View source for Talk:Knight/VersionHistory
- [View source↑]
- [History↑]
Contents
Thread title | Replies | Last modified |
---|---|---|
APM? | 7 | 01:43, 8 November 2017 |
First page |
Previous page |
Next page |
Last page |
Congratulations to the promising MC2K7 results! I also had a run and find pattern matchers really hurting me... like they used to be.
Anyway, did you tune your movement specially against PMs? I did so as well but never got some real improvement. (What I do is to come up with a new tree with attributes designed for pattern matchers, but it seems to have no effect ;() Or maybe I need some tuning in surfing algorithms instead, like adding some stop option or decel randomly like DrussGT does. Orbiting predicted enemy pos may help as well (although I never tried it), or that may be caused by the weakness of Fancy Wall stick...
Thank you!
I was tuning against AS and magically a firing wave flattener with time-since-decel and all that stuff, besides usual attributes, gave me a good result. Only thing a did different from my past attempts was to build *really* separate sets of trees. The flattener stats and the hit stats were on a single set of trees in the past. Now they are distributed over two sets, each one with a bunch of trees. Then I normalize the logged buffers (dividing by max) and weight them by something like 60%/40%. This helped me to give a proper weight to the flattener. Not sure, but I suppose you already do this.
The hard thing now will be to keep this good results without being hurt against simpler bots. I hope it is just a matter of tweaking the movement enough so that the flattener threshold is not hit that often. And well, close-rangers, I can handle them separately in the future.
You do not have permission to edit this page, for the following reasons:
You can view and copy the source of this page.
I was already doing better than ScalarBot, both with the old flattener and with no flattener at all, but yeah, adding the flattener gave a tremendous improvement.
Thanks for that information ;)
After a long time of experiment, I finally think the main difference from bots do good against PMs is not in the surfing algorithm... but the way surfing stats are handled. Maybe I should try some more traditional way before starting innovation... I've been already dropping old surfing stats (which uses 1-nn) since 0.012n1 (and finally got similar performance), now maybe I should start dropping crowd tree views ;) I use three simple views each with 3~5 attributes as main surfing stats now, maybe that's the reason why I got hit from PMs (and the top guns as well) badly.
I always thought strong APM was from lots of temporal attributes. It also makes sense that it would be from a small K size to prevent always dodging the same points and building repeated patterns.
First page |
Previous page |
Next page |
Last page |