User talk:Miked0801

From Robowiki
Revision as of 16:57, 4 June 2011 by Voidious (talk | contribs) (reply)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Hey! Do you have to remove the "Inactive Authors" category now? =) --Voidious 17:31, 28 April 2009 (UTC)

Lol - I think I already did. Gotta upload an Infinty page now that I fixed a minor bug with him.

Hey man, I happened to randomly notice that an old version of BrokenSword is appearing in the melee rumble. Did you point your client at the new RoboRumble/Participants/Melee page? Actually I don't know how that would happen anyway, unless it uses the last participants list it saw if it can't find the list. Or maybe you just had some cached / unuploaded results from the last time you ran it? --Voidious 20:06, 28 April 2009 (UTC)

Good call. Just fixed that. Sorry for the out of date files being added. I'm running the new link now. Someone's going to have to clean up my mess on that one. --Miked0801

Just to let you know, I've noticed you signing with --[[User:Miked0801|Miked0801]] or --[[User:Miked0801]]. Just to let you know, on the new wiki you can simply sign with --~~~~ (or the signature button if you have js enabled), and it'll automatically turn it to a complete signature including timestamp even. Just a tip :) --Rednaxela 17:01, 29 April 2009 (UTC)

Yep, someone else let me know that too. :) --Miked0801 20:36, 29 April 2009 (UTC)

Oh hey, speaking of code shrinking, I just thought I'd point out the new "Rules" class if you hadn't seen it. I resisted using it for the longest time to ensure compatibility with old Robocodes, but it's been in there for a while and we now have a way of filtering out old clients anyway. So now I'm compiling with Jikes and Robocode 1.4.9, so I can use Rules but still compile with Jikes. --Voidious 14:21, 7 May 2009 (UTC)

Alright, what is this rules class you speak of? --Miked0801 17:54, 7 May 2009 (UTC)

The robocode.Rules class can be used to calculate the max turn rate, bullet power, bullet damage and so on. It is the one use by game engine, and it is now public. » Nat | Talk » 18:12, 7 May 2009 (UTC)

(edit conflict) It's a utilities class that provides methods for a bunch of basic stuff, like bullet speed, damage, turn rate, etc. Rules class in Robocode API. It's definitely an unfair advantage over legacy bots, but that's your decision to grapple with. =) --Voidious 18:16, 7 May 2009 (UTC)

I checked it out and none of its functions would have use in DustBunny nor Infinity, though bullet speed would become useful for linear leading if I stored my firepower instead of just calculating in place. --Miked0801 20:33, 7 May 2009 (UTC)

Your client seems to be uploading 1-on-1 results for retired bots? stelo.PastFuture 1.0, for example. This is easy to fix but if there is a bigger problem going on... --Darkcanuck 03:55, 8 May 2009 (UTC)

Perhaps I've got the wrong page pointed at (again) for rumble. I'll check when I get home. --Miked0801 20:27, 8 May 2009 (UTC)

Hi, it's good to see a 'veteran' coming back and working on nanos. :D I have been on rest since the beginning of April, but now I think I should do something to get back the top 3 rank. ;) --HUNRobar 16:02, 13 May 2009 (UTC)

First Place in Melee and 1v1 are sooo close. I got back into this again because I've got some late night hours free with a newborn - just like last time I got into the fun. Come on, just 1% in each. There's gotta be something I can tweak to get that right? :) --Miked0801 18:49, 13 May 2009 (UTC)

Just watch out because I'm going to make an update on my top bots. :) See Pugio 1.3 for example. ;) --HUNRobar 19:13, 13 May 2009 (UTC)

Wow, nice update. Game on! --Miked0801 20:28, 13 May 2009 (UTC)

Are we allowed to use: Vector<Double> vect = new Vector<Double>();

type things and set the "-source 1.5" flag for our bots? If so, that will save some codesize for me.

Yep, Robocode has required Java 5 for a while now, so that should be fine. I use generics in Diamond and it definitely ran fine in my 1.5.4 RoboRumble clients, which is the oldest version allowed for the rumble. So that saves code size, eh? Hmm, thanks for the info... =) --Voidious 01:12, 27 May 2009 (UTC)

Oh yeah, but I forgot, Jikes only supports Java 1.4, that's why I never used generics in my MiniBots. But I'll definitely compare it to normal javac with generics now that I know this. --Voidious 01:15, 27 May 2009 (UTC)
Sorry for triple post, but what makes you say it will save you code size? I tried changing Komarious to use generics for its one collection, and it had no effect on code size. --Voidious 01:23, 27 May 2009 (UTC)
Lol, I just realized I never ran codesize on my micro with generics in place. Doh! Well, generics look cleaner if anything else :) --Miked0801 02:25, 27 May 2009 (UTC)
Generic only available in source, it will ease out at compile time and every casting will come back (try decompile the generic-ize class and see) The generic only to check the type sefty. This is from java specification. And because the codesize library check the byte code, not the not the source code so this has no effect. And this is not require Java 5 to run since it isn't in the byte code, ad long as you only use the classes from the target version.

While scanning my local harddrive, I found a newer version of Radnor.DoctorBob (1.50). Any objections to me entering this into the fray and seeing how well it does in tandem with 1.42?

And removed. Failed experiment there.

Does Kawigi's old floodGrapher stuff still work with the latest versions of Java for anyone else? I get tons of Javax.Swing exceptions when I try to view a graph. It looks like swing has be deprecrated. Is this so?

I've never tried FloodGrapher, but I'm almost certain swing is not deprecated, even in Java 7 beta versions, but may something changed that requires old code to work properly. --zyx 02:05, 9 June 2009 (UTC)

I use FloodGrapher myself. I think you are running Windows Vista then, because if you are running it as normal user, it will throw a lot of exceptions and generate NaN hit rate. You need to run it as administrator. I run with both Java 5 and Java 6 and it all work. If you are loading result from FloodGrapger saved file (not FloodMini saved file), you will need to add robocode.jar to classpath yourself. Andway, I bet PEZ's TGrapgher (aka FloodGrapher 1.0) is much more easier to use. » Nat | Talk » 02:44, 9 June 2009 (UTC)

I'm using XP Sp2 and while the program opens fine, it fails with all sorts of Swing exceptions for classes not found. No biggie as I got Pez's utility to work. Anyone know exactly what his FirePower/Distance values are that he decided to use A-E on? I'm guess 160 for the distance, but the rest is a guess.
Also, if you had to choose just 1 method of segmentation for guess factoring (against Nanos of course) do you think that distance would be the best one? My initial results seem to point at this, but was just wondering. Anyways, I'll have my first GF Nano done here shortly. --Miked0801 22:57, 9 June 2009 (UTC)
I don't know anything about PEZ's utility, but personally, I would definitely recommend lateral velocity over distance, and probably even absolute velocity over distance (seeing as code size is such an issue here and lateral is so expensive). But by all means, go with what your testing shows is best. =) --Voidious 02:38, 10 June 2009 (UTC)
It seem to run fine with my old XP Sp3 (and Sp2 too before upgrade) Can you post the actual (partial?) exceptions? For the segmentation, I think lateral velocity is much important. Because many bots will be hit at the same GF at every segment (taking distance into account for flattener). » Nat | Talk » 13:24, 10 June 2009 (UTC)

--- I took a glance on your new nano. It's fascinating! Ok, it loses against some of my bots decently, but it scores awesomely against others. It scored 68% against nanolaulektrik! I'm looking forward to see the final rank. ;) --HUNRobar 21:33, 14 June 2009 (UTC)

Thanks. I had some inspiration from the 200k max conversation and went to town. I need to change RoboGrapher a bit to output a string per bot in the rumble to save me time (and accuracy in typing). That will help him out in other size catagories. I also, ideally, would average enemy velocity over a few ticks and add that to the guess factor velocity to be able to target true dodge, stop and go bots. That of course costs bytes :). My random movement code needs some tuning as well. But overall, I'm very pleased with this bot. I will put a hard limit of no more than 200k on the string though just to be fair. Maybe less. This bot would probably do ok in melee as well. I'll investigate that soon. I have to admit I was tempted to release no source until I had it tuned, but that wouldn't be fair. --Miked0801 22:22, 14 June 2009 (UTC)

Is there any good way to get at the battle results from within a robot or are those classes restricted? --Miked0801 02:21, 16 June 2009 (UTC)

No, those classes are restricted unless you use -DNOSECURITY=true. I believe that in one-on-one, you can simulate the score calculator in your robot. » Nat | Talk » 02:55, 16 June 2009 (UTC)
Just saw this go by in the recent changes... You can get scores and stuff from the Robot.onBattleEnded(BattleEndedEvent) method (also accessible from AdvancedRobots). The BattleEndedEvent has a method getResults() which returns a [BattleResults] object which contains fields you can access for all the columns of the results that pop up at the end of a battle. Not certain this is what you are looking for. --BenHorner 03:18, 16 June 2009 (UTC)
Either of these will work for me. Thanks. Expect my next blackbook update sometime tomorrow. My testbot is built and running through all the nanos now. --Miked0801 03:22, 16 June 2009 (UTC)
You must use my method since Skilgannon still ran 1.5.4 client... » Nat | Talk » 03:24, 16 June 2009 (UTC)
No I must not - it's for my local statistical gatherer only. That said, I'm using Excel and RoboResearch this time around to manage all my data. --Miked0801 05:31, 16 June 2009 (UTC)
There's a BattleEndedEvent? Interesting... --Darkcanuck 04:30, 16 June 2009 (UTC)
Interesting indeed... I'd personally suggest we try to phase out 1.5.4 asap... ;) --Rednaxela 04:39, 16 June 2009 (UTC)
Darkcanuck, I think you should move on instead =) » Nat | Talk » 07:01, 16 June 2009 (UTC)
Yes, if comes through with no rumble bugs, we could probably roll forward. Actually, with so active, is there any reason to keep 1.5.4 or 1.6.0 alive? --Darkcanuck 07:16, 16 June 2009 (UTC)
No, no reason to keep 1.5.4 or 1.6.0. At least, not in my opinion :) --Rednaxela 13:26, 16 June 2009 (UTC)

I added all the repository bots to my install and it froze building the database, I'll file a bug this evening. --Skilgannon 08:29, 16 June 2009 (UTC)

Does anyone have a better site to upload bots to than the robocoderepository? Seems like it's down quite a bit. --Miked0801 02:28, 27 January 2010 (UTC)

I personally use my university web space, but I hear google sites works fairly well and some others around here have recommended it. --Rednaxela 03:10, 27 January 2010 (UTC)

I reccommend Google Site too, buT i personally use my own web hosting. --Nat Pavasant 05:23, 27 January 2010 (UTC)

Booted up my test suite. All still in order. Started with new bots. We'll see if I can get an update in by end of week... --Miked0801 04:34, 16 May 2011 (UTC)

And updated. A bit late. I don't have local copies of extra.sauce or extra.lightsauce though. Considering they appear to be competing for worst in division, I'd be willing to vot efor them removed if someone doesn't have a local copy I could get a hold of... --Miked0801 07:38, 4 June 2011 (UTC)

I removed them for now. They used a 1.7-only method, so always crashed with zeroish scores (as we tried to explain). Welcome back and good luck! --Voidious 15:57, 4 June 2011 (UTC)