Horizon/GunResearch

A page for my tests with Horizon's gun.

---

TargetingChallengeRM

Bot	Author	Gun	Aspd	Sprw	Fhqw	Yngw	FlMn	EASY	Tron	HTTC	RnMB	DlMc	Grbb	MEDIUM	SnDT	Cgrt	Frtn	WkOb	RkMc	HARD	TOTAL	Comments
0000 (Old)	AaronR	DC/GF	73.67	81.50	57.06	91.29	77.88	76.28	73.31	79.82	73.18	83.82	72.42	76.51	55.52	72.41	67.14	63.55	61.56	64.04	72.28	15.0 seasons
0000	AaronR	DC/GF	69.60	76.51	54.32	88.43	74.28	72.63	73.08	70.11	77.54	82.93	71.93	75.12	52.01	65.48	59.27	59.65	58.90	59.06	68.94	50.0 seasons
0001	AaronR	DC/GF	71.32	69.39	48.53	83.41	72.53	69.04	74.25	68.35	73.88	84.64	71.54	74.53	46.38	65.44	53.27	53.06	54.46	54.52	66.03	50.0 seasons
0002	AaronR	DC/GF	78.32	95.63	85.41	95.46	87.66	88.50	81.36	77.66	85.95	85.87	75.19	81.21	62.64	71.56	68.64	72.71	71.57	69.42	79.71	100.0 seasons
0003	AaronR	DC/GF	78.49	95.72	87.31	95.71	87.72	88.99	81.85	78.28	86.88	86.11	76.79	81.98	63.39	71.14	71.02	74.22	71.71	70.30	80.42	50.0 seasons
0004	AaronR	DC/GF	76.24	96.15	84.98	95.79	86.71	87.97	81.27	77.59	85.31	85.64	76.29	81.22	62.09	71.19	68.10	74.08	72.38	69.57	79.59	50.0 seasons
0005	AaronR	DC/GF	80.08	95.66	87.58	96.62	87.42	89.47	83.22	78.67	87.76	88.10	76.94	82.94	60.95	71.30	68.28	71.69	69.88	68.42	80.28	50.0 seasons
Bot	Author	Gun	Aspd	Sprw	Fhqw	Yngw	FlMn	EASY	Tron	HTTC	RnMB	DlMc	Grbb	MEDIUM	SnDT	Cgrt	Frtn	WkOb	RkMc	HARD	TOTAL	Comments
DCResearch 0071	Simonton	DC/GF	93.68	98.78	97.28	97.79	94.60	96.43	88.65	88.85	89.90	86.95	84.96	87.86	82.45	84.84	81.94	83.88	81.99	83.02	89.10	105.0 seasons
WaveSerpent 1.4	Kev	DC/GF	95.32	98.42	96.55	98.23	93.69	96.44	89.74	88.27	89.29	88.52	85.31	88.23	82.30	84.75	80.75	85.03	80.01	82.57	89.08	100 seasons
Phoenix 0.9DC17	David Alves	DC	93.63	97.30	95.36	97.65	91.75	95.14	88.28	86.97	88.47	89.38	85.99	87.82	77.91	79.60	83.48	81.10	84.01	81.22	88.06	155.0 seasons
Lukious 1.20	Voidious	DC	90.47	96.27	94.54	96.48	92.91	94.13	87.38	86.65	90.23	86.21	83.38	86.77	81.80	82.25	77.83	81.15	75.13	79.63	86.84	200 seasons

• 0000 (Old): Just Horizon 1.0beta11's gun with no modifications. Buggy! Faulty wave-intercept calculation!
• 0000: Just Horizon 1.0betaF3's gun with no modifications. Buggy! Faulty wave-intercept calculation!
• 0001: Trying out a much larger cluster size (30). Buggy! Faulty wave-intercept calculation!
• 0002: Just Horizon 1.0betaF4's gun with no modifications. These battles were run for me by Simonton after he found the bug.
• 0003: Playing around with the clustering weights; should rate lateral velocity much higher.
• 0004: Playing around with the clustering weights; rates time since acceleration as high as lateral velocity.
• 0005: Revert to 0003 weights; using sqrt(log size) as the cluster size.

---

Note how I have already proven my gun is a piece of junk. -- AaronR

Challenge complete, took 5 hrs 41 min 50 sec

... running with two threads on a dual-core processor. Man, I really need to upgrade Horizon to use Simonton's tree. -- AaronR

• These battles take 15-20 seconds each on my machine using DCResearch. It may be well worth your time to plug in the tree. Is this really the time to run 15 seasons of fast learning?? -- Simonton
  • Yes... and the battles take 2:30 - 3:15 each with Horizon. =( Of course, part of the problem is that my bot loses a higher percentage of the rounds, meaning there is more idle time than with a better gun like DCResearch. -- AaronR

I'll have 50 seasons of the new 0000 ready later. Using the tree really is much, much faster. It seems like there is a bug somewhere, though, since just about every one of the new version's scores is lower than the older release's. -- AaronR

Challenge complete, took 5 hrs 22 min 47 sec

... running with one thread on a single-core processor, over more than three times the number of seasons as before. I'm happy now. -- AaronR

You should send me your gun code. Maybe I can find what's holding you back. -- Simonton

• Thanks. I'll upload with source the TC version of the bot to my "tests" slot on the repository; I was just about to remove 1.0betaF3a from the rumble anyway. Chances are, the main bug is something really, really obvious, but that I haven't noticed because it's in the parts of the code I haven't looked at since the first release. =P Edit: the uploaded version is basically 0000, not 0001, so it has the smaller cluster size. -- AaronR
• Can you email it to me? As usual, the repository is down. -- Simonton
• Sent. You know, this is getting really irritating... -- AaronR