archive by month
Skip to content

SSCAIT tournament soon

I’ve just uploaded Steamhammer 3.3.5, which will be the SSCAIT tournament version unless it hits a last-minute bug. If you dare to rush through your opponent prep, now’s the time! Expect the change list after the deadline. Optically, this version fixes all the most visible bugs introduced in and since the AIIDE version; the games look cleaner, overlords live longer, bizarre expansion behavior does not happen. Results are only slightly improved, though, in part because of the learning hides bugs issue. I expected better.

Starting on 19 December, there’s been a rush of updates. In fact, every bot updated after 27 November was updated (or re-updated) on 19 December or later, so there’s a gap in the dates.

There is not much to predict about the tournament. I think everyone can foresee that the top finishers of the round robin phase will include Stardust, Krasi0 (if it competes as terran this year), Monster, and PurpleWave, and likely BananaBrain which has been doing well. Halo by Hao Pan is significantly weaker, and there is a gap below Hao Pan and adias (aka SAIDA) of nearly 100 elo before the remaining strong bots. Steamhammer is likely to finish near the middle of the top 16, and then survive not very long in the elimination phase, as in past years.

Trackbacks

No Trackbacks

Comments

Dilyan on :

Oi oi mate! SH vs Monsters just hour ago, still in opening book SH did not transfer drones from gaining gaz to minerals... Leading to 1000 gas and 60 minerals or so... Gl

Jay Scott on :

Ah, but you didn’t see the bug I fixed, which was worse. :-)

MarcoDBAA on :

Bereaver was updated!

Not completely unexpected anymore, because Monster is 99% (or is it completely clear now?) Bereaver zerg, but still nice. :)

The bot did play its first game just now (I did not watch it yet):
https://data.basil-ladder.net/bots/Bereaver/Bereaver%20vs%20Proxy%20Jade%20CTR_61C65507.rep

For the race balance in the tournament, it would still be better, if Monster played there however, because it is the best zerg.

Jay Scott on :

I think the “update” to Bereaver was disabling it for the tournament. Am I right?

MarcoDBAA on :

Yes, I saw now, that it is disabled @SSCAIT already...

If (probably not) there are any changes, they are surely small and the main reason for the "update" was to disable it for sure.
Did not play different vs Proxy.

Still hope for a surprise entry in general. ;)

Jay Scott on :

I see Krasi0 is disabled. It looks like Krasi0P will compete this year, like last year. That’s too bad, after its nice play against Stardust.

MarcoDBAA on :

Yes, don´t know why this happens every year.
krasi0 could have won each year, including this one.

And a tournament win isn´t that "real" for everyone else, if the strongest opponents don´t compete.

Dilyan on :

Can't tell if krasi0 is strongest nor it will win tour because of just now updated versions of stardust, monster and likely very soon PW. Also updated bananaBrain was beating him.

MarcoDBAA on :

I did not want to imply, that krasi0 would definitely have won the tournament, but the bot is for sure one of the strongest opponents for all the other bots, with only one (vs Bananabrain) clear losing matchup over the last 14 days (BASIL). Updated versions of other bots could have changed the balance of power, but this is not certain.

P.S: Sadly no new (or bots, that were not updated for a long time) entries.

MarcoDBAA on :

Not sure, if I should take responsibility (see http://satirist.org/ai/starcraft/blog/archives/1012-Steamhammer-on-BWAPI-4.4.0.html#comments) for these 3 possible games, which may run like this in the possible final or not? :P xD Of course Monsters author probably came up with these ideas himself already. Still cool (but...) to see it happen however.

https://data.basil-ladder.net/bots/Stardust/Stardust%20vs%20Monster%20Icarus%20CTR_8076B2CD.rep
https://data.basil-ladder.net/bots/Stardust/Stardust%20vs%20Monster%20Destination%20CTR_83524167.rep
https://data.basil-ladder.net/bots/Stardust/Stardust%20vs%20Monster%20Jade%20CTR_83D59DC1.rep

Jay Scott on :

Monster is exploiting specific weaknesses of Stardust. With that game plan, zerg loses if Stardust takes the opportunity to expand as much as it should. It’s kind of discouraging that even a specialized exploit takes so much skill, and even then doesn’t always work....

MarcoDBAA on :

Yes, my idea was to try a very similar strategy, but still take the natural and defend with sunkens there. Further expansions can be taken with Overlords again.

Author might have tried it like that, but the existing strategy was just more effective. And Monster has no learning, to have both strats active...

I believe in using Ensnare vs the dragoon army. This will stop them microing as effectively (rate of fire is reduced too) and retreats will end in disaster (regrouping with reinforcements will also be delayed). Simplicity won games vs the Locutus clone ZNZZ because (in part) of this. You may just use it, when protoss is deep in sunken territory, letting your sunkens produce some extra hits, if Protoss wants to disengage. Then annihilate the fleeing army. ;) Also: Zealots can be outmicroed and protoss air would not be a counter. Ensnare is easier to cast with bot APM ( and reactions), cannot be dodged and negates bot micro somewhat. Should be good? Think, that it is a mistake, that only Simplicity (no real top bot) tries this.

Jay Scott on :

Steamhammer uses ensnare, but it has no idea of when to cast it. How do you identify the key moments to use a limited resource? Simplicity relies on simple heuristics and misses many opportunities.

MarcoDBAA on :

Think it is cost effective for Simplicity, from the games I watched. But might be false. Easy to remember the working instances more of course...

Ok...

You could run the combat simulator for example, and if you are in a clear winning position, but your army (each unit) is mostly not in shooting range (you need to catch them).

Or if you want to retreat, and stop them from catching you, and they are in shooting range. And if you have enough slower troops (mutalisks should disengage from ground units without help) there, that are worth saving.

Or if there is an intense fight (many enemies and own troops are in attacking range, or the damage output, if a total value is known), and therefore you want to have the micro and rate of fire advantage in a big fight.

In all cases, there needs to be a place, where you hit enough enemies at once of course. Else you may use broodling maybe, if the queen energy is maxed (it should be more willing to use a spell, because you really lose potential energy, if you are maxed), or against High templars.

If you have lurkers there you may use it on an observer, then snipe him (priority target).

I can code a bit, but did not manage to set AIs up really locally, and creating an AI is surely difficult anyway (looked into it a bit), but combat simulator and looping units vs enemy units for "in attack range" should help here?

Jay Scott on :

Something along those lines, but to catch all the good uses it takes yet more. Ensnare wraiths but don’t engage them, only slow them down so you can escape safely. Ensnare marines here, attack with mutas there when the marines can no longer reach them. And to get the most out of it, you have to plan ahead so the queen and other units coordinate.

Starting with Simplicity’s simple heuristics sounds like a good plan!

Jay Scott on :

Steamhammer represents the effects of ensnare in its combat sim. Once an ensnare is down, it can make an informed decision about whether to engage. It does not use combat sim to predict whether ensnare will be useful. The idea is worth trying, but it might be hard to get it working well because the sim results are unstable.

Jay Scott on :

These are the kinds of ideas I will try out first with dark swarm. Swarm is powerful but hard to use well, and improved swarm skills would make a huge difference in late game.

Lines Prower on :

Well, it doesn't cost you much to use skills you already have, so it was worth trying. I'm sure Stardust can be beaten by something conceptually much more simple, like a pure speedling build: continuously ambush Stardust's troops as they move out (while probably backstabbing as well). Stardust is too eager to attack, so this should work. But Monster doesn't have the skills for this yet...

Jay Scott on :

For those who may not remember, Lines Prower was the name used by the author of Bereaver.

Add Comment

E-Mail addresses will not be displayed and will only be used for E-Mail notifications.

To prevent automated Bots from commentspamming, please enter the string you see in the image below in the appropriate input box. Your comment will only be submitted if the strings match. Please ensure that your browser supports and accepts cookies, or your comment cannot be verified correctly.
CAPTCHA

Form options

Submitted comments will be subject to moderation before being displayed.