Thread history
Viewing a history listing
| Time | User | Activity | Comment |
|---|---|---|---|
| 04:58, 28 April 2012 | CrazyBassoonist (talk | contribs) | New reply created | (Reply to micro/nano melee rumble) |
| 19:31, 27 April 2012 | Wompi (talk | contribs) | New thread created |
Man, you give me hard times :). I finally thought i have something new to conquer Capulet but it turned out you allready had the same idear. My last version (wallaby 1.9) did very well and i was quite excited about this, but then showed Mercutio as major threat up in the ranking table. And it turned out you have the same movement concept. The main idear was to combine the previous antigrav and the oscillation to something compact and still hold the close combat idear. I was hopeing to give you something exciting but well, you know it already.
One question would be how to decide if the bot is good enought for the competition. I test my new bots against 9 other bots from every part of the rumble (i mean 9x5 not only 9), with a couple of 1000 runs. After this i check some special cases (closecombat - smaller battlefield, and endgame- just the best 4 of each field part). If i win with a certain percentage i put it on the rumble. But unfortunatly is this kind of check not working very well. I'm always surprised how bad/good it worked out. So the question is, what would be a good benchmark check to deside if the bot is worth the rumble. I hope i can interest you in a little discussion.
btw. sorry for my bad english take care
I've never found a sure-fire way to know how a melee bot will do before it is released; I think there's a certain degree of randomness to melee scores as each individual battle has its own unique combination of bots that affects score outcome differently. That said, in order to determine if my robots are fit for the rumble I usually do basically what you do(long battles against 9 other bots that have been randomly selected from the different melee fields) except that instead of worrying about percentages I just check and see how it does in comparison to the bots I want it to beat. For example, I released Capulet when I noticed it was consistently beating other good micro's such as HawkOnFire and Shiz when they were paired together. I think being able to consistently bots that are good is an indicator that a bot will be good, as PL score almost always tends to exactly follow APS scores in melee fields.--CrazyBassoonist 02:58, 28 April 2012 (UTC)
