SSCAIT 2020 halfway point
The annual SSCAIT is past the halfway point of the round robin phase, and it’s time to take stock. The numbers keep changing, but here’s a snapshot.
Stardust has slowly climbed to #1 after its weak start, with only 4 losses after 50 games, as compared to #2 Monster with 7 losses after 67 games. Stardust has strong chances to hold its #1 position, though it has played fewer games. Stardust’s worst upset was against #29 ICEbot, while Monster’s was against #47 Junkbot. #5 PurpleWave is unexpectedly low, below #4 BananaBrain; from games I’ve seen, I suspect it did not get its usual thorough preparation, or perhaps the prep was concentrated on top opponents so that it can succeed in the elimination phase. #6 Iron is doing better than I expected, and is ahead of #7 Hao Pan, though they are ranked close and the edge may not stick. #9 Xiao Yi is also higher than I expected.
#14 Skynet by Andrew Smith is the only classic unupdated bot to hang on in the top 16. Other classics #17 UAlbertaBot by Dave Churchill and #18 XIMP by Tomas Vajda are just outside, and there is a gap with #16 Proxy so they may remain outside at the end of the round robin. #19 McRaveZ I had hoped to do better; its muta micro is good but its muta decision making (which target to seek, when to attack and when to run away) is not nearly as good as Monster’s. #20 Microwave has been slowly upping its win rate and has an outside chance of making it into the top 16 by the end; I imagine its learning is figuring out how to compensate for the bugs in this version.
Steamhammer is at #13 at the moment after a few losses, but I’m still forecasting that its most likely finish is #9 or #10. It has played more of its tough games than its easy games.
Some bots get special icons on the unofficial crosstable by Lines Prower. It’s a cute touch, though for me it makes the table harder to read. The funniest is Krasi0P’s linux penguin for 2 wins and Windows logo for 2 losses. I don’t understand McRaveZ’s icons. A salt shaker for losses, OK, but a secret agent for wins? I may be missing some background. PurpleWave gets a purple heart for wins. Maybe Lines Prower doesn’t know what a purple heart means to Americans?
Trackbacks
No Trackbacks
Comments
Johan de Jong on :
Jay Scott on :
Dan on :
Your latter suspicion is correct. For each tournament I prepare for the best chance to win the whole thing. For round-robin win-percentage formats, that usually involves incremental changes, broad testing, and bugfixes. For elimination formats, it means ensuring you beat (or have a reasonably good chance to beat) the top bots. And like it's poker, when behind I play to my outs.
The pre-tournament PurpleWave could not beat Stardust. The cause was straightforward: It could not win fights, even with significant edges, against Stardust's next-level micro. PurpleWave couldn't even convert 30+ midgame supply leads into wins. Stardust runs a very conservative PvP opener that consistently puts it at minor but manageable disadvantage against every other opening; it accepts this disadvantage in order to ensure it has the chance to trade favorably against you later. So there wasn't much juice left to squeeze on any front except being able to trade better in fair fights. And because I didn't expect anyone else to go in to the tournament with winning odds against Stardust, most options for a chip required that to succeed. PurpleWave is also not winning against Monster, but the road to winning that matchup was similarly long and I preferred the odds of a spoiler arising against Monster than a spoiler against Stardust.
So the preparation for SSCAIT involved a total overhaul of PurpleWave's micro. PurpleWave's micro had been neglected for years; outside of threat-aware pathfinding and shuttle handling, it's basically the same tech I ran in AIST 2018. Having to prioritize other things (the 4.4 tournament module client timing bug took up most of my development time on the year). Most of the micro code by volume was thrown out and replaced in order to have a chance of beating Stardust in an elimination series.
As you've remarked upon, new features expand capabilities but in the short term tend to reduce winrates due to lack of polish. This was especially true of the development cycle for this SSCAIT; unlike my usual preparation which has multiple rounds of testing and escalating code freezes, for this I didn't have roughly-at-parity micro until two days before the deadline, and the time afterwards was a rush to convert Stardust into a near-even matchup, which by the deadline I failed to do.
This process meant going into the deadline with known, and new, weaknesses. Four of the twelve losses are against Wuli and Legacy, caused by a novel-and-known-but-not-prioritized inability for goons to perfectly trade against slow zealots (I didn't have time to replace the logic for withholding fire to open wider gaps against slow short-ranged targets). Of the two losses against Hao, one was caused by defense squad caution tuned against Stardust being bad against a Terran contain; the other was a game where the bot wasn't running at all, for unknown reasons, and I haven't inquired about it because I don't think the small seeding implications are worth requesting intervention.
Combine the known unpolished edges to the new micro, with PurpleWave having played almost all its top-end matchups first, and the results are about what I was expecting going into it. I'm actually pretty happy with how the new micro has held up in matchups where it received no testing at all.
The curious emoji are mostly references to conversation on Discord. You might enjoy even just lurking to stay abreast of things; a lot of the questions you pose have previously been discussed or answered there.
Jay Scott on :
Every time I go to sign up for Discord, they ask me my age and lie about why, and it sticks in my craw.
MicroDK on :
Microwave did play many of the top bots in the fist half of the tournament, so the next half will be a little easier. But it needs more than two games to learn than 4pool/5pool will loose (on crash) and mutas are weaker. Many of its matchups that it normally wins 2-0 (a microwave) have already been converted to 0-2 (a explosion): vs Arrakhammer, Yuanheng Zhu and KaonBot. So I don't think the learning will compensate in this short tournament.
Jay Scott on :
Does Microwave learn anything on a crash? I thought the data would not be written then. Yet another reason to never crash. :-/
MicroDK on :