Difference between revisions of "Geomancy"

From Robowiki
Jump to navigation Jump to search
m (Geomancy 11)
m (Geomancy 12)
Line 12: Line 12:
| released        = 19 June 2009
| released        = 19 June 2009
| best_rating    = 59{{sups|th}} with G2-1937.0.
| best_rating    = 59{{sups|th}} with G2-1937.0.
| rating          = [http://darkcanuck.net/rumble/RatingsDetails?game=roborumble&name=synapse.Geomancy+11 RoboRumble]
| rating          = [http://darkcanuck.net/rumble/RatingsDetails?game=roborumble&name=synapse.Geomancy+12 RoboRumble]
| current_version = 10
| current_version = 12
| download_link  = http://synapticflame.googlepages.com/synapse.Geomancy_11.jar
| download_link  = http://synapticflame.googlepages.com/synapse.Geomancy_12.jar
| isOneOnOne      = true
| isOneOnOne      = true
| isMelee        = false
| isMelee        = false

Revision as of 13:38, 25 September 2009

Pages: Main page · Code · Version History · Discussion

Author(s) Synapse
Extends AdvancedRobot
Targeting AS GF
Movement AS WS
Released 19 June 2009
Best Rating 59th with G2-1937.0.
Current Rating RoboRumble
Current Version 12

What's special about Geomancy?

It's a 1v1 bot based on Watermelon. Geomancy takes a relatively new approach to the Segmentation problem: it keeps every possible segmentation updated, and just uses the one with the most specific suggestion for each scenario. This lets it learn very quickly and still have improvement over the course of a match. It also has debug graphics enabled, which may help you understand why it does the things it does.

What does the name mean?

Geomancy is an ancient divination mechanism that used self-similarity to generate random numbers from an impulsively created data set. It's mathematically interesting and has some cool patterns. More info: Geomancy on Wikipedia.

How does it move?

Autoselected segmentation of wave surfing visit count stats. Bot width is calculated precisely, but bot movement is not accounted for. It also has a partial flattener which helps performance against pattern matchers.

How does it fire?

Autoselected segmentation of traditional guessfactor targeting data.

What does it look like?

Waves are labelled with the axes used. Bins are marked as arc segments - brighter bins are fuller. Gray circles are full forward and full reverse positions that can be reached before the nearest wave arrives.

Geomancy 5 debug graphics