Difference between revisions of "Dookious/DookiSaberRRGC"

From Robowiki
Jump to navigation Jump to search
(moving DookiSaberRRGC page)
 
m (fixing bad subpage link)
Line 28: Line 28:
 
** Weighting of main stat buffers is dependent on amount of data in them, up to a point. (Previously, they were weighted equally no matter what.)
 
** Weighting of main stat buffers is dependent on amount of data in them, up to a point. (Previously, they were weighted equally no matter what.)
 
* '''1.042''' - '''Rating: 1868'''
 
* '''1.042''' - '''Rating: 1868'''
** Two fast learning buffers instead of just one, and with a shared segmentation separate from the rest of [[/DookiSaber|DookiSaber]]'s segmentations.
+
** Two fast learning buffers instead of just one, and with a shared segmentation separate from the rest of [[Dookious/DookiSaber|DookiSaber]]'s segmentations.
 
** Various segmentation tweaks.
 
** Various segmentation tweaks.
 
** Tweaks to VirtualGuns logic:
 
** Tweaks to VirtualGuns logic:

Revision as of 17:34, 12 November 2007

The RRGunChallenge version of Dookious is called "DookiSaberRRGC". Up through version 1.0, the RRGC entries all mirrored an actual version of Dookious. However, with DookiSaberRRGC 1.042, I've begun testing gun changes with no corresponding version of Dookious; until I find a gun change worth making, of course. I'll list the updates to DookiSaberRRGC here.

Note: Dookious has not had an RRGC entry in a long time, so these scores do not represent the recent versions of Dookious.


  • 1.06 - Rating: 1884
    • Rollback to 1.04.
    • Bullet times instead of distances in Main Gun segmentations.
    • Weighting of two stat buffers dependent on weighted visit counts, instead of fixed weighting for each stat buffer.
    • Tweaks to bullet power management.
  • 1.048 - Pulled
    • Change to velocity-change time segment calculation in Main Gun.
  • 1.047 - Pulled
    • Major change to how velocity-change time segment is calculated in Main Gun.
  • 1.045 - Rating: 1873
    • Rollback to 1.043.
    • Lower VirtualGuns rating threshold for firing power 3 bullets.
    • Added VirtualGuns rating threshold for firing power 2.4 bullets.
  • 1.044 - Pulled
    • Added distance and accel segments to fast learning buffer of AntiSurfer Gun, a change made to Main Gun a while ago.
    • Lower visit threshold for maximum weight in stat buffers of Main Gun, and added consideration for visits in buffer weights to AntiSurfer Gun:
      • Maximum weights reached at 12/8 weighted visits in low/high segmented buffers of Main Gun. (Was 20 weighted visits in Main Gun. Firing waves = 1, non-firing = .2.)
      • Maximum weights reached at 3/2 weighted visits in low/high segmented buffers of AntiSurfer gun. (Firing waves = 1, non-firing = .02.)
    • High segmented buffers are rated 1.5 times as much per weighted visit; maximum weights are equal.
  • Note: Hopefully the last time I am unwise enough to try so many changes at once. =)
  • 1.043 - Rating: 1873
    • Weighting of main stat buffers is dependent on amount of data in them, up to a point. (Previously, they were weighted equally no matter what.)
  • 1.042 - Rating: 1868
    • Two fast learning buffers instead of just one, and with a shared segmentation separate from the rest of DookiSaber's segmentations.
    • Various segmentation tweaks.
    • Tweaks to VirtualGuns logic:
      • Don't refresh best gun when gun is close to cooled.
      • Above a rating of .26 (about 26% virtual hit rate), default to Main Gun.
    • Data saving is off, so as to better compare to AscendantRRGC and BeeRRGC.
  • 1.0 - Rating: 1869
    • Data saving inspired by WikiTargeting / SuperNodes concepts. (Credit to Vic.)
    • Fast learning buffer in Main Gun has three segments (LateralVelocity, distance, accel) instead of just one (LateralVelocity).
    • Non-firing waves in Main Gun weight restored to 1/5, after being 1/10 for a few versions.
  • 0.965 - Rating: 1861
    • Bug fixed in wave tracking. When target is below a certain velocity, Dookious uses the last direction instead of their current direction (calculated by that tick's data); the bug was causing it to reverse the orientation in some cases when it shouldn't, and not reverse it in some cases when it should. Change affects both movement and gun, but primarily the gun, since the movement velocity threshold is 0.1 while the gun's is 1.0.
    • Two small flaws (maybe not quite "bugs") fixed in gun:
      • "Start time" of gun waves was off by 1 tick (was 1 too soon). (Lined up visually in RobocodeGL, but were off by 1; fixed in movement in 0.945.)
      • If Dookious disabled himself by firing a shot, all gun waves after that were marked as "firing" instead of "non-firing", which affected the weight of the waves if/when they hit. (Obscure logic bug.)
    • Reworked how target direction is handled throughout Dookious's wave tracking, fixing obscure but long-standing bugs in this system.
    • Fixed the typo-ish bug introduced in 0.963's method which gets the "windowed" bin score. (It was setting both the low and high indexes the same, so it was only considering a bin along the edge, instead of all bins from one edge to the other.)
  • 0.96 - Rating: 1866
    • First RRGC entry.