Difference between revisions of "SimpleBot/Version History"

From Robowiki
Jump to navigation Jump to search
m (Better expression)
m (0.024d && 0.024c.c)
Line 5: Line 5:
 
}}
 
}}
 
----
 
----
 +
* {{SimpleVersion|date=2017-08-27|version=0.024c.c}}
 +
** '''Based on 0.024c'''
 +
** REFACTOR EVERYTHING but break anything.
 +
** Add Acceleration & WallDistance dimension in movement (in a separate tree to make sure not breaking existing tree).
 +
* {{SimpleVersion|date=2017-08-27|version=0.024d (skipped)}}
 +
** Use kernel density in movement.
 +
** Remove most recent hit gf.
 +
** Try add straight wall distance in surfing dimensions.
 +
** Try to decay GuessFactors with gaussian distribution.
 +
** Performance decreased a lot.
 
* {{SimpleVersion|date=2017-08-26|version=0.024c}}
 
* {{SimpleVersion|date=2017-08-26|version=0.024c}}
 
** '''{{SimpleStats|link=http://literumble.appspot.com/BotDetails?game=roborumble&name=aaa.SimpleBot%200.024c|rumble=RoboRumble|scorelabel=APS|score=83.06|rank=36th|win=1093|loss=59|plrank=54th|glicko2=N/A|pwin=94.88|vote=2.81|anpp=87.3|score2label=Survival|score2=89.84}}'''
 
** '''{{SimpleStats|link=http://literumble.appspot.com/BotDetails?game=roborumble&name=aaa.SimpleBot%200.024c|rumble=RoboRumble|scorelabel=APS|score=83.06|rank=36th|win=1093|loss=59|plrank=54th|glicko2=N/A|pwin=94.88|vote=2.81|anpp=87.3|score2label=Survival|score2=89.84}}'''

Revision as of 12:01, 28 August 2017

SimpleBot Sub-pages:
SimpleBotVersion History

  • 0.024c.c – 2017-08-27
    • Based on 0.024c
    • REFACTOR EVERYTHING but break anything.
    • Add Acceleration & WallDistance dimension in movement (in a separate tree to make sure not breaking existing tree).
  • 0.024d (skipped) – 2017-08-27
    • Use kernel density in movement.
    • Remove most recent hit gf.
    • Try add straight wall distance in surfing dimensions.
    • Try to decay GuessFactors with gaussian distribution.
    • Performance decreased a lot.
  • 0.024c – 2017-08-26
    • RoboRumble ‒ APS: 83.06% (36th)
    • Add second-wave surfing.
    • Weight waves on bullet damage. Not yet tried weighting on distance or so.
    • Add stationary risk (risk if I hit the brake immediately).
  • 0.024b (skipped) – 2017-08-26
    • Remove flattener completely.
    • Merge branch knn to master.
  • 0.023h.knn.b – 2017-08-26
    • RoboRumble ‒ APS: 82.15% (46th)
    • Tune gun against RaikoMicro & FloodMini:
      • Use manhattan distance for knn
      • Use gaussian distribution kernel density
      • Use max overlap for selecting best firing angle
      • Use accel for lateral direction in gun.
      • Give up PreciseMea but in gun attribute.
      • Add accel & currentGf dimension.
      • Thanks Skilgannon for the great idea of using currentGf! Also thanks Skilgannon for the ideas of using manhattan distance & gaussian distribution kernel density.
  • 0.023h_knn – 2017-08-25
    • RoboRumble ‒ APS: 79.56% (63rd)
    • Based on 0.023h.
    • Try kNN (my Bucket PR KdTree implementation) in gun and movement.
  • 0.024 – 2017-08-24
    • RoboRumble ‒ APS: 71.14% (134th)
    • WaveSurfing: Add a flattener.
    • Gun: fire gfs from last 50 tick waves.
  • 0.023i (skipped) – 2017-08-24
    • Fire lower power at first shot.
    • Add another tier (which is lower than max) for killer shot.
  • 0.023h – 2017-08-24
    • RoboRumble ‒ APS: 70.1% (155th)
    • Fix a bug in 0.023g: now reset energy power when init round.
  • 0.023g (skipped) – 2017-08-24
    • Rework energy management.
  • 0.023f – 2017-08-23
    • RoboRumble ‒ APS: 68.04% (192nd)
    • Gun: try to match enemy bullet power.
  • 0.023e – 2017-08-22
    • RoboRumble ‒ APS: 67.91% (194th)
    • Fix a bug in 0.023d that caused WallSmoothing not working.
  • 0.023d – 2017-08-22
    • RoboRumble ‒ APS: 67.07% (212th)
    • Try some non-iterative way of WallSmoothing in both gun and movement. Also as a side effect, a trivial bug in the old simplest implementation of WallSmoothing is eliminated. And the total running time is halved comparing to the old implementation when fighting against RaikoMicro.
  • 0.023c (skipped) – 2017-08-22
    • Add a switch to debug graphics, and is off by default.
    • Try to truncate gun gf history but no noticeable increase or decrease in performance against RaikoMicro.
    • Fix a bug causing wave to be logged after enemy die.
  • 0.023b – 2017-08-21
    • RoboRumble ‒ APS: 67.96% (192nd)
    • Aim at predicted fire location instead of current location;
    • Fix a long living bug which caused fire power to mismatch aim power;
    • Add graphics for firing waves.
  • 0.023 – 2017-08-21
  • 0.022d – 2017-08-20
    • RoboRumble ‒ APS: 67.65% (199th)
    • Tune against some RamBots:
      • Fire a dummy wave when there is no wave to surf.
    • Precise prediction now ends after wave passes center instead of touches.
    • Limit attack angle
    • distancing on enemy distance instead of distance of wave source.
    • Add most recent gf in combination of random gf of last 20.
  • 0.022c (skipped) – 2017-08-20
    • Take botWidth into account when calculating wave danger;
    • Show selected GuessFactors in graphics;
    • Add Precise WaveTracking.
  • 0.022b – 2017-08-20
    • RoboRumble ‒ APS: 59.65% (376th)
    • Tune against RaikoMicro —
      • Simple Energy Management;
      • WaveSurfing select randomly in last 20 hits;
      • BulletHitBullet event handled.
    • Switch to orbiting fire position. Thanks Dsekercioglu!
  • 0.022 – 2017-08-19
    • RoboRumble ‒ APS: 59.64% (376th)
    • Surf next wave immediately when the current wave touches.
    • Tune danger function.
  • 0.021 – 2017-08-19
    • RoboRumble ‒ APS: 59.61% (377th)
    • Minor fix in movement (a long living bug).
    • Add debug graphics.
    • Take distance into account in danger calculation.
  • 0.02 – 2017-08-19
  • 0.011 – 2017-08-18
  • 0.01 – 2017-08-18