MiniMeleeRumble

Wow, you're at the top of the MiniMeleeRumble with a MicroBot - amazing! Congrats again. (Maybe it will drift in and out, but it's at least a tie.)

Thanks man. Yes it was quite a challenge to reach GlowBlowMelee and i'm really happy that i got at least the same score now. Thanks to RoboRunner i could start some more extravagant tests and it paid off very well :)

Indeed really amazing! Only 1 loss (close to tie) and 1 loss in survival. If you make a mini, will it also reach top-3 in the meleerumble?

Thanks. Hmm there is something going on in the rumble Wallaby is right now shown in micro but not in mini or general.

What the ... its now back and has lost 0.3% . This makes me quite suspicious :(

Anyway thanks, i guess i need a very good 1vs1 strategy to reach the top-5 and right now that is not something i'm very good at :)

Looks like something is really messing with the rumble stats right now :(. Diamond also lost 0.2% and to me it looks like anyone above top ten also. GlowBlowMelee has gained 0.1% and that after 65k battles that shouldn't happen i guess. Oh man, Neuromancer is back to 280 pairings? This looks very bad right now ... :(.

I've seen weird stuff like this when someone points a MeleeRumble client at the RoboRumble participants list. Maybe that's what happened. Normally I wouldn't be worried about Darkcanuck being able to fix it, or even just rebuilding the rankings with the amount of Rumble Power we have these days =), but Darkcanuck hasn't responded to much lately. I guess at worst we could switch over to a LiteRumble config, or I could put up a new instance of Darkcanuck's server code.

Hmm. Looks like it lost some pairing informations but if it gets one of the lost parings it also gets back to normal. Every bot with missing pairings has right now some sort of score glitch. Maybe it comes back if all pairings are reestablished.

Yes too sad that Darkcanuck is not around anymore. LiteRumble would be nice, especially with his new bot stats but i guess it can not handle the amount of contributions right now. Well, lest see what is coming out of all this.

I think it might have been my clients, I just checked it and it was crashing at the end of every upload then trying to reupload everything next time. I didn't see it removing any bots though, so I'm not sure. I've restarted them, see if it helps.

Oh, and really great work you've been doing with Wallaby =) I'm worried about what you'll do when you have limitless codesize available for Numbat!

Oh, right, back on topic. =) I was going to say that my best guess as to where Numbat needs improving is his gun, not 1v1 skill. I remember being really surprised at how strong a Melee gun can get as I climbed the top 10. You might do well to try a test with Diamond's gun in Numbat (or Numbat's movement in Diamond, if that's easier =)) to get an idea how it compares. IIRC Neuromancer is still fairly weak in 1v1, though I'm not sure when it was last tested there.

Yes you are right the gun needs to make a big jump, it crashes everything with a weak movement but is useless against everybody else. In its current state Numbat can not handle the top-10 micros it ends 5th or worth (compared to Wallaby who can handle it very well) but against the top-10 nanos it works superior. But with a little time i'm sure i can make the gun a little better.

My guess for Numbat-Move and Diamond-Gun is, it will not work very well. It looks like your gun needs a lot of scan info and Diamond is hugging the edges/corners to bring the radar in a good scan position. But my risk movement brings the bot most of the time, if its too crowded, out of the edges/corners. So i'm not sure if i can get the needed scan info to feed the gun. But it sounds like worth a try, just for the curiosity.

If i say 1vs1 it also means in battle 1vs1's it happens to often that i get caught in a 'duel' and loose a big chunk of my energy just to put down an oscillator mover. And loosing 1vs1 against top-30 bots is something i really have to solve :). I bet Neuromancer is quite good at 1vs1. maybe not against Diamond :), but everybody else should have a hard time to bring him down.

Neuromancer actually isn't that great at 1v1 - I think it is due to a combination of distancing and bullet power selection. I know the gun could be improved, and some of the movement mechanics aren't quite correct in 1v1 (and even less so in melee due to the uncertainty).

Hmm, I'm not sure how much Diamond's gun would be hurt by the fewer scans. But every gun needs scan data - if it would hurt Diamond's gun, it's hurting your too. =)

Numbats radar works very different to Diamonds. If Diamond gets in the open field he starts to spin the radar. Numbat locks on his target and only rescan near bots if they could reach him so i can just run around in the open field and still get a quite long scan time to collect my patterns. If i put my movement to your gun and radar it would spin the radar most of the time and i guess this is nothing what your gun likes :). Hmm thinking of it again, it should work to take my radar as well, i guess this would show if a good gun is what i need.

But his radar does that to get the most information as quickly as possible - it scans in the direction of the enemy which was scanned longest ago. He builds DV through interpolation, so it shouldn't matter that much if it spins a lot. Also, he builds DV against *everybody* all the time, from all the bots' perspectives, so it is better if it doesn't point at one bot a lot. He basically does Shadow/Melee Gun but with DV which means it doesn't help to scan a single enemy.

Good point, i forgot he uses DV. Well, in this case the radar indeed doesn't matter that much.

Really it's nothing to do with DV specifically so much as the fact that interpolation and [Shadow/Melee Gun]-style techniques are used. All of the melee bots from at least Glacier up use some form of interpolation and [Shadow/Melee Gun]-style techniques in their targeting, and because of that the top priority for the radar is "minimize the largest gaps in time between sightings of the enemy, for every enemy" instead of "get the maximum number of scans of the enemy we want to hit".