Difference between revisions of "Talk:Wintermute/Version History"

From Robowiki
Jump to navigation Jump to search
(Revert of 0.5?)
(revert answer)
 
Line 10: Line 10:
  
 
Hmm, why does 0.6 revert the change of 0.5? The 0.5 change appears to only improve things, particularly in PL. --[[User:Rednaxela|Rednaxela]] 00:01, 11 June 2009 (UTC)
 
Hmm, why does 0.6 revert the change of 0.5? The 0.5 change appears to only improve things, particularly in PL. --[[User:Rednaxela|Rednaxela]] 00:01, 11 June 2009 (UTC)
 +
 +
If you look at the Common APS you'll see there's actually no difference between them, and I'd already reverted my codebase to 0.4 a while ago. I might release a 0.7 with the inverse weighting enabled again, but it seems without a virtual gun it doesn't really help. And as far as PL goes, this bot doesn't even have a flattener yet, so I'm not that concerned about it. --[[User:Skilgannon|Skilgannon]] 16:20, 11 June 2009 (UTC)

Latest revision as of 17:20, 11 June 2009

Very strange, applying EXACTLY the same inverse-distance change in DrussGT's gun gained me around 0.5APS, whereas in Wintermute it actually lost me 0.1. I'm starting to wonder if that was the thing that actually gained me my score in DrussGT... --Skilgannon 08:37, 23 May 2009 (UTC)

Well, are you weighting your DC dimensions exactly the same? For example, if the average distance values are larger in Wintermute than DrussGT, it could completely change the outcome of that change. --Rednaxela 13:57, 23 May 2009 (UTC)

Doing the diff with DrussGT 1.3.7 & 1.3.8 and Wintermute 0.4 & 0.5 told me that the weight is the same, although the PM gun in DrussGT change from latVel/advVel to velocity/deltaHeading so that may make DrussGT more points. » Nat | Talk » 14:16, 23 May 2009 (UTC)

The only difference between DrussGT's gun and Wintermute's is that Wintermute is missing the non-precise stuff, like the VCS antisurfer gun and the pattern matcher (although that would be easy to make 'precise'). I think I'm going to release a version of DrussGT with just the inverse weighting removed, and see what happens. And that switch with the PM gun was actually a switch back after I had just tested latVel/advVel. --Skilgannon 14:22, 23 May 2009 (UTC)


Hmm, why does 0.6 revert the change of 0.5? The 0.5 change appears to only improve things, particularly in PL. --Rednaxela 00:01, 11 June 2009 (UTC)

If you look at the Common APS you'll see there's actually no difference between them, and I'd already reverted my codebase to 0.4 a while ago. I might release a 0.7 with the inverse weighting enabled again, but it seems without a virtual gun it doesn't really help. And as far as PL goes, this bot doesn't even have a flattener yet, so I'm not that concerned about it. --Skilgannon 16:20, 11 June 2009 (UTC)