For anyone attempting to rate the AI scripts

[Timid_Turtler]

Simply play a game, then watch your replay.


It should be in your recent replay list.

This way you can monitor every aspect of the AI

[sLayed]

Don't you think an AI should be judged by how well it emulates a player and interacts with the one playing the AI?  I suppose there are 2 aspects you could look at this, perhaps 2 ratings are justified. One from a gamers perspective and one from a programmer+gamers perspective. EG the first while you're playing the AI and the second watching the replay and grading it as a whole.

[Timid_Turtler]

You could still monitor that by watching a replay, i can't think of why you wouldn't be able to do that.

[chetjan]

ya...

[1337]

I run 4-player AI only games as observer, using SC2Launcher 2.0. I restart until I get at least one of each race (with one race duplicated) and then watch them duke it out.


I compare their resource income and harvester rate...
and their unit counts,
and number of expansions,
tech level,
and attack strategy
at every stage of the game.


That is the best way to test ai




Currently I am about to run a test game of AMAI v0.01 vs IAI (Protoss) vs my own custom Zerg AI

[briantheman]

I'm having trouble watching replays that I have played by myself against AI 5.5++.
I can watch other peoples replays but I can't watch mine.


When I try to load it using lazylaunch or using battle.net it says unable to open map.


Any suggestions? I know about replacing the s2liberty.mod thingy with the default.

[Rain[sun]]

I think AI rating should be entirely relative to our ressources; this is to say, compare the StarCrack and whatever other scripts to eachother, not to yourself. I know the ultimate goal is for a realistic player vs player but who are we kidding, it won't happen overnight.

Consistently using the feedback on what the AI is missing, as well as comparing its stats to the previous versions, will help you get an idea on how the progress is goin!

[turdburgler]

I think AI rating should be entirely relative to our ressources; this is to say, compare the StarCrack and whatever other scripts to eachother, not to yourself. I know the ultimate goal is for a realistic player vs player but who are we kidding, it won't happen overnight.

Consistently using the feedback on what the AI is missing, as well as comparing its stats to the previous versions, will help you get an idea on how the progress is goin!

Exactly, this is why I insist on regular releases, I get a good idea of what people think needs work.