Difference between revisions of "Monk/VersionHistory"

From Robowiki
Jump to navigation Jump to search
 
(31 intermediate revisions by the same user not shown)
Line 1: Line 1:
 
<!--
 
<!--
'''{{subst:rumblestats:meleerumble|rsalesc.melee.Monk 0.1}}'''
+
'''{{subst:rumblestats:meleerumble|rsalesc.melee.Monk 0.3}}'''
 
-->
 
-->
 +
 +
; 0.3.6c - 10.09.2017
 +
* '''PENDING'''
 +
** Rollback performance enchancing bug.
 +
 +
; 0.3.6b - 10.09.2017
 +
* '''skipped'''
 +
** Small fix.
 +
 +
; 0.3.6 - 10.09.2017
 +
* '''skipped'''
 +
** Add wall avoidance.
 +
** Fix a max(0) bug in risk function, which probably turns out to be a performance enchancing bug.
 +
 +
; 0.3.5 - 10.09.2017
 +
* '''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.5|rumble=MeleeRumble|scorelabel=APS|score=66.83|rank=11th|win=386|loss=13|plrank=13th|glicko2=N/A|pwin=96.74|vote=1.5|anpp=90.37|score2label=Survival|score2=40.67}}'''
 +
** Add 1v1 guns back.
 +
** Incorporate changings from [[Knight]] 0.4.4.
 +
** Improve painting management.
 +
** Add bullet shadows.
 +
** Put PIF-Gun back into the Swarm component. Hope it doesn't perform much worse than 0.3.4x.
 +
** Fix a small bug in the Swarm component which would pick no best angle and return null, possibly crashing the bot.
 +
** Add a stop option in the movement, which was not present until this version.
 +
** Set the maximum number of neighbors to 100 and add a procedure to calculate the maximum # of queryable data for each enemy, then take the min and use that as our K for the knn, so we can ensure that we use the same amount of neighbors for each enemy.
 +
 +
; 0.3.4e to 0.3.4h
 +
* '''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.4h|rumble=MeleeRumble|scorelabel=APS|score=66.75|rank=11th|win=385|loss=14|plrank=12th|glicko2=N/A|pwin=96.49|vote=0.5|anpp=90.87|score2label=Survival|score2=39.8}}'''
 +
** Try to figure out what is causing the score inconsistency.
 +
 +
; 0.3.4d - 09.26.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.4d|rumble=MeleeRumble|scorelabel=APS|score=65.42|rank=14th|win=374|loss=25|plrank=26th|glicko2=N/A|pwin=93.73|vote=0.25|anpp=88.72|score2label=Survival|score2=33.34}}'''
 +
** Stick only with PIF gun again. There seems to be a problem using the guessfactor guns in melee.
 +
 +
; 0.3.4b to 0.3.4c - 09.25.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.4c|rumble=MeleeRumble|scorelabel=APS|score=66.17|rank=13th|win=385|loss=14|plrank=13th|glicko2=N/A|pwin=96.49|vote=0.75|anpp=89.81|score2label=Survival|score2=39.75}}'''
 +
** Take out random gun.
 +
 +
; 0.3.4 - 09.25.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.4|rumble=MeleeRumble|scorelabel=APS|score=65.7|rank=13th|win=380|loss=19|plrank=18th|glicko2=N/A|pwin=95.24|vote=3.26|anpp=89.25|score2label=Survival|score2=36.56}}'''
 +
** Try 1v1 guns.
 +
 +
; 0.3.3 - 09.25.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.3|rumble=MeleeRumble|scorelabel=APS|score=67.1|rank=10th|win=388|loss=11|plrank=12th|glicko2=N/A|pwin=97.24|vote=5.3|anpp=91.47|score2label=Survival|score2=46.32}}'''
 +
** Increase danger of enemies and add perpendicularity danger factor inversed by distance. As a consequence, stand more close to the wall while having enough mobility to stand against enemies from every side of the field.
 +
 +
; 0.3.2g - 09.25.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.2g|rumble=MeleeRumble|scorelabel=APS|score=67.29|rank=10th|win=380|loss=19|plrank=18th|glicko2=N/A|pwin=95.24|vote=11.11|anpp=86.54|score2label=Survival|score2=42.75}}'''
 +
** Use the distance to the predicted enemy position instead of the current position to determine bandwidth and inverse distance in the swarm gun (as it really was in the old gun, nothing new here).
 +
 +
; 0.3.2f - 09.25.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.2f|rumble=MeleeRumble|scorelabel=APS|score=66.63|rank=12th|win=382|loss=17|plrank=18th|glicko2=N/A|pwin=95.74|vote=0.25|anpp=90.47|score2label=Survival|score2=42.26}}'''
 +
** Try some changes which to make the new gun almost equal to the older. I still don't know where the score discrepancy is coming from.
 +
 +
; 0.3.2e - 09.25.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.2e|rumble=MeleeRumble|scorelabel=APS|score=66.54|rank=12th|win=257|loss=17|plrank=138th|glicko2=N/A|pwin=93.8|vote=0|anpp=0|score2label=Survival|score2=39.77}}'''
 +
** Compute firing angle every tick like in the old gun.
 +
 +
; 0.3.2d - 09.24.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.2d|rumble=MeleeRumble|scorelabel=APS|score=66.61|rank=11th|win=379|loss=19|plrank=19th|glicko2=N/A|pwin=95.23|vote=3.08|anpp=90.54|score2label=Survival|score2=43.2}}'''
 +
** My final effort to equalize the new gun and the old gun.
 +
 +
; 0.3.2c - skipped
 +
 +
; 0.3.2b - 09.24.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.2b|rumble=MeleeRumble|scorelabel=APS|score=66.25|rank=12th|win=381|loss=18|plrank=18th|glicko2=N/A|pwin=95.49|vote=1.75|anpp=89.95|score2label=Survival|score2=38.5}}'''
 +
** Put the new gun back with some changes. Nothing is really a fix.
 +
 +
; 0.3.2 - 09.23.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.2|rumble=MeleeRumble|scorelabel=APS|score=67.51|rank=10th|win=388|loss=11|plrank=11th|glicko2=N/A|pwin=97.24|vote=3.01|anpp=91.8|score2label=Survival|score2=43.61}}'''
 +
** Use old gun from 0.2.3c, but everything else is the same.
 +
** Clearly something is wrong with the new gun.
 +
 +
; 0.3 to 0.3.1b - 09.23.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.3.1b|rumble=MeleeRumble|scorelabel=APS|score=63.98|rank=24th|win=368|loss=31|plrank=34th|glicko2=N/A|pwin=92.23|vote=4|anpp=86.89|score2label=Survival|score2=35.99}}'''
 +
** Incorporate new guns from [[Knight]] (1v1).
 +
** Tweak PIF gun and add HOT for when there is no data.
 +
** Performance shouldn't be that bad. Every day I feel like I broke something but I still don't know what.
 +
 +
; 0.2.3c - 09.15.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.2.3c|rumble=MeleeRumble|scorelabel=APS|score=67.64|rank=10th|win=386|loss=12|plrank=12th|glicko2=N/A|pwin=96.98|vote=0|anpp=0|score2label=Survival|score2=44.51}}'''
 +
** Fix 1v1 movement.
 +
** Try to avoid gun thrashing between enemies.
 +
 +
; 0.2.3b
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.2.3b|rumble=MeleeRumble|scorelabel=APS|score=67.18|rank=10th|win=383|loss=15|plrank=14th|glicko2=N/A|pwin=96.23|vote=1.51|anpp=91.49|score2label=Survival|score2=41.47}}'''
 +
** Movement tweaking.
 +
 +
; 0.2.3 - 09.13.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.2.3|rumble=MeleeRumble|scorelabel=APS|score=67.05|rank=10th|win=382|loss=16|plrank=15th|glicko2=N/A|pwin=95.98|vote=3.02|anpp=91.04|score2label=Survival|score2=41.18}}'''
 +
** Incorporate [[Knight]] 0.1.1 changes which are most 1v1 changes, detailed below.
 +
** ''Incorporate [[Roborio]]'s movement trees with different use conditions. No flattener, though.''
 +
** ''Revert order my afterRun() function is executed.''
 +
** ''Log when wave passes me. Still have to log when my wave passes an enemy, though.''
 +
** ''Make a class dedicated to hold stats from the battle, which is used, for example, to decide which tree to use in the movement. It can be used just for logging purposes as well.''
 +
 +
; 0.2.2 - 09.13.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.2.2|rumble=MeleeRumble|scorelabel=APS|score=67.41|rank=10th|win=384|loss=14|plrank=13th|glicko2=N/A|pwin=96.48|vote=9.92|anpp=91.91|score2label=Survival|score2=39.7}}'''
 +
** Add energy management.
 +
 +
; 0.2.1 - 09.13.2017
 +
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.2.1|rumble=MeleeRumble|scorelabel=APS|score=66.26|rank=12th|win=380|loss=18|plrank=17th|glicko2=N/A|pwin=95.48|vote=6.32|anpp=90.28|score2label=Survival|score2=34.47}}'''
 +
** Fix an amateur bug when switching from melee movement to one-on-one movement. Probably was hurting my survival score.
 +
** Take distance to target into account when picking an angle to shoot (1 / sqrt(distance)).
 +
 +
; 0.2 - 09.12.2017 (skipped pairings)
 +
** Stored bullet, scan and death events and handle them separately. Hope I didn't break anything.
 +
** Add wave surfing from [[Roborio]], but with simple attributes and weightings.
 +
** Don't expect it to improve over the older version, just wanna make sure I didn't break anything.
  
 
; 0.1 - 09.12.2017 - first release
 
; 0.1 - 09.12.2017 - first release
'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.1|rumble=MeleeRumble|scorelabel=APS|score=65.21|rank=14th|win=368|loss=30|plrank=30th|glicko2=N/A|pwin=92.46|vote=0|anpp=0|score2label=Survival|score2=27.12}}'''
+
*'''{{RumbleStatsDefault|link=http://literumble.appspot.com/BotDetails?game=meleerumble&name=rsalesc.melee.Monk%200.1|rumble=MeleeRumble|scorelabel=APS|score=65.21|rank=14th|win=368|loss=30|plrank=30th|glicko2=N/A|pwin=92.46|vote=0|anpp=0|score2label=Survival|score2=27.12}}'''
* My first release of any bot based on my framework BackAsFrontRobot2.
+
** My first release of any bot based on my framework BackAsFrontRobot2.
* Simple DC-PIF gun based on [[RoborioPorradeiro]] with some basic parameters and some fixes.
+
** Simple DC-PIF swarm gun based on [[RoborioPorradeiro]] with some basic parameters and some fixes.
* [[Minimum Risk Movement]] based on some past experiments. No wall avoidance and reaction to bullets at all.
+
** [[Minimum Risk Movement]] based on some past experiments. No wall avoidance and reaction to bullets at all.
* Radar which sweeps in the direction of the robot which has the oldest latest scan.
+
** Radar which sweeps in the direction of the robot which has the oldest latest scan.
* No 1v1 strategy.
+
** No 1v1 strategy.
 +
** No energy management at all. Always fires 1.95.

Latest revision as of 02:23, 10 October 2017


0.3.6c - 10.09.2017
  • PENDING
    • Rollback performance enchancing bug.
0.3.6b - 10.09.2017
  • skipped
    • Small fix.
0.3.6 - 10.09.2017
  • skipped
    • Add wall avoidance.
    • Fix a max(0) bug in risk function, which probably turns out to be a performance enchancing bug.
0.3.5 - 10.09.2017
  • MeleeRumble ‒ APS: 66.83% (11th), PL: 386-13 (13th), Survival: 40.67%
    • Add 1v1 guns back.
    • Incorporate changings from Knight 0.4.4.
    • Improve painting management.
    • Add bullet shadows.
    • Put PIF-Gun back into the Swarm component. Hope it doesn't perform much worse than 0.3.4x.
    • Fix a small bug in the Swarm component which would pick no best angle and return null, possibly crashing the bot.
    • Add a stop option in the movement, which was not present until this version.
    • Set the maximum number of neighbors to 100 and add a procedure to calculate the maximum # of queryable data for each enemy, then take the min and use that as our K for the knn, so we can ensure that we use the same amount of neighbors for each enemy.
0.3.4e to 0.3.4h
  • MeleeRumble ‒ APS: 66.75% (11th), PL: 385-14 (12th), Survival: 39.8%
    • Try to figure out what is causing the score inconsistency.
0.3.4d - 09.26.2017
  • MeleeRumble ‒ APS: 65.42% (14th), PL: 374-25 (26th), Survival: 33.34%
    • Stick only with PIF gun again. There seems to be a problem using the guessfactor guns in melee.
0.3.4b to 0.3.4c - 09.25.2017
  • MeleeRumble ‒ APS: 66.17% (13th), PL: 385-14 (13th), Survival: 39.75%
    • Take out random gun.
0.3.4 - 09.25.2017
  • MeleeRumble ‒ APS: 65.7% (13th), PL: 380-19 (18th), Survival: 36.56%
    • Try 1v1 guns.
0.3.3 - 09.25.2017
  • MeleeRumble ‒ APS: 67.1% (10th), PL: 388-11 (12th), Survival: 46.32%
    • Increase danger of enemies and add perpendicularity danger factor inversed by distance. As a consequence, stand more close to the wall while having enough mobility to stand against enemies from every side of the field.
0.3.2g - 09.25.2017
  • MeleeRumble ‒ APS: 67.29% (10th), PL: 380-19 (18th), Survival: 42.75%
    • Use the distance to the predicted enemy position instead of the current position to determine bandwidth and inverse distance in the swarm gun (as it really was in the old gun, nothing new here).
0.3.2f - 09.25.2017
  • MeleeRumble ‒ APS: 66.63% (12th), PL: 382-17 (18th), Survival: 42.26%
    • Try some changes which to make the new gun almost equal to the older. I still don't know where the score discrepancy is coming from.
0.3.2e - 09.25.2017
  • MeleeRumble ‒ APS: 66.54% (12th), PL: 257-17 (138th), Survival: 39.77%
    • Compute firing angle every tick like in the old gun.
0.3.2d - 09.24.2017
  • MeleeRumble ‒ APS: 66.61% (11th), PL: 379-19 (19th), Survival: 43.2%
    • My final effort to equalize the new gun and the old gun.
0.3.2c - skipped
0.3.2b - 09.24.2017
  • MeleeRumble ‒ APS: 66.25% (12th), PL: 381-18 (18th), Survival: 38.5%
    • Put the new gun back with some changes. Nothing is really a fix.
0.3.2 - 09.23.2017
  • MeleeRumble ‒ APS: 67.51% (10th), PL: 388-11 (11th), Survival: 43.61%
    • Use old gun from 0.2.3c, but everything else is the same.
    • Clearly something is wrong with the new gun.
0.3 to 0.3.1b - 09.23.2017
  • MeleeRumble ‒ APS: 63.98% (24th), PL: 368-31 (34th), Survival: 35.99%
    • Incorporate new guns from Knight (1v1).
    • Tweak PIF gun and add HOT for when there is no data.
    • Performance shouldn't be that bad. Every day I feel like I broke something but I still don't know what.
0.2.3c - 09.15.2017
  • MeleeRumble ‒ APS: 67.64% (10th), PL: 386-12 (12th), Survival: 44.51%
    • Fix 1v1 movement.
    • Try to avoid gun thrashing between enemies.
0.2.3b
  • MeleeRumble ‒ APS: 67.18% (10th), PL: 383-15 (14th), Survival: 41.47%
    • Movement tweaking.
0.2.3 - 09.13.2017
  • MeleeRumble ‒ APS: 67.05% (10th), PL: 382-16 (15th), Survival: 41.18%
    • Incorporate Knight 0.1.1 changes which are most 1v1 changes, detailed below.
    • Incorporate Roborio's movement trees with different use conditions. No flattener, though.
    • Revert order my afterRun() function is executed.
    • Log when wave passes me. Still have to log when my wave passes an enemy, though.
    • Make a class dedicated to hold stats from the battle, which is used, for example, to decide which tree to use in the movement. It can be used just for logging purposes as well.
0.2.2 - 09.13.2017
  • MeleeRumble ‒ APS: 67.41% (10th), PL: 384-14 (13th), Survival: 39.7%
    • Add energy management.
0.2.1 - 09.13.2017
  • MeleeRumble ‒ APS: 66.26% (12th), PL: 380-18 (17th), Survival: 34.47%
    • Fix an amateur bug when switching from melee movement to one-on-one movement. Probably was hurting my survival score.
    • Take distance to target into account when picking an angle to shoot (1 / sqrt(distance)).
0.2 - 09.12.2017 (skipped pairings)
    • Stored bullet, scan and death events and handle them separately. Hope I didn't break anything.
    • Add wave surfing from Roborio, but with simple attributes and weightings.
    • Don't expect it to improve over the older version, just wanna make sure I didn't break anything.
0.1 - 09.12.2017 - first release
  • MeleeRumble ‒ APS: 65.21% (14th), PL: 368-30 (30th), Survival: 27.12%
    • My first release of any bot based on my framework BackAsFrontRobot2.
    • Simple DC-PIF swarm gun based on RoborioPorradeiro with some basic parameters and some fixes.
    • Minimum Risk Movement based on some past experiments. No wall avoidance and reaction to bullets at all.
    • Radar which sweeps in the direction of the robot which has the oldest latest scan.
    • No 1v1 strategy.
    • No energy management at all. Always fires 1.95.