Page 1 of 1

NullBot starting position sensitivity

Posted: 26 Jan 2013, 18:18
by PkK
I've been playing a few matches against AIs on the 5-player map that comes with 3.1.

I played at the player at the bottom of the map, and used the default AI for the middle left and middle right players. I used NullBot and SemperFi for the upper two players. The map is rather symmetric. I'm not a very good player, but I play warzone once in a while, and even won in multiplayer once in a while, and usually win against the AI.

Now here's what happens in the matches on the 5-player map: When I have NullBot as the upper right player, it manages to hold its ground, and does battle other players once in a while but that's it. I've played three matches in this cofiguration and always did win easily.
When I have NullBot as the upper left player, it manages to defeat the other AIs, and becomes very strong, giving a real challenge. I've played three matches in this configuration, and while I lasted longer each time, I've lost them all.

Why does NullBot perform so differently depending on starting position on this quite symmetric map?

Philipp

Re: NullBot starting position sensitivity

Posted: 26 Jan 2013, 18:41
by NoQ
I wouldn't call this map symmetric. It is oriented counter-clockwise; inner ramp may be preferable for attacking, at least for AIs, while outer ramp will most likely be rarely used by AIs. That's why, whatever player is chosen by nullbot@0 to die first, he will most likely move his tanks near nexus@4; and nullbot@1 will meet semperfi@0 most of the time instead.

Do i make sense?

Re: NullBot starting position sensitivity

Posted: 26 Jan 2013, 19:44
by PkK
NoQ wrote:I wouldn't call this map symmetric. It is oriented counter-clockwise; inner ramp may be preferable for attacking, at least for AIs, while outer ramp will most likely be rarely used by AIs. That's why, whatever player is chosen by nullbot@0 to die first, he will most likely move his tanks near nexus@4; and nullbot@1 will meet semperfi@0 most of the time instead.

Do i make sense?
On one hand yes, on the other hand I doubt that meeting a SemperFI vs. meeting a Nexus first will mkae such a huge difference for NullBot. If I find time I'll see if I can reproduce the issue with only the NullBot and one other AI type.

Philipp

Re: NullBot starting position sensitivity

Posted: 26 Jan 2013, 19:45
by NoQ
That's not the only difference. In one of the cases he meets your neighbor.

Re: NullBot starting position sensitivity

Posted: 26 Jan 2013, 22:22
by PkK
NoQ wrote:That's not the only difference. In one of the cases he meets your neighbor.
Is there a way to have an AI-only skirmish, to test if that's what makes the difference?

Philipp

Re: NullBot starting position sensitivity

Posted: 26 Jan 2013, 23:34
by PkK
From what I saw during the games, I got the impression, that NullBot, when starting in the upper right corner is likely to get some groups of tanks stuck somewhere instead of using them efficiently.

Philipp

Re: NullBot starting position sensitivity

Posted: 27 Jan 2013, 06:37
by NoQ
Quite possible too (:
Is there a way to have an AI-only skirmish, to test if that's what makes the difference?
No, unless you modify the map.