The PvT project

[Taranok]

Updated v1.1! This version is much more solid in play! Added the second bunker. Doesn't seem to hurt it much and it helps in defense, even though it's in the main, as units get lured into the main. Made a few changes to attack loop to make it more consistent.

Stats: (tested on BWAIWarV and BWAIWarIV reps)
3-1 vs TaranokAI v1.71 - loss on Jade: this map seems a little bit harder to defend with? anyway the classic 10 minute attack somehow breaks through defenses
4-0 vs Unleaded v2.41 - steam rolled
2-2 vs ZBwaiP - loss on Icarus: attack loop freezes loss on Jade: units get stuck
3-1 vs Baja P - loss on Byzantium: loses to rush...

I'm not sure how I feel about these numbers. 12-4 is a 75% win rate. However, this rushing still seems to be the main problem.

I think Protoss AI's will have to modifiy unit compositions into one of these three:
zealot + dragoon with arbiters - zealot dragoon doesn't work well
zealot + archon - actually does surprisingly well, probably because Archons can tank the damage from tanks.
carrier + zealot - I don't think this works, but maybe tweaking with how ZBwaiP works could fix this.

Or they will just have to rush like Baja P.

[bajadulce]

[Taranok]

- Combined replays from BWAIWarIV and BWAIWarV. Also added my edited destination rep if anyone wants it.

Some notes: As far as the PvT ones, the rush distances are similar. The difference between maps is the distance between going diagonally across the map and vertically/horizontally. My least favorite maps for PvT are Python and Jade.

Btw, I do all my testing with the entropy.exe that was in the new BWAILauncher. I find that it results in better targeting and control of spell casters (no more irradiating overlords ), although the changes to storm might have made it a little bit overpowered . It's also uploaded here: .

I might add that your baja P AI could benefit from a little tweaking. Kind of sits with a lot of minerals around the 10 minute mark. But the rushing is pretty good .

[bajadulce]

[Taranok]

As far as I know, there wasn't Terran AIvsAI script that Myk wrote. There were the normal Myk 3,5 scripts, and I don't know whether those should be used or not. I also don't know if I should be using Unleaded 2.41 or Unleaded 3.31 (or whatever the latest one is), because I don't know if 2.41 was specifically intended for AIvsAI.

As for ai scripting potential, I think there are still quite a few unexplored options left. Seeing that simple stuff like Carrier rushes and the fact that Mech never was successfully attempted before kind of shows that scripting hasn't quite developed it's full potential. I wrote a ZvP script that just won against even some good AIs just by building drones and lings (no joke). And if BWScriptEmulator works out it could maybe open new options (if it is what I think it is).

Scripting for human play kind of dropped off because I think players found them useful mainly for training macro. You don't really need 50 different scripts to practice macro. Additionally, activity lessened after a bit.

EDIT: suggestion: your AI could benefit from more SCV's. 20 minutes into a game it still only has about 30 SCVs. It took quite a long time (almost 50 minutes) to defeat Blizzard P on Icarus. It just doesn't have any money to build anything. I would try using a command place_guard or get_oldpeons(20) to force more SCV's.

[bajadulce]

Ah for some reason I thought Mike had written a few Terran scripts?

Good point on the different ideas and always fun to explore new stuff (tho I have to say that the non-cheating avenues are rather limited). The FMP or crazy give_money scripts can do all sorts of off-the-wall stuff.

Ya. Activity dropped that's for sure, but there are still plenty of ppl that enjoy the old game including single player stuff like this.

@ script benefiting from more SCV's than 30 @ 20 minute mark
30 is not enough? I never tested anything past 15 minutes cause there weren't any replays that lasted that long when I wrote these. And If I hadn't killed the other AI by the 10 minute mark it was prob a wash anyways. Mainly I wrote the most simple script I could with a rather static opening build order that tried to burn up as much resources as it could in the first 5 minutes. After that, generic attack loops and called it done. I was only interested in the first 5-10 minutes and squeezing the most out of what was harvested. No doubt those scripts of mine are dinosaurs.

I might tinker w/ some a bit here tho. Need a challenge. What AI is the strongest PvT on those reps right now? And those reps ONLY. We should have a challenge the 3 of us to come up w/ a terran AI that beats the strongest P AI on all of those maps. Has to win every single map vs a single selectively chosen P.

[Taranok]

Maybe Myk did? Hopefully he'll tell us.

Myk's carrier rush and ZBwaiP would technically get the same score versus current Terran AI's (I'm pretty sure they 2-0 every one on the WarV maps). Myk's carrier script is slightly more gimmicky in that it relies solely on carriers. Once you can deal with carriers, you've beaten it. So if you want a better test I would go with ZBwaiP, which adds zealots (actually quite complicates things sometimes).
Of the Protoss AI's that don't go for carrier rushes, TaranokAI v1.71 is the strongest. Unleaded 2.41 does slightly worse according to scores. However, both carrier scripts are stronger than these. (Although, with some tweaking, TaranokAI could pose more of a threat. need to fix mineral-gas imbalance and expanding issues)

If you pick ZBwaiP I already have a head start My mech v1.1b I think already has a 3-1 score, and if not, 4-0.
[spoiler]Maybe we can all try building a PvT script vs the Mech script (that wins on all maps and doesn't dt rush)?[/spoiler]I will upload v1.1b soon, after some more testing. The main change is the fix to the attack loop with the goliaths added. It used to stop because it used wait_force for vultures when the loop doesn't tell the AI to build any

[krasi0]

"...Maybe we can all try building a PvT script vs the Mech script (that wins on all maps and doesn't dt rush)?"
For examples you could have a look at how Skynet plays. It has a few BOs

[bajadulce]

A Mech script that works! Cool! can't wait to see it in action.

@ Terran script
Last night I slapped together a super-duper chessy T script (never built a single factory ala CounterZerg tier 1 simplicity) that slammed every single no$ P script on that current BWAILauncher on all 4 of the replays (no game lasted more than 10 mins and if the 1st rush didn't wipe em out, the 2nd cleanup wave did). So I looked at about a dozen replays from the old BWAIWar replay pak (need to reupload here soon) vs the stronger ones and it also won every one I looked at.

[Taranok]

You quite like your rush scripts, don't you As if defending existing 2rax rushes wasn't hard enough. Probably medic marine firebat rush with more barracks, right?

Heh, Mech v1.1b seems a little bit disapppointing (at least to me) in performance.
TaranokAI: 3-1
ZBwaiP: 2-2

ZBwaiP tends to win because it is able to kill the third base on Python, where the 3rd is super open, and Byzantium, where the AI expands toward the Protoss (bad idea versus carriers). Plus, on Python it gets a weird supply block at 110 supply.

If you want a fun game I would suggest watching TaranokAI vs Mech on Icarus.