Author Topic: The great NAR AI round robin. (Read 8432 times)

System32 · « **on:** August 08, 2012, 02:41:20 PM »

The latest edition of NAR AI availible as of today is being used in a massive round robin tournament. I'm cutting it up into chunks to deal with the scale, and devolving everything to a single battle each.

Basically, this is to show two things; The current state and balance of the AI pack and to be a soapbox for my RA2 musings while I update.

The first segment is a MW round robin. Segment one includes all the bots from the "stock" teams. the MW, LW, BW and AW packs will be in BBEANS v2 arena. Let's watch.

http://challonge.com/6gg5kkmore

The 28 contestants will fight for 27 rounds. That is over 300 battles. over 1000 if I did a best of three.

Luckily, The fact it's a round robin should equalize the unfair losses as the tournament drags on. Not only this, the best bots from the chunks will go into a few "high tier" chunks to really sort out the best of the best. This should essentially serve to show what bots truly need updating, what bots don't and what the metagame is looking like at the moment.

freeziez · « **Reply #1 on:** August 08, 2012, 02:42:13 PM »

MassimoV · « **Reply #2 on:** August 08, 2012, 02:49:18 PM »

Dang, you got alot of work ahead of you.

Naryar · « **Reply #3 on:** August 08, 2012, 02:56:55 PM »

Cool, I'll use the results for RA2 science.

nicsan2009 · « **Reply #4 on:** August 08, 2012, 02:57:32 PM »

That will take a Really long time.

System32 · « **Reply #5 on:** August 08, 2012, 03:01:18 PM »

Okay, First thing:

As of round 2, I'll be doing all of scavengers matchups before I continue. The UI of challaonge allows for this and Scavenger nasty pickes a lot, taking up a ton of time.

System32 · « **Reply #6 on:** August 08, 2012, 04:53:16 PM »

Aaaand done with scavenger. At 2-25, he's arguably the worst bot in Segment 1.

The design is solid as a concept and fun to watch, and he was never Redzoned at all. Most of his losses were due to a lack of DPS, Trying to nasty pickle and being knocked out of the arena.

He beat Iceberg due to out weaponing it and the opponent flipping itself due to the motor speed of it's drive, and 742 due to the fact that bot has issues with being unable to self right from the back and sides.

@ guy who AI'ed Explosive Material:

Quote from: Naryar on August 03, 2012, 06:11:50 AM

Nice 0 radius there.

The bot is another solid design, but the AI literally punishes it to a ridiculous degree. Most of it's losses are due to it turning broadside because it's wedges snagged. Up the radius and it'll be a solid competitor.

System32 · « **Reply #7 on:** August 08, 2012, 10:20:13 PM »

I think I'm done for the night.

The 4 top winners will go to the high tiers.

Also, the BBEANS hammer bot Reier made is better than the two in Segment one. They need updating.

Mr. AS · « **Reply #8 on:** August 08, 2012, 10:23:53 PM »

this (very huge) tourney should give us insight into what bots need updating in nar ai

Scrap Daddy · « **Reply #9 on:** August 08, 2012, 10:25:08 PM »

This is awesome

Naryar · « **Reply #10 on:** August 09, 2012, 04:01:56 AM »

lol scavenger actually won something ?

I'll see what I can do with Explosive Material's AI. And upgrade the bot as well.

System32 · « **Reply #11 on:** August 09, 2012, 06:08:03 AM »

Quote from: Naryar on August 09, 2012, 04:01:56 AM

lol scavenger actually won something ?

I'll see what I can do with Explosive Material's AI. And upgrade the bot as well.

742 was unable to flip it out before it put itself to its side. that design needs a slight update, preferably to emulate the UK flippers, which should fix that problem.

Iceberg is just bad. Up the DPS, Look at Starcore AI hammers for self righting, Curve the hammer, Better the wedge.

System32 · « **Reply #12 on:** August 09, 2012, 07:58:33 AM »

Samurai is amazing. It has great throwing power from it's weapons and DPS. The problem is the lack of weaponry leaves it useless in any sustained matchup, even moreso than other popups. Also, a lack of "opponent control" means a lot of it's potential goes to waste as harrassment.

Finally, it has self righting issues, even worse than erectus. Samurai has to backflip until it loses power to actually self right.

These flaws ruin an otherwise solid design that holds up to opponents with weapons and wedges that thoretically outclass it.

Also, erectus 2 has serius self righting issues. This costs it matches. Even worse than ExMa's bindings failures.

System32 · « **Reply #13 on:** August 09, 2012, 10:52:00 AM »

My favorite image so far.

I should have taken the pic where feeblemind jammed itself againt the blade and spent the whole match using that as a weapon, or scavenger diving onto slashback by ignoring he central bit, clinbing up and attacking by rolling off it onto it's chassis.

Scrap Daddy · « **Reply #14 on:** August 09, 2012, 11:43:44 AM »

Are you doing every middleweight in nar AI?

System32 · « **Reply #15 on:** August 09, 2012, 11:47:12 AM »

Quote from: Scrap Daddy on August 09, 2012, 11:43:44 AM

Are you doing every middleweight in nar AI?

By the end of the project, I'll have done every robot.

Scrap Daddy · « **Reply #16 on:** August 09, 2012, 11:49:00 AM »

awesome, really.

MassimoV · « **Reply #17 on:** August 09, 2012, 01:08:36 PM »

No, only the ones in the first 16 teams.

System32 · « **Reply #18 on:** August 09, 2012, 01:15:43 PM »

Quote from: MassimoV on August 09, 2012, 01:08:36 PM

No, only the ones in the first 16 teams.

I broke it up into segments. Otherwise, I'd end up doing over 1000 matches, and that's just for MW's.

MassimoV · « **Reply #19 on:** August 09, 2012, 01:25:13 PM »

Oh, I was just looking at the challonge page and only saw the first 16 teams. But I think splitting it up is deffinently a wise move.