Difference between revisions of "Scalar/Version History"
< Scalar
Jump to navigation
Jump to search
m |
m (Maybe the fastest normalRelativeAngle ever ;)) |
||
Line 26: | Line 26: | ||
* ? | * ? | ||
** Thanks [[Skilgannon]] for mentioning paper "Fast Polynomial Approximations to Sine and Cosine" in his code ;) I'll try some approximated function as the replacement of StrictMath. | ** Thanks [[Skilgannon]] for mentioning paper "Fast Polynomial Approximations to Sine and Cosine" in his code ;) I'll try some approximated function as the replacement of StrictMath. | ||
+ | ** Just benchmarked my legacy normalRelativeAngle (uses two %), robocode.Utils.normalRelativeAngle (branching), [[Skilgannon]]'s normalRelativeAngle (floor) and my new normalRelativeAngle (which is shorter and cleaner than Skilgannon's, and uses INVERSE_TWO_PI to get ride of divisions), and find my version a little faster than [[Skilgannon]]'s (5.4x faster than robocode's), and my legacy version (not used in my recent bots though) is 50% slower than the robocode's ;p | ||
* {{SimpleVersion|date=2017-11-02|version=0.01d}} | * {{SimpleVersion|date=2017-11-02|version=0.01d}} |
Revision as of 13:51, 4 November 2017
- Scalar Sub-pages:
- Scalar - Version History - Understanding Scalar
- Planned 0.011
Migrate ScalarBot’s 1v1 movement & gun- Migrate ScalarBot’s 1v1 gun.
- Rewrite a 1v1 movement that’s first optimized for simple guns, then slow learning guns. And find out a way to dodge pattern matchers well.
- After that, add a tick flattener to help dodging the top guns, and a firing wave flattener agaisnt AS guns. (Only after I can dodge anything else very well can I start working on this and set a better threshold).
- Planned 0.01f
- kNN melee gun
- Planned 0.01e
- Melee wave surfing
- ?
- Thanks Skilgannon for mentioning paper "Fast Polynomial Approximations to Sine and Cosine" in his code ;) I'll try some approximated function as the replacement of StrictMath.
- Just benchmarked my legacy normalRelativeAngle (uses two %), robocode.Utils.normalRelativeAngle (branching), Skilgannon's normalRelativeAngle (floor) and my new normalRelativeAngle (which is shorter and cleaner than Skilgannon's, and uses INVERSE_TWO_PI to get ride of divisions), and find my version a little faster than Skilgannon's (5.4x faster than robocode's), and my legacy version (not used in my recent bots though) is 50% slower than the robocode's ;p
- 0.01d – 2017-11-02
- MeleeRumble ‒ APS: 63.94% (25th), PL: 362-39 (39th), Survival: 26.62%
- Fixed a lot of bugs in PIF gun ;)
- Implemented melee wave tracking (no gun heat waves yet), and no reaction to that information yet.
- 0.01c – 2017-10-31
- Simple melee movement that dodges a lot of HOT bullets without surfing ;)
- Simple PIF gun that selects scans randomly