Well, it was made one day when Kawigi was talking to Iiley over MSN IM about the pre-release (and yet unnamed) version of BlestPain. I toyed with sticking FloodMini's gun into Cigaret in place of the Wave pattern-matching, and found that it was never far behind SandboxDT 1.91 in score. What's really special about it is how it fights against SandboxDT 2.01.
https://www.robocoderepository.com/BotDetail.jsp?id=1676
It's relatively untested, actually. But from the tests I've done, it seems to be able to beat just about any bot given about 300-500 rounds. This is what you've all been waiting for, though. No saved data on either side:
1st: wiki.mini.Sedan 83189 22700 4540 49918 6031 0 0 455 546 0 2nd: pe.SandboxDT 2.01 73419 27300 5460 36036 4622 0 0 547 454 0And with the saved data from the previous round:
1st: wiki.mini.Sedan 81893 22750 4550 48510 6081 1 0 456 545 0 2nd: pe.SandboxDT 2.01 73709 27250 5450 36357 4648 3 0 545 455 0
It uses Cigaret 1.31's superb movement. Basically it randomly picks a point to go to when it's fired at and goes there, and once it's there, it waits for its opponent to fire again. It's not much more complex than that.
It uses constantly-compiled stats (sacrificing accuracy against some bots for learning speed) segmented on distance of projected movement to the wall, acceleration, LateralVelocity, and projected bullet travel time. This is the same gun as is used by FloodMini 1.4.
Moves randomly when it's fired at.
Non-existance vs. proficiency.
Arbitrarily.
Saves its stat buffer between rounds and matches, saves arbitrary variables between battles.
When Iiley was telling me about this new MegaBot he was building, and sending me scores against SandboxDT, the dev version didn't have a finalized name yet, and it was called "Van". The day I proposed to Iiley that we really release this bot, he had just figured out what "Van" means in English (in his words, "a big car"). Since BlestPain and this bot (which was named "CigarQwhgads?" at the time) were similar bots conceptually, I suggested naming it after a smaller car, possibly "coupe" (which is usually a small, 2-door, sporty car), or "Sedan" (its eventual name). Iiley knew "Sedan" as a town in France, I explained to him that a Sedan is a medium-sized car, usually with 4 doors. He seemed to find the prospect of a robot named after a "car with 4 doors" after he accidentally named his dev robot after "a big car" extremely hilarious. So we settled on that.
If I said no, would it be hypocritical? In all fairness, this is already Iiley's code with Kawigi's code copied and pasted into it, so you might as well use Cigaret's and FloodMini's code. But do ask permission if you use a large amount of either of our code, and give credit where credit is due. And read the KPL?.
Maybe update it when either of its inspiring robots gets updated? Win every MiniBot contest that will ever exist from now on (is that too optimistic?)?
Maybe the old SandboxDT.
Cigaret 1.31 and FloodMini 1.4
After having gathered data from around 3000 battles, this is the result of a 5000-battle run:
1st: wiki.mini.Sedan 1.0 403410 107550 21510 245239 29108 2 0 2157 2849 0 2nd: pe.SandboxDT 2.01 381378 142450 28490 185743 24689 6 0 2855 2151 0
Using the formula from IsTheBotBetterOrWorse, we get ((w-l)/r)/(120/sqrt(r))=2.59 ... This means we're more than 95% sure Sedan is a better bot than SandboxDT! The king is dead, long live the king.
We're really in need of EternalRumble or RoboRumble@Home to put this in a prespective ...
-- FnH
We'll see what Iiley does with this, too, he's got the ball now, and he may even be able to improve the movement. Keep in mind, also that this is a MiniBot. With space to spare (1451 bytes currently). Tobe proposed the MiniBotChallenge with the question, "How good can you make your robot if it is restricted in terms of code-size?" Mine and Iiley's answer to that is "pretty friggin' good". DT is around 24000-25000 bytes.
What's more significant about these results is that they show that whatever Paul Evans did with his movement doesn't work well enough, because it assumes too much about his opponent's targeting system. Not that it's a bad idea, of course, just that his old movement was better. -- Kawigi
We're a few 2000-round tests later, and I can say with more than 95% certainty: Sedan beats SandboxDT 2.01, SandboxDT 2.01 beats SandboxDT 1.91, and SandboxDT 1.91 beats Sedan ... looks like rock, paper, siscors to me :). Untill we get EternalRumble back or finish RoboRumble@Home, we in Robocode-land will be ruled by a triumvirate instead of a king I guess ... -- FnH
...And BlestPain beat DT 1.91 and lost to DT 2.01. I'm not sure, but I think Iiley said Sedan loses to BlestPain, but he speculated that given his latest movement tweak, Sedan might still have the edge. -- Kawigi
I tweaked the movement of Sedan yesterday with the experience of BlestPain,but it did worse,I think this is why BlestPain beat SandboxDT 1.91 but lost to SandboxDT 2.01,if i return BlestPain's movement to preversion,it maybe can beat SandboxDT 2.01,but then lost to SandboxDT 1.91,this seems like another rock.~;],however i will not give it up,i will try more,when i figured it out,next version of Sedan will born.btw,Sedan has many codesize space to improve,i squeezed it yesterday to about 13XX. -- iiley
Using some tricks from FhqwhgadsMicro, I may be able to save us another 30-50 bytes on top of that. -- Kawigi
5k rounds:
1st: pe.SandboxDT 2.11 422655 168700 33740 191442 28773 0 0 3385 1625 0 2nd: wiki.mini.Sedan 1.0 335436 81250 16250 216181 21754 0 0 1633 3374 0-- Kuuran