Editing GresSuffurd/GresVretter

Robo Home | Changes | Preferences | AllPages

[[GresSuffurd/GresVretter|GresVretter]] is the gun of GresSuffurd. It starts out as a really really (really?) simple GuessFactorTargeting gun, based upon [[Kawigi]]'s [[GuessFactorTargeting/Tutorial|GuessFactorTargeting Tutorial]]. It will gradually improve until it is good enough for [[The2000Club]].

----
'''version history''' (newest entry on top)

<font size = -1>
* '''0.2.6''' (20070523) 
** removed decay in gunstats
** removed GF1 protection factor
* '''0.2.5''' (20070222) 
** changed wallsegmentation from 3 to 5 bins (no - heading within 75 dist - heading 150 - retreat 75 - retreat 150)
* '''0.2.4''' (20070207) 
** do not fire virtual waves if opponent has not moved yet
* '''0.2.3''' (20070125) 
** small bugfix weighting non-bullet waves in unsegmented buffer
** introduce some decay in gunstats
* '''0.2.2''' (20061212) 
** reverted anti-ram behaviour
* '''0.2.1''' (20061127) 
** adapted energymanagement when enemy is close (anti-ram)
* '''0.2.0''' (20061116) 
** also firing non-bullet waves with 0.1 weight
* '''0.1.5''' (20061107) 
** using proper onDeath (by David_Alves) and keep events in line with WeefSuffurd
* '''0.1.4''' (20061007) 
** bugfixing hitted bin detection
** fire HOT and without waves on disabled enemies
* '''0.1.3''' (20061005) 
** as 0.1.1 plus : record all hitted bins and fire on highest concentration (3-bin wide)
* '''0.1.2''' (20061001) 
** segmenting on lateral ( 0 - 2.5 - 5.5 - 8) and advancing ( -8 - -4 - 4 - 8) velocity only
* '''0.1.1''' (200609xx) 
** no changes, just some cosmetical adjustments. Released in the rumble as part of GresSuffurd v0.1.0
* '''0.1.0''' (20060630) 
** added trick for faster learning (quite rudimentary (and probably not effective))
** added [[Voidious]]' tip about asin(8/bulletvelocity)*constant
* '''0.0.7''' (20060627) 
** segmenting on acceleration, distance and wallproximity (segments as below)
* '''0.0.6''' (20060624) 
** segmenting on wallproximity only ( none - heading for wall within 90 pixels - retreating from wall within 90 pixels )
* '''0.0.5''' (20060622) 
** segmenting on acceleration only ( accell - decell - time < 6 - time < 16 - time < 36 - time >= 36 )
* '''0.0.4''' (20060618) 
** segmenting on absvelocity only ( 0 - 1.9 - 6.1 - oo )
* '''0.0.3''' (20060612) 
** segmenting on distance only ( 0-300-450-600-oo )
* '''0.0.2''' (20060531) 
** unsegmented gun (37 bins), firing only bulletwaves, with one-bin hitdetection

</font>
==== [[RRGunChallenge]] ====
| '''Version''' | '''RR Rating''' | '''PL Rating''' | '''Date''' | '''Comments'''
|  0.2.0        |  1829 (59)      |  420-58         | 20061118   | still 40 points behind topguns
|  0.1.0        |  1791 (66)      |  357-67         | 20060630   | still very much to do
|  0.0.3        |  1758 (87)      |  348-74         | 20060621   | Just for reference (close to GrubbmGrbRRGC)

----

==== [[TargetingChallenge2K6|TC2K6 500 rounds]] ====
| '''Version''' | '''BFly''' | '''CC''' | '''Chk''' | '''Cig''' | '''Cya''' | '''DM''' | '''FM''' | '''Grb''' | '''RMB''' | '''Tig''' | '''Total''' | '''Seasons'''
| 0.2.5 | 97.25 | 59.09 | 80.76 | 80.67 | 71.38 | 93.77 | 91.31 | 88.01 | 86.68 | 74.44 | '''82.34''' |  2
| 0.1.4 | 96.11 | 48.47 | 70.18 | 79.79 | 64.73 | 93.43 | 91.17 | 87.72 | 86.97 | 68.87 | '''78.75''' |  1
| 0.0.7 | 96.24 | 45.31 | 72.57 | 78.25 | 59.62 | 91.74 | 88.12 | 89.72 | 86.50 | 66.17 | '''77.42''' |  1
|  Grb  | 89.78 | 48.80 | 57.37 | 73.56 | 49.48 | 89.37 | 86.28 | 83.25 | 88.96 | 63.44 | '''73.03''' |  1
| 0.0.5 | 87.97 | 36.39 | 62.10 | 74.15 | 33.74 | 91.45 | 87.84 | 87.13 | 81.95 | 54.35 | '''69.71''' |  1
| 0.0.3 | 88.50 | 34.65 | 60.53 | 76.20 | 36.42 | 89.22 | 85.10 | 84.80 | 85.25 | 43.89 | '''68.46''' |  1  
| 0.1.2 | 93.83 | 27.22 | 53.39 | 69.37 | 23.43 | 89.08 | 80.99 | 81.76 | 76.48 | 44.86 | '''64.04''' |  1
| 0.0.2 | 88.95 | 28.09 | 54.46 | 67.79 | 22.75 | 90.24 | 80.80 | 81.07 | 77.82 | 38.60 | '''63.06''' |  1

----

==== [[TargetingChallenge2K6/FastLearning|TC2K6 FastLearning]] ====

| '''Version''' | '''BFly''' | '''CC''' | '''Chk''' | '''Cig''' | '''Cya''' | '''DM''' | '''FM''' | '''Grb''' | '''RMB''' | '''Tig''' | '''Total''' | '''Seasons'''
| 0.2.5 | 94.34 | 57.69 | 76.85 | 75.39 | 64.94 | 89.50 | 88.91 | 84.95 | 82.14 | 77.15 | '''79.19''' | 15 
| 0.1.4 | 93.03 | 53.38 | 67.97 | 73.09 | 61.09 | 89.98 | 92.40 | 85.57 | 82.37 | 68.60 | '''76.75''' | 15
| 0.2.0 | 95.68 | 45.79 | 65.23 | 74.72 | 62.78 | 89.94 | 90.15 | 88.03 | 85.99 | 65.65 | '''76.40''' | 15
| 0.0.7 | 92.43 | 52.60 | 68.10 | 70.93 | 60.62 | 85.52 | 87.83 | 82.29 | 81.45 | 68.01 | '''74.98''' | 15
| 0.1.0 | 90.74 | 50.43 | 63.50 | 72.02 | 56.28 | 89.43 | 88.04 | 85.19 | 81.56 | 67.24 | '''74.45''' | 15
|  Grb  | 84.30 | 45.48 | 55.82 | 67.14 | 47.19 | 87.89 | 88.25 | 83.04 | 87.54 | 65.45 | '''71.21''' | 15
| 0.0.5 | 91.55 | 41.02 | 62.08 | 72.22 | 41.73 | 90.14 | 81.67 | 82.82 | 80.03 | 57.42 | '''70.07''' | 15
| 0.0.3 | 80.62 | 37.42 | 60.53 | 68.74 | 42.43 | 87.89 | 85.44 | 81.79 | 80.57 | 46.82 | '''67.23''' | 15
| 0.1.2 | 90.67 | 32.98 | 52.26 | 65.75 | 31.82 | 88.81 | 78.97 | 81.65 | 77.77 | 46.04 | '''64.67''' | 15
| 0.0.2 | 79.29 | 27.96 | 55.30 | 64.31 | 27.63 | 88.68 | 77.88 | 81.20 | 78.95 | 38.94 | '''62.01''' | 15
----

==== [[TargetingChallenge|TC old 500 rounds]] ====
| '''Version''' | '''DT''' | '''Asp''' | '''TAOW''' | '''Spar''' | '''Cig''' | '''Tron''' | '''Fhqw''' | '''HTTC''' | '''Yngw''' | '''Funk''' | '''Total''' | '''Comment'''
| 0.2.5 | 82.61 | 90.71 | 100.00| 97.95 | 80.48 | 88.09 | 94.85 | 89.35 | 98.00 | 89.03 | '''91.11''' | 2 seasons
| 0.1.4 | 81.10 | 90.90 | 100.00| 98.33 | 81.17 | 86.26 | 95.53 | 88.93 | 97.62 | 88.22 | '''90.81''' | 1 season
| 0.2.0 | 82.07 | 90.45 | 99.91 | 97.56 | 78.32 | 88.86 | 95.92 | 88.28 | 97.77 | 87.57 | '''90.67''' | 2 seasons
| 0.0.7 | 79.15 | 89.45 | 99.86 | 96.94 | 79.17 | 85.19 | 92.49 | 88.49 | 97.06 | 87.13 | '''89.49''' | 1 season
| 0.0.5 | 78.27 | 89.13 | 100.00| 94.65 | 71.98 | 86.75 | 88.43 | 90.00 | 95.71 | 85.10 | '''88.00''' | 1 season
| 0.0.3 | 61.50 | 83.19 | 99.96 | 92.59 | 74.43 | 80.22 | 81.09 | 87.07 | 94.58 | 84.84 | '''83.96''' | 1 season
| 0.0.6 | 63.86 | 85.07 | 99.78 | 89.17 | 72.18 | 83.98 | 82.62 | 84.71 | 92.37 | 82.92 | '''83.67''' | 1 season
|  Grb  | 58.37 | 79.62 | 87.36 | 95.28 | 71.76 | 80.97 | 91.17 | 84.51 | 96.64 | 88.57 | '''83.42''' | 3 seasons
| 0.1.2 | 70.23 | 86.24 | 99.73 | 89.86 | 67.31 | 82.11 | 77.11 | 84.73 | 94.14 | 80.89 | '''83.23''' | 1 season
| 0.0.4 | 69.71 | 85.57 | 99.96 | 90.63 | 63.73 | 81.59 | 72.50 | 86.26 | 94.38 | 80.85 | '''82.52''' | 1 season
| 0.0.2 | 64.01 | 85.48 | 98.91 | 87.55 | 69.22 | 80.30 | 71.78 | 86.39 | 95.30 | 80.92 | '''81.98''' | 1 season

==== [[TargetingChallenge/FastLearning|TC old FastLearning]] ====
| '''Version''' | '''DT''' | '''Asp''' | '''TAOW''' | '''Spar''' | '''Cig''' | '''Tron''' | '''Fhqw''' | '''HTTC''' | '''Yngw''' | '''Funk''' | '''Total''' | '''Comment'''
| 0.2.0 | 75.33 | 87.16 | 99.99 | 95.35 | 73.53 | 85.29 | 94.97 | 83.98 | 95.80 | 87.08 | '''87.85''' | 30 seasons
| 0.2.5 | 75.11 | 86.98 | 99.87 | 96.63 | 74.50 | 83.41 | 93.94 | 84.71 | 95.58 | 85.60 | '''87.63''' | 15 seasons
| 0.1.4 | 70.69 | 87.93 | 99.73 | 95.32 | 72.30 | 83.03 | 92.50 | 85.18 | 95.38 | 84.91 | '''86.70''' | 15 seasons
|  Grb  | 58.72 | 76.04 | 86.00 | 94.01 | 67.05 | 80.57 | 90.32 | 82.37 | 96.37 | 86.50 | '''81.79''' | 15 seasons

As [[PEZ]] already stated with his GFTargetingBot, WaveSurfers are designed to evade (basic) GF-guns. Therefor the 'old' TargetingChallenge might be better to measure the performance in the early stages of development.
----

Hey Grubbm - I took a glance through your gun's source, just because I was curious. I didn't find anything like a bug in it, and I'll hold off on any tuning-type suggestions for now, because I see you are taking development one step at a time.

However, one thing stood out as somewhat "mathematically incorrect" in my mind: you use Math.asin(9 / bulletvelocity) instead of Math.asin(8 / bulletvelocity). If this is to give a buffer above the 1.0 mark for some reason, I would suggest instead using something like Math.asin(8 / bulletvelocity) * 1.2. The reason being that using 9 instead of 8 (their max speed) will not be as consistent with different bullet powers. For instance, using 9 instead of 8 results in about an extra 15% when bulletvelocity = 11, but an extra 17.6% at bulletvelocity = 14. A very small point, and it will only make a difference with varying bullet powers, but every bit of precision counts =) Best of luck with this 2K-destined bot. -- [[Voidious]]

Nice hint, i am doing this asin(9/velocity) in all my bots :) Thanks! --[[Krabb]]

Indeed I am using 9 instead of 8 for GF>1.0 reasons. I figured that 90% of the bullets would be of power 1.9, and I just needed something to fight an occasional index-exception. You are right though and I will change it. With my gun I still have lots of ground to cover, like taking into account the botwidth with hit-detection, firing a wave each tick, and ofcourse segmentation issues. I haven't even checked if my waves are aligned yet. I think I am just too old to comprehend everything at once, so I absorb this stuff with little bites because I want to know what I am doing and why. ;-)  -- [[GrubbmGait]]

Yeah, if you're using the same bullet power mostly, the effect will be negligible. It's more like the kind of thing that would bother me on a philosophical level ;) Also, I noticed that your gun code seemed like it might be choosing different bullet powers (like maybe in melee?). I think that the way you are developing it is quite wise, and that you will probably save yourself a lot of headaches in the long run. -- [[Voidious]]

If I had known that a GF-gun was this simple, I had never wasted my time with a PM-gun. Now that my rudimentary GF-gun outperforms GrubbmGrb (at least in the TargetingChallenge2K6), I must put some time in a WaveSurfing movement that can evade the simplest targetters.  -- [[GrubbmGait]]

Hmm, your original gun is doing far better then mine is so far, i'm trying to make it so that I can get a high rate against TheArtOfWar, however it seems that no matter how hard I try (though different bin sizes do show different preformace levels), I can't seem to get a relatively high hit rate dispite how high other guns can get. -- [[Chase-san]]

Your version 0.1.57 looks quite similar to my v0.0.2 (except for Fhqw). Don't forget that I (think I) have the gamephysics things right(like radar, events before run, fire before turngun) because I have two years experience with robocode. On the other hand, this is my first GF-gun so there still might be some buggy things in there.  -- [[GrubbmGait]]

I thought that using non-bullet waves would increase the score, apparantly I thought wrong. -- [[GrubbmGait]]

Summary:
Consciousness check: 4 + 10 =
This change is a minor edit. Send email notification that GresSuffurd/GresVretter has been changed.
(Visit Preferences to set your user name.)

View other revisions
Robo Home | GresSuffurd | Changes | Preferences | AllPages