From Robowiki
Jump to navigation Jump to search


Thread titleRepliesLast modified
Congratulations on recent updates!518:45, 19 November 2017
APM?700:43, 8 November 2017

Congratulations on recent updates!

It's rather hard to improve even a tiny at the top, and now I see Knight ascended to #6 with noticeable APS gain after a few versions, congratulations!

I also came back to ScalarBot and decided now to do what I really want to do. I found myself changing things I don't want to change and ended up losing points, so now I switched to focus on what I want to do the most, and gained noticeable improvement, e.g. ScalarBot 2.8 wins around 60% constantly against 2.7.101. I think the most important part about this game is not about tweaking style only tiny things, rather, aggressive changes that gets the point gives you the most gains. Yes, tiny things are still important, but I think you can lose score easily by making mistake in it, but you can never earn a tiny from that.

Btw, I'm really interested in the recent changes of Knight ;) Well you add that at some point?

Xor (talk)15:00, 18 November 2017

Excited to see ScalarBot moving to the GoTo-style movement! I was planning to do so this weekend, but probably for different reasons as I'm almost sure I won't gain any points solely because of it, I struggle a lot with GoTo :P

I've mostly worked on hitting and dodging random movers recently, I never run battles against Knight itself. Seems to be a nice thing to do for my next experiments.

Sure, I'll post them soon but they were mostly anti-random gun changes and the experiments are still ongoing. The reason I've not been posting too much is that I moved to another city to an internship and I have a very low cost laptop right now, so I barely open the browser :P But I'm writing down all my change log on file and I'll update the wiki with it soon. Hope I can grab some good improvements over this weekend.

Rsalesc (talk)15:55, 18 November 2017

Just read more carefully the change logs now and it seems that it is Scalar that is actually moving to GoTo. :P

Rsalesc (talk)16:00, 18 November 2017

Well the name seems confusing ;) And only its 1v1 movement is GoTo ;)

Xor (talk)17:21, 18 November 2017

I second that, really nice improvements the last weeks from your sides. But indeed, the air is getting thin so it is hard to get higher in the ranking/APS. Although my big changes, to be able to use BulletShadows, are a bit postponed due to failed preconditions (0.3.21-0.3.23), fixing/improving some small stuff when you encounter them can also be satisfying. I rather fix things and then take the big step, then the other way round, because that big step will blur the things to fix/improve.

GrubbmGait (talk)16:26, 18 November 2017

I'm mostly doing this right now. While developing Roborio I could take a glimpse of what it is like to do a feature-driven development, and it certainly haven't paid out. When I rewrote the code I could find a dozen bugs lurking around my gun and my movement and eventually Knight became really different from the former just because of those bugs. To avoid losing my motivation (which is very important as well, we gotta have fun at the end of the day), I'm interpolating fixes/tweaks with new features/experiments.

Rsalesc (talk)18:45, 19 November 2017

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...

Xor (talk)06:58, 7 November 2017

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.

Rsalesc (talk)08:19, 7 November 2017

Well I think the only noticeable difference in handling flattener now is that I normalize buffers based on area, instead of max ;) AFAIK Neuromancer is also doing this, although without flattener or multi-trees (but he is dealing with multiple enemies ;) ).

My flattener also uses time-since-decel, as I find my surfers having weakness about that attribute in the past (that is VCS surfers, though).

Btw, are you doing with PMs that well in the past, or just after recent updates? I'm wondering whether flatteners help against PMs ;), since I've been long suffering about flatteners being turned on occasionally against PMs, which hurts.

Xor (talk)10:16, 7 November 2017

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.

Rsalesc (talk)14:39, 7 November 2017

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.

Xor (talk)17:46, 7 November 2017

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.

Skilgannon (talk)21:29, 7 November 2017