Difference between revisions of "Talk:Moebius"

From Robowiki
Jump to navigation Jump to search
(Awesome results!)
 
Line 37: Line 37:
  
 
I discovered a serious bug in Pugio's distance controller. I forgot a pair of braces which made my bot closing forever and never retreating. I fixed that bug both in Pugio and Gladius, and now it seemes they gained 1-1,5% APS. ;) Start to prepare for the battle, Pugio is a happy member of 71+ % bots. ;) --[[User:Robar|HUNRobar]] 20:11, 12 June 2009 (UTC)
 
I discovered a serious bug in Pugio's distance controller. I forgot a pair of braces which made my bot closing forever and never retreating. I fixed that bug both in Pugio and Gladius, and now it seemes they gained 1-1,5% APS. ;) Start to prepare for the battle, Pugio is a happy member of 71+ % bots. ;) --[[User:Robar|HUNRobar]] 20:11, 12 June 2009 (UTC)
 +
: Sweet.  Welcome to the club.  I look very much forward to Pugio's challenge!  --[[User:Miked0801|Miked0801]] 23:54, 12 June 2009 (UTC)

Latest revision as of 00:54, 13 June 2009

Wow, you made that look easy. =) No updates for years and then 100 points in one jump. Good luck with those last few points between you and the NanoBot throne! --Voidious 17:53, 12 May 2009 (UTC)

My changes were rather lame. I looked at the top 5 bot's source codes and mismashed things around a bit and got a bot that's about 90% NeophytePattern 1.0 and 10% my own code. I'll be making another pass through it in the next few days to see if I can make it more unique more effective. The only justification I have is that I completely understand every line of code and why it was done - and that Nanos have always been about borrowing innovations from each other and giving credit where credit is due. --Miked0801 22:26, 12 May 2009 (UTC)

Oh no! I think I am needed to work my OcnirpSNG again. I think NeophytePattern is the easiest of Neophyte series (Id o understand it, along with NeophytePRAL, but I never understand NeophyteSRAL) Yes, nano always beem about borrowing each other code, sometimes without credit. The best gun in nano is pattern matcher (if you don't count PRAL and SRAL), since since Simonton do reduce the codesize most, now every top seven nanobot, exclude PRAL and SRAL, do use Simonton's pattern matcher gun Yes, NeophytePattern borrowed WeekendObsession gun. I too borrowed it. Robar first borrow FunckyChicken, but later replace by WeekendObsession's since it smaller.

Taking for your version history: "I'm just taking tem and playing with them a bit to see if I can improve them or find any extra code space." :I bet you can't find any extra spaces. Neophyte series are super optimized for codesize. If you want to improve it, you may want to borrow BlackWidow instead. It has a LOT of room left (it contains 15-byte color too) » Nat | Talk » 23:42, 12 May 2009 (UTC)

I made a pass through a stelo bot that had a nano-sized wave matcher and almost peed myself. I'm trying to compress that code now. --Miked0801 03:11, 13 May 2009 (UTC)

BTW, it is because nanos are almost all open source that they have progressed as fast as they have in skill. You should have seen how lame the original bots were way back when. Somewhat linear aim with crappy radar and wall death common. Then Infinite radar, then perfect linear, then rolling average linear, then circular-like linear, then pattern matcher. The latest bots are top 30% bots. That is absolutely amazing to me. --Miked0801 03:16, 13 May 2009 (UTC)

Hi, I see the rank of your new version is like my bot's: a lot better against top bots but lose some % against the rest. Unfortunately, I can't watch your robot now or look at the code. :( --HUNRobar 07:07, 18 May 2009 (UTC)

I'm putting source code in with my bot, but for some reason, it doesn't allow you to grab it always. The source is also available at the robocode robot repository. All my bots are open source - I have nothing to hide :) --Miked0801 15:00, 19 May 2009 (UTC)

FretNano outcome

Well, this seems to say FretNano uses a "LastSuccessfulGuessAngleTargeting" gun. Given that you use stop-n-go style movement... it kind of makes sense that would hit you very well and also adapt almost instantly when you change from back-and-forth to orbital too. --Rednaxela 12:25, 19 May 2009 (UTC)

Yep - The thing is, that bot has a massive movement weakness against 2.5 power guns. It gets locked in place against walls and gets destroyed - hence other bots that shoot at 2.5 get a big score against it. I'm using 3 power because, well because I want to :) --Miked0801 15:00, 19 May 2009 (UTC)

Champion

Congrats dude! --Voidious 14:17, 20 May 2009 (UTC)

Oi! That push me down to 3rd PL =( Anyway, OcnirpSNG is two from seven of your lost list! » Nat | Talk » 14:43, 20 May 2009 (UTC)

I wasn't surprised when I saw Moebius at 1# rank. I haven't examined the code so I don't know how it works but it's totally absolutely amazing. :) I would faint if you reached the 2000 ELO. ;) --HUNRobar 17:45, 20 May 2009 (UTC)

You should look - the original string matching algorithm that everyone's been stealing back and forth has a bug in it that I fixed. Basically, it wasn't taking into account the 1 tic lag between gun movement and firing. That, plus the fact that robots tend to be further away than my 'desired' distance means that a simple +2 to the number of angular velocities to add made my gun a ton more accurate. That's about 3 bytes of code that all nano PM bots could add to get a nice boost in performance. If I had another 3 bytes, I could make it even more accurate - but that will be for a future revision. --Miked0801 18:13, 20 May 2009 (UTC)
Uhh I thought your new movement idea brought you the high ranks. Is it under development or simply trashed?
(edit) I took a look at your source. Your gun is inaccurate because you use a pre-defined distance divisor instead of e.distance/bullet velocity. I tried it but it was so bad that I even replaced distancing with a fixed approaching anno. --HUNRobar 19:03, 20 May 2009 (UTC)
My new movement idea would work fine, but it requires way, way too much work to get right. The reason my predefined devisor works is because my movement code puts me at that range fairly quickly - though it tends to lag me outside of it a touch - hence the addition of 2 ticks to the divisor for more accuracy. That plus increasing the chance of a less than full speed movement length (by 10% or so) is what actually vaulted me up in the ranks. I'm going to go back to a orbit movement with a better gun for a bit in a new bot to see if I can top where this one is at. --Miked0801 20:39, 20 May 2009 (UTC)

Ok, the latest version keeps me slightly further from the target and moves a touch more random. My local testing showed a 1% or so APS, but in the rumble, I'm faring about .2% lower with a better win percentage against some of the tougher bots. So, is it better to beat more bots, or score higher? Dunno. I've got another version where I stay closer and aim less ahead that local testing showed to be roughly equal in gain. Perhaps that one will do better overall. Apparently my 20 bot test bed wasn't quite a good representation of the pool as I thought...

I discovered a serious bug in Pugio's distance controller. I forgot a pair of braces which made my bot closing forever and never retreating. I fixed that bug both in Pugio and Gladius, and now it seemes they gained 1-1,5% APS. ;) Start to prepare for the battle, Pugio is a happy member of 71+ % bots. ;) --HUNRobar 20:11, 12 June 2009 (UTC)

Sweet. Welcome to the club. I look very much forward to Pugio's challenge! --Miked0801 23:54, 12 June 2009 (UTC)