TormentedAngel
| TormentedAngel | |
 Tormented Angel | |
| Author(s) | lRem |
| Extends | AdvancedRobot |
| Targeting | GuessFactor Targeting |
| Movement | WaveSurfing |
| Released | November 16th, 2008 |
| Best Rating | 1v1: 1729.7 (228th) |
| Current Version | Unlimited |
| Code License | None |
| Download | |
Background
This is my first experiment with WaveSurfing. It has been done before I had a chance to read the excellent Wave Surfing Tutorial, so it included some WaveSuffering. But to be frank, that wasn't that bad. On the other hand I solved most problems in a different way, probably less efficient.
This bot is uncompetitive by design. It doesn't use any segmentation - it's just to check how much different implementation issues matter.
History
- Alpha - 1630.5 points
Doesn't care about which waves are associated with bullets that actually hit it. This makes it essentially a curve flattener. No wonder it scores around MakoHT.
- Beta - 1709.8 points
Pretty much the same code, but gives 10 times more attention to waves that carry a bullet that hit it. Big advance, improved against almost all reasonable bots. Some exceptions are Wolverine, Cigaret, FloodNano and Leopard. See the pattern? ;)
- Unlimited - 1706.6 points
Added Rolling Averages, exponential style with 1.4x decay every approximately 14 waves. Mixed blessing: gained against the strong and lost against the weak. This goes as far as +6.1% against Diamond, +7.4% against WaveSerpent, +7.5% against Dookious +9.3% against DrussGT and +10.9% against SilverSurfer. Despite the tiny rating drop I'm happy with the change.
- Arabian - 1729.7 points
Added Bin Smoothing and increased significantly the number of bins. A bug surfaced and sometimes the score is pretty close to 0, but I'm not sure whether it was introduced now or just evaded detection. Anyhow a nice overall increase, but problem bots got a bit more problematic. --Lrem 18:30, 20 May 2010 (UTC)
- Antiquitie - 1699.7 points
This version reached a limit before publishing! The limit of characters in version name - it lacks an s...
My own invention - instant prediction. Not exactly precise, but I hope it can be good enough. Surely it's much better than assuming I can always go 8 * escapeTime. But there seems to be an error. Ranking dropped a lot. Even despite the fact that it's winning quite good against older versions of itself.
