from 2005, on nachtrabe’s blog, which I’m gradually tearing apart like a a mountain lion and hiding around the landscape in small pieces.
Some observations about gnugo’s style:
Gnugo is very greedy when it comes to stones it can capture. It will often set up a capture rather than take moves that would net a greater profit elsewhere. Sometimes even ignoring the safety of its own group to go for a capture.
Gnugo likes to play monkey jumps and play hane-on-the-first-line [hanetsugi] endgame moves very early regardless of their actual value.
It completely ignores the proverbs about not letting the opponent get a star-capture [ponnuki, I guess].
It likes to tenuki, sometimes leaving a critical flaw open or ignoring a forcing move to do so. Sometimes the “forcing move” wasn’t, in other cases it would lose a group to it.
It seems to avoid ko-fights. Occasionally I would catch it increasing the value of the ko rather than going for an actual ko threat.
Gnugo is very very good at reading out ladders but it does so extremely slowly. So slowly that at one point, in an experiment game, I would make a move and then go do something else for ten minutes while it figured out what to do next. It would then proceed to recalculate the ladder on every move.
Gnugo almost never seems to play strategic forcing moves, such as peeps, or strategic sacrificial moves, such as probes.
It has trouble handling any situation where the opponent feints or otherwise plays in a “two-handed” manner.
Just some random observations from a handful of test games and two semi-serious games against gnugo.