[Home]GresSuffurd/GresVretter

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

version history (newest entry on top)

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?)

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

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

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

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

Robo Home | GresSuffurd | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited May 23, 2007 11:19 EST by GrubbmGait (diff)
Search: