Maps that BWTA has trouble on

[heavy_metal]

I am working on a map analysis algorithm and I need some difficult maps to test against. I think that there were some maps that BWTA didn't do so well on. I don't mean maps that it outright crashes on, but those where it gets the Baselocations wrong, where the base is built on the wrong side of the mineral line, things like that. If anyone knows some of these maps off the top of your head that would be great.

[Kae]

I am working on a map analysis algorithm and I need some difficult maps to test against. I think that there were some maps that BWTA didn't do so well on. I don't mean maps that it outright crashes on, but those where it gets the Baselocations wrong, where the base is built on the wrong side of the mineral line, things like that. If anyone knows some of these maps off the top of your head that would be great.

I know AIUR has some problems to take B4 on Destination, certainely because of BWTA.

[krasi0]

I am working on a map analysis algorithm and I need some difficult maps to test against. I think that there were some maps that BWTA didn't do so well on. I don't mean maps that it outright crashes on, but those where it gets the Baselocations wrong, where the base is built on the wrong side of the mineral line, things like that. If anyone knows some of these maps off the top of your head that would be great.

Concentrate on maps where some chokepoints are blocked by neutral units / resources. Also on Andromeda for example BWTA considers the closest chokepoint to the base although it's very wide and one would rather defend on the next one which is shorter.

[heavy_metal]

So it should add a property 'blocked' and maybe a set of Unit* blocking the choke?

For the wide chokes a terrain analyzer should include a choke width property, so bots can decide which one to defend.

[krasi0]

So it should add a property 'blocked' and maybe a set of Unit* blocking the choke?
For the wide chokes a terrain analyzer should include a choke width property, so bots can decide which one to defend.

1) Yeah, that'd be great!

2) BWTA gives the width ATM and yeah, this way I choose which one to defend... It's just that chokepoint (especially on Andromeda) is useless considering the close proximity of the tighter one. My point is a human would not even consider the wide chokepoint so it looks more like a false positive...

BTW if you base your algorithm on BWTA and make, say, a branch out of it, that'd be great because ppl utilizing BWTA would easily switch to using yours.

[Kae]

I agree, Andromeda is tricky because of destructible reactors blocking the way near to your main but consider by BWTA as a choke. My defenders are often confused, wainting near to that point instead of defending the proper choke.

[krasi0]

I agree, Andromeda is tricky because of destructible reactors blocking the way near to your main but consider by BWTA as a choke. My defenders are often confused, wainting near to that point instead of defending the proper choke.

Yeah, I had to add special heuristics logic to exclude such spots although heavy_metal states that my bot still gets confused on other maps.

[Taranok]

Concentrate on maps where some chokepoints are blocked by neutral units / resources. Also on Andromeda for example BWTA considers the closest chokepoint to the base although it's very wide and one would rather defend on the next one which is shorter.

Electric Circuit, Medusa, Triathlon?
Arkanoid has a ton of neutral units.
EDIT:Hitchhiker has quite a few too.

[heavy_metal]

Electric Circuit, Medusa, Triathlon?
Arkanoid has a ton of neutral units.

Thanks for the tip. Arkanoid looks like a fun map... I wonder how many bots would get trapped by the neutral buildings.

[krasi0]

Electric Circuit, Medusa, Triathlon?
Arkanoid has a ton of neutral units.

Thanks for the tip. Arkanoid looks like a fun map... I wonder how many bots would get trapped by the neutral buildings.

prolly about 95% of all