Starcraft AI blog

I feel that it may be discouraging to new players when bots are disabled for being too weak. For example, on SSCAIT, a bot will sometimes be disabled with a note saying "It plays worse than the built-in AI." On BASIL, a bot will sometimes be disabled if its win rate isn't high enough.

But new bots are almost always going to be weaker and have lower win rates. In addition, when weaker bots are disabled, the field gets tougher, which compounds the problem.

I think it feels discouraging and a little mean, like the community is saying, "We don't want to play with them."

I know it's not your decision. I've just been thinking about your post, and I know a lot of the community follows this space.

I have a suggestion: I think instead of disabling "weaker" bots, SSCAIT should disable "inactive" bots. Bots should automatically disable if they aren't updated for a long time, such as 180 days, regardless of strength. Updates could be as minimal as just re-uploading the same binary. This rule makes it the author's choice whether to play or not, not the community's choice whether to exclude someone.

I'm sorry if I sound critical of SSCAIT; I know that putting strong bots on stream is important to everyone and I don't think there's anything wrong with that. I also realize that you and your readers have much more time and energy invested in this community than I do.

In my experience there are some significant barriers to entry for building a new bot.

- It's difficult to get off the ground; I ran into a ton of issues trying to get the demo AI module to compile and run successfully.

- Development more or less has to be done in Windows, so there's a bit of a learning curve and adjustment for those of us who are used to developing in a UNIX environment.

- There's a high bar (or at least the appearance of one) that a bot needs to reach before it can be run on SSCAIT. A lot of bots are disabled for "playing worse than the default AI" or "just another rush bot."

Sorry but I don't want to watch 90 min long games because of bots got bugged, slows down whole game, or simply does not know how to finish opponent.
When I watch stream I want to see a bot that is significantly stronger than build in a. I., does not lag, crash and actually can play decent game. Even if its outdated by 5 years inactive is still great benchmark for all new bots and even top ones as long as its smooth and does something challenging to opponent.

@Jay, any reason that my comment didn't register? :)

Really, weird. I can now see that another of my posts hasn't registered to other topic, namely: `beating SAIDA`. This has happened in the past, too :(

Here is one more attempt at posting my response, although *very* late to the party and I doubt it that anyone would read it :D
---------------
1) To LetaBot, I don't think that the whole effect could be attributed to the availability of the SC2 API, although certainly the latter plays its role in attracting newcomer bot developers towards other games like SC2.
Attracting mind share is one area where we as a community should work harder even in terms of better PR of BWAPI, the Twitch stream, Nepeta's video casts, the Undermind podcast, the Discord server, etc. :D

2) To the anonymous bot developer: you have a good point there. And this is not a new observation BTW. Some background to the current status quo:
Having just one 24/7 running ladder until recently (the SSCAIT one) had made it difficult to suit the needs of everyone: new bot authors who want to be given their 15 minutes of fame on the stream; hardcore developers and teams who wish to win at all costs; Twitch viewers who would like to watch some good action and have fun while at it; trolls, etc. Those groups who sometimes have intersections between each other all have different priorities and preferences. We've striven to find a balance to suit everyone (save for the trolls :D ) but it's not easy with just one ladder and just one active Twitch stream. Then came BASIL which is more objective and tries to cover the play of even more deserving bots than SSCAIT could due to the video streaming and play speed limitations of the latter. But both of those ladders together still have limitations, of course. Hardware capabilities aren't infinite. You can only run so many games per day...
If ladder admins don't periodically prune weak / inactive bots, things would just slowly stall and BW AI improvements momentum would stop. Just imagine what would happen otherwise: which bot author would have the patience to wait for say a whole week until their bot meets an *interesting* opponent on the stream?
I (and everyone else) believe that each new bot (author) should be given a fair chance to prove its worth so it is highly unlikely that a new submission would get disabled right away as long as it doesn't crash too often, lag the games it plays in, and wins at least a small percentage of games against some mediocre opponents. That said, it's not OK for someone to submit a weak 4 pool bot once and expect it to still be enabled after say half a year of lack of updates? Who would that please except for the ego of the author perhaps?

tl;dr if you (and I am not directing this to just one person) are serious about the whole stuff, you should try to make some noticeable improvements and introduce some novelty in the AI scene at least once every couple of weeks, submit them (to SSCAIT) for everyone to notice that you have *REALLY* been trying to make a dent in the BW bot world. If you succeed at the above, I assure you that you will continue enjoying watching your bot play on Twitch (SSCAIT) and (hopefully) BASIL.
OTOH, some oldies but goldies have earned their place on the ladder even after years of inactivity by just doing something kinda unique and doing it well enough. Some examples: Tomas Vajda, Andrew Smith, Florian Richoux and many others. Why don't you try to be one of them? ;)

Nepeta's video casts definitely got a lot more attention than previous years. The typical game of the RR phase - not so much...