0.3.7 (codesize ???)
- Now still does stop-and-go for those last bullets that are less than 0.1, but doesn't detect them otherwise
- Since I had the spare codesize I added in a pattern-length-limiter that will allow 35+ round battles without slowing down
0.3.6 (codesize 724)
- Code squeeze - did you know that putting objects (rather than doubles, or ints) as the first declared variables will make subsequent access cheaper?
- Enemy energy tracking from onBulletHit? now uses rules class and bullet power instead of just subtracting 10 (damage from standard power 2 bullets)
- Made the attack/retreat angles slightly less steep.
0.3.5 (codesize 748)
- Shoot with a combination of PM and CT for the first round. Hopefully this will make me lose one of the first 3 games if the enemy can hit stop and go.
0.3.4 (codesize 738)
- Fixed the variable-distance StopAndGo 'tuning'
0.3.3 (codesize 738)
- Some refactoring
- Aim from the position I'll be at next tick (learned from my experience with wavesurfers)
- Tuned the variable-distance StopAndGo
0.3.2 (codesize 723)
- Back to standard anti-rambot code
- Only enable flattener if I lose in the first 3 rounds
- Shrunk some code
0.3.1 (codesize 731) - the 'cleanup' version
- Pattern-matcher only uses a 'key' that fits in this round
- Keeps up to 30k ticks of data (from 15k)
- Doesn't add data in first scan (to get a real DeltaHeading?)
- Uses my own flattener - finally!
- accidently used the untested anti-rambot code
0.3 (codesize 700)
- 1.1.17 with all extras removed:
- max retreating angle
- max advancing angle
- precise energy management
- different wall bounce for StopAndGo
- Default distance set to 600
- Need new things to do with the codesize - ideas guys? =)
0.2 - multiple choice - dropped
0.1.16 (codesize 746)
- less wallbounce for StopAndGo
- max retreating angle of PI/24 (7.5 degrees) when using StopAndGo
- hopefully the last release in the 0.1.x line
0.1.15 (codesize 737)
- Fixed StopAndGo (it didn't travel the full distance in 0.1.14, codeshrinking again)
- Simpler wall bouncing
- Max retreating angle of PI/12 (15 degrees) when using StopAndGo
- Less conservative with energy in endgame
- Non-exact enemyEnergy tracking
- Doesn't use rules class
0.1.14 (codesize 748)
- A revert to 0.1.9 (ie. no Rules class)
- Fixed bug that deleted most recent data instead of oldest data
- Better anti-ramming code
- Always moves away if time < 20
0.1.13 (codesize 705) --- BUGGY!!! SHOOTS HOT AGAINST StopAndGo !!!
- base firepower always 2
- removed many (small) features that weren't helping
- fixed bug that deleted most recent data instead of oldest data
0.1.12 (codesize 745)
- Firepower is 900/distance
- Moves directly away from rammers
- Uses normal distancing for StopAndGo
- Increased BEST_DISTANCE to 500
- Rules class seems to be safe
0.1.11 (codesize 748)
- Firepower is 3 only for distance < 150
- Another release with the Rules class
0.1.10 (codesize 744)
- Always retreats (slightly) with StopAndGo
- Distance based firepower
- Uses Rules class to calculate several variables
0.1.9 (codesize 746)
- Reversed the direction data is stored (saved 30 codesize)
- Moves if time < 20 (hope to escape bad situations) even with StopAndGo
- Made enemyEnergy tracking work for all bullet powers
- Improved Wall bouncing
0.1.8 (codesize 744)
0.1.7 (codesize: 739)
- Removed BitShifting?
- Added smooth distancing
- Added variable distance StopAndGo
- Removed 'match only from 3 ticks before fire'
0.1.6 (codesize: 726)
- Removed variable distance StopAndGo
- Removed smooth distancing
- Kept BitShifting?
0.1.5 (codesize: 744)
- Added BitShifting?
- Added smooth distancing (from Decado)
- Varied distance of StopAndGo depending on power of enemy bullet
0.1.4 (codesize: 749)
- Decreased number of possible values for angular velocity
- Decreased minimum match length
0.1.3 (codesize: 745)
- Reverted to non-multiple choice (too many bug opportunities, not enough codesize for comfort)
- Added StopAndGo
0.2.3 (codesize: 746)
- Fixed the right bug this time
- Rating is still below 0.1
0.2.2 (codesize: 748)
- An attempted bugfix that fixed the wrong bug
0.2.1 (codesize: 748)
- Better energy management
- Contained a fatal bug, dropped quickly
0.2 (codesize: 747)
0.1.1, 0.1.2
- Unknown, both decreased ratings, reverted to 0.1
0.1 (codesize: 725)
In the spirit of open-source-ness and wiki-friendly-competition-ness - would you care to fill us in? It looks like Waylander took a flying leap to the top of the micro rumble with the last version or two. -- Simonton
Will do...its kinda late right now but I'll fill it in tomorrow.
The basic changes were changing the angle 'scaling' from 128 down to 24, decreasing the minimum match length from 8 to 2 (I'd rather use pattern data than CircularTargeting) and adding StopAndGo. My latest (0.1.5) changes the distance traveled in the StopAndGo depending on the power they fired with, and adds a favourite distance of 400 (this is the number that worked best for Decado). I also started using bit shifting to put velocity and angular velocity in one character, it seemed to speed up the matching and decreased the codesize as well. I may now be able to take out my 'match only when about to fire' if statement, adding room for better wall bouncing. But first to see what the rating stabilises at =). -- Skilgannon
Oh yes, just in case you're wondering, it is OpenSource. With the same restrictions as Decado (I'll get around to making a decent page tomorrow - it's 22:12 here and I've got school tomorrow). -- Skilgannon
Version 0.1.10 does not do quite well, I see at least 3 very strange results (after 210 battles). cjk.Merkava, areb.Union and jaw.KarenCain?. -- GrubbmGait
It sounds like you weren't listening to the recent talk about PEZ rule #1 (.3) - don't change gun & movement in the same release. But seriously, you're making me nervous here. -- Simonton
I know, I was the one that said that, but I couldn't wait =). Anyways, I don't consider it to be much of a movement change. When I plotted the functions of the Raiko flattener and my flattener, they are virtually on top of each other. So for all my tweaking and rounds against FloodMini, I got a 1 byte smaller version of the same flattener.
With all that extra codesize in 0.3, I decided to fix all those microbot-isms that have never seemed important enough to implement. Together (hopefully) they will contribute 8+ points to my score, enough to get me #1. Just maybe. -- Skilgannon
- Well, I have 4 more bytes to shed in WeeklongObsession to fix the major bug in 3.1.x, so I'm still considering myself in the race :). -- Simonton