Difference between revisions of "Gaff/VersionHistory"

From Robowiki
Jump to navigation Jump to search
m (migrated page from old wiki)
 
(1.50 notes)
 
(7 intermediate revisions by the same user not shown)
Line 1: Line 1:
* 1.28: Minor tweaks to cornering and movement selection
+
;1.50 - Back to your corner!
** Cornering no longer selected when Gaff is already cornered.
+
* After 1.44's poor performance, a more drastic revision was in order.
** Damage differential now reset every round
+
* Gaff's new movement is a return to his original roots: cornering is back.
 +
** This version uses [[Holden]]'s wave surfing (far superior to the old bullet dodging) and a simpler cornering plan -- not as crafty but Gaff takes far less hits than before.
 +
** As a fallback, Gaff reverts to a standard orbital movement but with very tight distancing for that up-close, personal feeling.
 +
* Some targeting tweaks to boost [[TCRM]] performance above 88.
 +
* Plus bugfixes and a bit of optimization.
  
* 1.26:  Tweaked movement to improve performance against worst [[ProblemBots]] and a selection of top bots.  No targeting changes.
 
** Cornering movement now used only when Gaff has a significant energy advantage
 
** Random and dodging orbital styles selected based on current damage differential; now switches with same round if needed.
 
** Bullet predictor decay rate decreased dramatically for faster adaptation.
 
** Distancing adjusted to make orbital modes similar and decrease advantage of segmenting on distance (makes a big difference).
 
** Fixed some bugs in the random mode which resulted in strange moves.
 
** Ranked 106 in OneOnOne (1805.94pts, 2059 battles, 20-Aug-2008)
 
  
* 1.24: Mostly movement updates
+
'''1.44''': Surfer killer?
** Smoothed out dodging orbit movement mode
+
* {{RumbleStatsDefault|link=http://darkcanuck.net/rumble/RatingsDetails?game=roborumble&name=darkcanuck.Gaff+1.44|rumble=RoboRumble|scorelabel=APS|score=68.06|rank=108th|win=617|loss=126|plrank=111th|glicko2=1845.6|score2label=Survival|score2=76.47}}
** Added a random orbit (non-dodging) movement mode
+
* Updated targeting -- same as 1.43g which topped the heap in the [[Anti-Surfer_Challenge/Results | AS Challenge]]
** Movement selection now settles on new random mode if other two modes lose too often
+
** Now uses a slightly different input data set for NNs (10 variables)
** Increased bullet predictor decay rate
+
** Features approximately doubled in number
** Minor tweaks to targeting NN to adapt faster and handle lower bullet powers better (but no challenge score improvement)
+
* Many, many small framework tweaks carried over from [[Pris]] and [[Holden]] development
** Ranked 97 in OneOnOne (1816.7pts, 2061 battles, 17-Aug-2008)
+
** Now adjusts aim for tank movement
 +
** Hopefully fixes radar lock bug
  
* 1.20: Further targeting improvements, many movement bug fixes
+
'''1.42''': Revamped movement
** Targeting network no longer uses hidden layers.
+
* Restored the orbital dodging mode which was keeping scores up vs simple targeters.
** Network is retrained at a slower rate but more often.
+
* Dropped the older random orbit movement in favour of the newer Raiko-style one.
** Fixed many, many bugs in the orbital movement and bullet dodging classes.
+
* Tweaked some distances and fixed a minor distancing bug.
** Removed the "oddball" movement style
+
* Expanded my test bot set and this one worked out better than 1.34.
** Changed the movement style selection method
+
* Ranked 91 in [[RoboRumble]]: 69.9 APS, 1546 ELO, 1869 Glicko2, 2841 battles (14-Aug-2009)
** Ranked 117 in OneOnOne (1800.74pts, 2122 battles, 6-Aug-2008)
 
  
* 1.10:  Continued improvements to targeting and some new movement
+
'''1.40''': Improved movement
** Targeting method now trains multiple samples per firing wave
+
* Created a pretty slick, low cost orbital movement to replace the ugly iterative one in previous versions
** Fixed a major bug in the bullet dodging code
+
* Smoothed out dynamic distancing
** Credits for both of the above go to [[Rednaxela]]
+
* Replaced the buggy orbital dodge movement mode with another random movement inspired by Raiko's algorithm.
** Added two new movement styles (orbital and "oddball") for situations where cornering isn't working out.
+
* Ranked 126 in [[RoboRumble]] -- 65.10 APS, 1573 ELO, 1803 Glicko2, 2089 battles (30-May-2009)
** Ranked 163 in OneOnOne (1768.96pts, 2225 battles, 31-Jul-2008)
+
** Not-so-improved movement?  My tests showed a noticeable increase vs top bots, the rumble showed that I neglected to test against simpler bots, which ate Gaff alive using HOT!
* 1.04:  Major improvements to targeting after spending some time with the targeting challenges and RoboResearch
+
 
** Now uses only one targeting method based on a small neural network with only one hidden layer
+
'''1.34''': Combined + improved upon 1.32 targeting methods
** Added [[GuessFactor]] range limiting using my FutureMove class
+
* After 15 seasons, squeaks into top spot of fast-learning [[TC2K7]]
** Fixed a nasty bug in FutureMove that also affected dodging
+
* Uses targeting methods from 1.32 and 1.32rm and combines the outputs
** Ranked 229 in OneOnOne (1688pts, 2007 battles, 27-Jun-2008)
+
* Improved input data classification by using radial basis functions
* 1.00:  Initial release
+
* Tweaked movement to avoid corners somewhat
** Ranked 313 in OneOnOne (1608pts, 23-Jun-2008)
+
* Ranked 91 in [[RoboRumble]] -- 69.66 APS, 1646 ELO, 2188 battles (26-May-2009)
 +
 
 +
'''1.32rm''': Alternate version of 1.32 with improved random mover targeting
 +
* Same input data as 1.32, using NN training scheme designed to learn vs. random movers.
 +
* Ranked 96 in [[RoboRumble]] -- 69.16 APS, 1675 ELO, 2241 battles (13-May-2009)
 +
 
 +
'''1.32''': Featuring new & improved anti-surfer targeting
 +
* New input data method to NN targeting scores very high against surfers in [[TC2K7]].
 +
** Targeting data inspired by data segmentation in [[Skilgannon]]'s [[DrussGT]]
 +
* Fixed bug where waves were not cleared at the end of every round.
 +
* Improved min/max reachable GF calculation and fixed related bug in NN training.
 +
* Ranked 95 in [[RoboRumble]] -- 69.32 APS, 1689.8 ELO, 2239 battles (12-May-2009)
 +
 
 +
'''1.30''': Reworked movement
 +
* Orbital movements now take into account max. reachable angle
 +
* Added dynamic distancing based on energy differential
 +
* Removed cornering movement
 +
* Ranked 96 in [[RoboRumble]] -- 69.29 APS, 1714.9 ELO, 3705 battles (9-May-2009)
 +
 
 +
'''1.28''':  Minor tweaks to cornering and movement selection
 +
* Cornering no longer selected when Gaff is already cornered.
 +
* Damage differential now reset every round
 +
* Ranked ''xx'' in [[RoboRumble]] == 68.45 APS, 1819 ELO, 2175 battles (17-Dec-2009)
 +
 
 +
'''1.26''':  Tweaked movement to improve performance against worst [[ProblemBots]] and a selection of top bots.  No targeting changes.
 +
* Cornering movement now used only when Gaff has a significant energy advantage
 +
* Random and dodging orbital styles selected based on current damage differential; now switches with same round if needed.
 +
* Bullet predictor decay rate decreased dramatically for faster adaptation.
 +
* Distancing adjusted to make orbital modes similar and decrease advantage of segmenting on distance (makes a big difference).
 +
* Fixed some bugs in the random mode which resulted in strange moves.
 +
* Ranked 106 in OneOnOne (1805.94pts, 2059 battles, 20-Aug-2008)
 +
 
 +
'''1.24''':  Mostly movement updates
 +
* Smoothed out dodging orbit movement mode
 +
* Added a random orbit (non-dodging) movement mode
 +
* Movement selection now settles on new random mode if other two modes lose too often
 +
* Increased bullet predictor decay rate
 +
* Minor tweaks to targeting NN to adapt faster and handle lower bullet powers better (but no challenge score improvement)
 +
* Ranked 97 in OneOnOne (1816.7pts, 2061 battles, 17-Aug-2008)
 +
 
 +
'''1.20''':  Further targeting improvements, many movement bug fixes
 +
* Targeting network no longer uses hidden layers.
 +
* Network is retrained at a slower rate but more often.
 +
* Fixed many, many bugs in the orbital movement and bullet dodging classes.
 +
* Removed the "oddball" movement style
 +
* Changed the movement style selection method
 +
* Ranked 117 in OneOnOne (1800.74pts, 2122 battles, 6-Aug-2008)
 +
 
 +
'''1.10''':  Continued improvements to targeting and some new movement
 +
* Targeting method now trains multiple samples per firing wave
 +
* Fixed a major bug in the bullet dodging code
 +
* Credits for both of the above go to [[Rednaxela]]
 +
* Added two new movement styles (orbital and "oddball") for situations where cornering isn't working out.
 +
* Ranked 163 in OneOnOne (1768.96pts, 2225 battles, 31-Jul-2008)
 +
 
 +
'''1.04''':  Major improvements to targeting after spending some time with the targeting challenges and RoboResearch
 +
* Now uses only one targeting method based on a small neural network with only one hidden layer
 +
* Added [[GuessFactor]] range limiting using my FutureMove class
 +
* Fixed a nasty bug in FutureMove that also affected dodging
 +
* Ranked 229 in OneOnOne (1688pts, 2007 battles, 27-Jun-2008)
 +
 
 +
'''1.00''':  Initial release
 +
* Ranked 313 in OneOnOne (1608pts, 23-Jun-2008)

Latest revision as of 06:59, 16 September 2009

1.50 - Back to your corner!
  • After 1.44's poor performance, a more drastic revision was in order.
  • Gaff's new movement is a return to his original roots: cornering is back.
    • This version uses Holden's wave surfing (far superior to the old bullet dodging) and a simpler cornering plan -- not as crafty but Gaff takes far less hits than before.
    • As a fallback, Gaff reverts to a standard orbital movement but with very tight distancing for that up-close, personal feeling.
  • Some targeting tweaks to boost TCRM performance above 88.
  • Plus bugfixes and a bit of optimization.


1.44: Surfer killer?

  • RoboRumble ‒ APS: 68.06% (108th), PL: 617-126 (111th), Survival: 76.47%
  • Updated targeting -- same as 1.43g which topped the heap in the AS Challenge
    • Now uses a slightly different input data set for NNs (10 variables)
    • Features approximately doubled in number
  • Many, many small framework tweaks carried over from Pris and Holden development
    • Now adjusts aim for tank movement
    • Hopefully fixes radar lock bug

1.42: Revamped movement

  • Restored the orbital dodging mode which was keeping scores up vs simple targeters.
  • Dropped the older random orbit movement in favour of the newer Raiko-style one.
  • Tweaked some distances and fixed a minor distancing bug.
  • Expanded my test bot set and this one worked out better than 1.34.
  • Ranked 91 in RoboRumble: 69.9 APS, 1546 ELO, 1869 Glicko2, 2841 battles (14-Aug-2009)

1.40: Improved movement

  • Created a pretty slick, low cost orbital movement to replace the ugly iterative one in previous versions
  • Smoothed out dynamic distancing
  • Replaced the buggy orbital dodge movement mode with another random movement inspired by Raiko's algorithm.
  • Ranked 126 in RoboRumble -- 65.10 APS, 1573 ELO, 1803 Glicko2, 2089 battles (30-May-2009)
    • Not-so-improved movement? My tests showed a noticeable increase vs top bots, the rumble showed that I neglected to test against simpler bots, which ate Gaff alive using HOT!

1.34: Combined + improved upon 1.32 targeting methods

  • After 15 seasons, squeaks into top spot of fast-learning TC2K7
  • Uses targeting methods from 1.32 and 1.32rm and combines the outputs
  • Improved input data classification by using radial basis functions
  • Tweaked movement to avoid corners somewhat
  • Ranked 91 in RoboRumble -- 69.66 APS, 1646 ELO, 2188 battles (26-May-2009)

1.32rm: Alternate version of 1.32 with improved random mover targeting

  • Same input data as 1.32, using NN training scheme designed to learn vs. random movers.
  • Ranked 96 in RoboRumble -- 69.16 APS, 1675 ELO, 2241 battles (13-May-2009)

1.32: Featuring new & improved anti-surfer targeting

  • New input data method to NN targeting scores very high against surfers in TC2K7.
  • Fixed bug where waves were not cleared at the end of every round.
  • Improved min/max reachable GF calculation and fixed related bug in NN training.
  • Ranked 95 in RoboRumble -- 69.32 APS, 1689.8 ELO, 2239 battles (12-May-2009)

1.30: Reworked movement

  • Orbital movements now take into account max. reachable angle
  • Added dynamic distancing based on energy differential
  • Removed cornering movement
  • Ranked 96 in RoboRumble -- 69.29 APS, 1714.9 ELO, 3705 battles (9-May-2009)

1.28: Minor tweaks to cornering and movement selection

  • Cornering no longer selected when Gaff is already cornered.
  • Damage differential now reset every round
  • Ranked xx in RoboRumble == 68.45 APS, 1819 ELO, 2175 battles (17-Dec-2009)

1.26: Tweaked movement to improve performance against worst ProblemBots and a selection of top bots. No targeting changes.

  • Cornering movement now used only when Gaff has a significant energy advantage
  • Random and dodging orbital styles selected based on current damage differential; now switches with same round if needed.
  • Bullet predictor decay rate decreased dramatically for faster adaptation.
  • Distancing adjusted to make orbital modes similar and decrease advantage of segmenting on distance (makes a big difference).
  • Fixed some bugs in the random mode which resulted in strange moves.
  • Ranked 106 in OneOnOne (1805.94pts, 2059 battles, 20-Aug-2008)

1.24: Mostly movement updates

  • Smoothed out dodging orbit movement mode
  • Added a random orbit (non-dodging) movement mode
  • Movement selection now settles on new random mode if other two modes lose too often
  • Increased bullet predictor decay rate
  • Minor tweaks to targeting NN to adapt faster and handle lower bullet powers better (but no challenge score improvement)
  • Ranked 97 in OneOnOne (1816.7pts, 2061 battles, 17-Aug-2008)

1.20: Further targeting improvements, many movement bug fixes

  • Targeting network no longer uses hidden layers.
  • Network is retrained at a slower rate but more often.
  • Fixed many, many bugs in the orbital movement and bullet dodging classes.
  • Removed the "oddball" movement style
  • Changed the movement style selection method
  • Ranked 117 in OneOnOne (1800.74pts, 2122 battles, 6-Aug-2008)

1.10: Continued improvements to targeting and some new movement

  • Targeting method now trains multiple samples per firing wave
  • Fixed a major bug in the bullet dodging code
  • Credits for both of the above go to Rednaxela
  • Added two new movement styles (orbital and "oddball") for situations where cornering isn't working out.
  • Ranked 163 in OneOnOne (1768.96pts, 2225 battles, 31-Jul-2008)

1.04: Major improvements to targeting after spending some time with the targeting challenges and RoboResearch

  • Now uses only one targeting method based on a small neural network with only one hidden layer
  • Added GuessFactor range limiting using my FutureMove class
  • Fixed a nasty bug in FutureMove that also affected dodging
  • Ranked 229 in OneOnOne (1688pts, 2007 battles, 27-Jun-2008)

1.00: Initial release

  • Ranked 313 in OneOnOne (1608pts, 23-Jun-2008)