TassieDevils - Wompi/robocode-bots GitHub Wiki

The development diary of the TassieDevils. The official version history can be found at RoboWiki (TassieDevils). The diary is just to keep track of the thoughts and odds while programming on this twins. I guess it is not appropriate for the official wiki and therefore i link it to this repository.

Current Notes: Things to try

Date: 11/09/2012 Size: v1.7 1979

It is quite a while when i last changed the Tassies, and it is not a big change i made. I'm just curious if the new random component of the movement is as strong as i hoped. I made some changes to the combat setup as well and it should be now have a little more survivability. I hope it is still good against close combat teams. I got rid of the minion rule for 1vs2 (the minion got to close combat) because it was not good at all. I plan to bring the Tassies at least to top 5 and will do some major changes in the next days. Well lets see how this is working out.

Credit: nothing changed .. i feel how i get hooked again on this twin stuff.

Date: 06/06/2012 Size: v1.6 1996

Another day without sleep and i could fix some issues of the last version. It lost major score against some mid class bots just because the movement was wrong. So i got rid of the 'jiggle' stuff. Some minor changes to the movement if the bot goes for close combat and a different protect point for both bots. The minion now goes for protection to the last guessed point (collected by the gun) and the leader goes to the opposite site (so the enemy stays between those both). The code size is a little bit of a problem (mostly to bring in some stuff against the top teams) The overall forces got a little change (not sure about this). The distance gets more weight for the targeting now. I spotted some serious issues of the movement and this need some heavy debugging soon. Sometimes the call for help got no response and the bot drives to the opposite (cowardly) direction. Also somtimes the bot get stucked in the middle of the field (i guess the risk function reaches it local maximum .. escpecialy if the dodge variables are true). I played with lostCounts and gooing close combat if i loose to much but this wasn't realy successful. The gun has much trouble to even hit on a very close distance to some teams and this bugs me most. I think i should change the avgVeocity to something more dynamic to picture the velocity changes more precise.

Credit: nothing today
tiered but still excited.

Date: lost Size: v1.5 1888

This one got some nice improvements and i started to clean the code a little bit. For now both teammembers react on protect calls and help each other in serious situations. I changed the fire target distance by 100 to save some energy against the long shot teams. The movement got a little tweak by lowering the calcuated riskpoints and a sin move if the bot goes to the riskpoint. The targets are now hold by its own variables, this saved alot of code size but got uglier than ever. The radar got some love and loos now a little bit better. The protect react force is now 100k to make a faster engagement of the protected target. Nice new battlestate variable for clean code size and smooth decision making. The target is now not ony the closest it is rated by energy and distance. This means if the target is near it will be selected but if it is far away it depends on his energy state. The target distance is now correct caculated and if the target gets an update by team messages it is advanced by 1 step (roughly guessed). The team message only has the vaus of the last round and if the target shots blind it was always one step behind the target.
Rednexala has changed his Lunartwins and after this they have beaten me very badly. This should be fixed with this version. But i call it a success if the topleader changes his bots just because of me.

Credit: hmm
Still hooked and excited about this robocode stuff.

Date: lost Size: v1.4 1952

Just a more enhanced draft this time. I like the new team messages, they realy make the Tassies look like a team. Now they share every information on targets and give them to the other team member. Also new is the scan behavior. The leader takes his target (closest) and send a message to the minion. The minion now looks for the other target and scan lock it. Because every bot has the same information about the position, heading, avg velocity and so on every bot can shoot blind on his best target and dont have to loose the scan lock, very neat i guess :). I'm still disappointed about my gun. The gun is still useless against the 1v1 stallions but i dont want to switch to a more enhanced weapon until i didn't squieze the last bit of performance out of the 'simple' gun. So still no chance against wave teams but very nice performance against everyone else. The code is still draft status and could realy use some love. Right now i reached the 2000 byte barrier but there is still hugh code size left if proper coded.

Credit: i still try my own ideas
Man i'm so hooked into this twin stuff this is realy good clean fun.

Date: lost Size: v1.3 1300+

Well not much to say. I just include the source, so you can see my horrible programming draft. I realy would spend more time with this, but unfortunately do i have not much time these days. The tassie devils have a little more team spirit and the leader will be guarded at certain battle field states. I really would like to implement more TeamMessages for the enemy state and change the radar and and and ... but well no time. This one is a very rough draft so be gentle with me :). I know there are a lot of code fails and way to naiv functions but for a quick look it should be enough. I dont have the right feeling right now for the twin rumble but i'm sure this will come after more serious testing and implementing.

Credit: not much to say, its all based on Wallaby glued with naive team functions
.. off we go