Robo Home | Changes | Preferences | AllPages

by Pulsar

/History - Chat Start - Chat Newest


PulsarMax is my attempt at a really sophisticated robot. One on One only for now. As a software architect/developer I couldn't stand doing it any other way than using different classes and general stuff, so no codesize restrictions :) First of my bots which has more than distance segmentation for targeting and the first which has a decent movement. Been working on it for quite a few weeks before releasing the first version (0.1)


See Pulsar

How does it move?

WaveSurfer mainly. Selects other movements under specific circumstances.

How does it fire?

Multiple Guess factor targetings with waves.

How does it dodge bullets?

It is a WaveSurfer.

How does the melee strategy differ from one-on-one strategy?

It tries to stay away from everyone else and the middle of the battlefield. It also tries to avoid head on and linear targeting.

What does it save between rounds and matches?

Matches: Nothing. Rounds: Movement and targeting stats.

Where did you get the name?

My alias is Pulsar, MAX is something me and my collegues tend to add to software stuff when we want a new and improved version :)

Can I use your code?

Please read and understand the RWPCL before looking at the code.

What's next for your robot?

I still have a long todo list, some things are fairly basic too actually. More and better segmentation for the targetings. Faster learning for targetings. Code refactoring. Dynamic data segmentation (looking forward to this one!), a secret targeting thing, selection based targeting, etc.


What other robot(s) is it based on?

SnippetBot! Lots of things from the Wiki! Uses apv's FuturePosition for now. Inspirations are from persons rather than bots I guess as the wiki is the source of many ideas. I tend to read and take notes. Everyone seems very helpful and I hope I can return the favour to some small amount one day. Especially PEZ has helped a lot and while no code of his is directly in use in PulsarMax he has inspired and given hints. The actual gun segmentations are fairly similar to some older version of CassiusClay.

Chat - /OldChatSummer2004

Looks like I need to improve my learning speed too, good, something more to add to the never ending todo-list. Example, DevilFish?, 1000round match is great but 35 tends to not be as good (relatively, everything considered). That's understandable but can be improved. Something simliar (although to a lesser extent) with head on targeters is kind of bad though. PM shouldn't need to get hit even once to learn that. At least this version (0.5.20) seems much improved, finally, after some 25 versions without any improvements in score. -- Pulsar

Totally agreed. Great attention to learning speed is the next logical step in WaveSurfing evolution. -- PEZ

Pulsar will be away from robocoding for I think a week or so. What about we prepare a present for him when he comes back? His gun has some peculiar property. It is a good gun. Together with a top class WaveSurfing movement it can bring home 2035+ rating points. But. It's totally lame against surfers. What could cause that? Let's brainstorm folks! Maybe we can figure it out. Then we would maybe be part of helping forward a new King of Robocode land even. -- PEZ

Well, my experience in GF guns is aprox. null (PulsarMax gun should be a PM, dont u think?:)... But u can count on me. My initial guess to itīs problem would be learning velocity (but its a guess only)... Btw: it would be good if its source was open... -- Axe

What do you mean by "learning velocity"? -- PEZ

I assume Axe menas learning speed. I have a minute or two until wednesday then I'm back sunday evening at the earliest. Thank you PEZ you are too kind! Regarding the source code I've previoulsy said: "For now I'm not supplying the source, I most likely will later on. It's nothing secret though so go ahead and ask!" and I guess it's stable enough now to be able to be released. I'll see if I have some time later today to release it in the jar or at least send it to PEZ. Regarding the gun it must be a bug somewhere. The last versions have the exact same complexity as the Bee gun so learning speed shouldn't be a problem, not the same segments exactly but still. I've tried the exact same segmentation as the Bee gun and that made it better but not nearly as good as plugging in the real Bee gun. Current version is rather similar but using my own utilities etc. -- Pulsar

17% against ad.last.Bottom ... And 56% against CassiusClay. Interesting! =) -- PEZ

Heheh that made my morning at least :) Think I know what it is, a null pointer. --Pulsar

There's nothing wrong with a null pointer as long as you don't use it. =) -- PEZ

You can check off "2040+ score" on your goals. Congratulations! :-) --David Alves

Thanks! Finally! --Pulsar

I saw it yesterday, PM 2050+ (beating also SSīs score). Congratulations man! I knew u could get there! -- Axe

Thanks, well it would never have happened without you all. After I finally found that bug with the targeting that has annoyed me for 3months or so everything is fun again and I can experiment again! Now 2100 looks a goal I would love to reach. ;-) This is the first time I feel that I can reach top 3 for a few days and maybe for a really short while top 1 position ;-) -- Pulsar

I'm impressed as well. You missed top 3 by only 3 points. Well done. --Mue

Great work! Maybe it's time I tune in to RC again, looks like I'm about to lose my top3 place and my PL highscore. Have you tested the latest PulsarMax against Shadow in 500/1000 rounds? -- ABC

Yeah, really good work. And good work stirring things up again. Your surfing act might be best in breed now. Or tied with Shadow's or something like that. Feel free to try out a .bee version. Could be interesting. At least for me to see if #1 is reachable through surfing changes only. I just can't improve the gun it seems... -- PEZ

Looks you've reached your next goal, pushing Shadow out of the top-3 for the first time since . . . ? ever ? Congratulations !! --GrubbmGait

Thanks, but I'm not so sure, it is such a small margin! -- Pulsar

Well the margin is 3 points now, which is clear enough for me. So congratulations from me too! And i'd like to suggest a new goal:

Defeat Shadow in a direct duel, to break its domination in PremierLeague ;-). Its incredibly strong in battles vs other strong bots and i didn't manage to beat it yet (well, i haven't had much time for robocode recently, but still...) --Mue

Thanks! I'm currently busy with something else for PulsarMax, when I have the time to robocode. I'm happy enough with 3rd place for now and beating Shadow sounds like a nice goal but it will have to wait! -- Pulsar

Congratulations from me too, 3 points is definitely clear enough. You are now seriously threatening CC's 2nd place. About beating Shadow, I'll probably try to regain my top3 position soon, and it seems that you have to score worse against Shadow 3.55.1 to rank higher in the 2050+ club. I might have to do some optimising against mid-range bots, and that'll probably decrease my PL performance... -- ABC

Glad to see you dedicating some time to melee, I'm sure you'll quickly enter the top10 (and top3). All it takes is a good melee capable gun and a well tuned anti-gravity movement. I'll be away next week, I expect PulsarMax to be challenging Shadow's all-around best bot title when I return... ;) -- ABC

Hehe. Thanks for the encouraging words. I'll be away most of next week (CeBIT? in Hannover, Germany) so I'm afraid it will have to wait. Well... "well tuned anti-gravity movement" sure seems harder than it sounds, loads of fun though. But yes that title is my goal! If only for a day! -- Pulsar

Hey man, I just thought I'd let you know that PulsarMax still rams his enemy if they're disabled when he's set to "movement challenge" mode. It doesn't affect my own targeting testing against him, but it would certainly throw off his MovementChallenge scores... (if you care :)) -- Voidious

Thanks! -- Pulsar

Robo Home | Changes | Preferences | AllPages
Edit text of this page | View other revisions
Last edited May 4, 2006 21:46 EST by Wcsv (diff)