Archived talk:Jen 20090527

From Robowiki
Jump to navigation Jump to search
Jen Sub-pages:
JenVersion History - Code - Archived Talk 20090527
       Archive        This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.     

A MicroBot duelist from Voidious. I'll make a proper page tomorrow. No idea where this will rank, but I'd been tinkering a bit a couple weeks ago, so I may as well just release an initial version and start from there... -- Voidious

I was wondering when this was going to happen. I also saw a PhoenixMicro on the page.... =) -- Skilgannon

Time will tell if I actually put much focus on it. I didn't see much appeal in making another StopAndGo / RandomMovement MicroBot, but I figured it would take two seconds to connect Komarious' gun to the decent RandomMovement from a TwinDuel team I was tinkering with. Now that I see it so far behind in the rankings, I almost can't help myself from going for a better rating, though. =) -- Voidious

Seems that you got hooked even at micro's. When can we expect your first nano? Note that you rating is better than the #2 of a year ago (which is now at #10!) -- GrubbmGait

  • Yes, I am quite obsessed! =) Before making a MicroBot, I thought, "well, they're kind of all the same - a little StopAndGo, a simple GF gun, and some RandomMovement == 1950+ rating". Then I did that and was still 100 points behind the top bots. It turns out that while less CodeSize might mean less features in an individual bot, it can also mean more possible combinations because adding one feature means removing another. I had similar revelations when working on Komarious, but it's been a while. Not sure when I'll post a NanoBot, but it's definitely on my mind. =) -- Voidious
  • The most impressive ranking of the micros, IMO, is RaikoMicro. He's still using the MusashiTrick! And I'm surprised that Connavar isn't getting a much higher score than Waylander, there must be a lot of simple movements in the microrumble that simple GF guns can't quite pick up on, but PM guns can. -- Skilgannon

Hmmm... have you tried FloodGrapher on your movement? I tuned my movement for 10000 rounds against Aristocles....after getting it in the ballpark with FloodGrapher. A slightly higher reverse chance helped against Aristocles (not much, just a constant of 0.02 added to the one side). I think it confuses the time-since-decel segment a bit better. -- Skilgannon

  • No, but that's a good idea. I haven't done any movement graphing since the very earliest days of Dookious! I used StatistRobot then, I think, but maybe I'll give FloodGrapher a try. -- Voidious

Nice jump for just tweaks! Are your "Won't stop" and "different distancing" working off of the same thresholds? I use a function for the attack/retreat angle in Waylander, which I find to be sufficient, but Waylander will run away (ie. don't reverse from the flattener or stop for the StopAndGo) if the enemy is closer than 100. This also helps against RamBots. -- Skilgannon

  • Thanks! IIRC, I am not-stopping for StopAndGo within 100 and using a bigger (er, more negative) attack angle within 150. That first version (which lost tons of points) used 250 and 300 or something, which was way too much. I might toy with an attack angle function, like you said, but for all my distancing / attack angle tweaks in Komarious, I still ended up with simply using the same negative attack angle at all times! So I'm not optimistic about that gaining much. (But clearly MiniBots and MicroBots are more different than I originally thought, so who knows...) -- Voidious
  • Another reason I'm using a function is because it's cheaper in codesize...its a sort of k + distance/m function, nothing fancy =) -- Skilgannon

Wow... Waiting until gun is aimed before firing is worth 7 points for Komarious but -14 points for Jen? Unless I mixed up the source code for 1.087 somewhere, which I can test with a re-release. If these data are right, I guess the conclusion is: if your gun isn't very accurate, you're better off just firing as often as possible instead of waiting an extra tick or two until you're aimed correctly. -- Voidious

I would say that the point difference is due to the different movements. I've wondered often whether I would be better off firing less often (never mind less accurately) when I'm getting more point from survival, and losing more energy due to my own firing instead of to enemy bullet hits. Komarious is able to dodge better, so gets more points from survival, and thus doesn't depend on bullet hit points as much. In fact, those few rounds where Komarious does lose are more due to running out of energy from missing, than being pounded into submission. Jen is exactly the opposite: it has to hit the other bot A LOT to be able to keep its energy higher than the other bot, because it loses much more energy from being hit. It's all a fight for those survival points. At least, that's my speculation =) -- Skilgannon

There are no threads on this page yet.