Difference between revisions of "User talk:Zyx"
(melee testing?) |
(→Melee Testing: my thoughts) |
||
Line 29: | Line 29: | ||
== Melee Testing == | == Melee Testing == | ||
How do you guys test melee bots? Not tool-wise, I know [[Voidious]] patched [[RoboResearch]] which I will probably use to test the melee bot as well soon. But I mean battle-wise, I was testing with the top 6 bots from different authors and 2 ''random'' middle bots, the other two spots were taken by two versions of my bots and ran 100 round battles and compare the results of my two bots to decide which one is better. But I knew that testing wasn't suitable, YersiniaPestis 3.0 can beat [[Shadow]], [[Aleph]] a lot of times (~40% of times beats one of them), [[Portia]] every now and then and [[Fermat]] every time under that circumstances, but I knew that in reality battles don't such a high concentration of top bots so results were kind of irrelevant as rumble performance meant. Thanks. --[[User:Zyx|zyx]] 21:03, 19 August 2009 (UTC) | How do you guys test melee bots? Not tool-wise, I know [[Voidious]] patched [[RoboResearch]] which I will probably use to test the melee bot as well soon. But I mean battle-wise, I was testing with the top 6 bots from different authors and 2 ''random'' middle bots, the other two spots were taken by two versions of my bots and ran 100 round battles and compare the results of my two bots to decide which one is better. But I knew that testing wasn't suitable, YersiniaPestis 3.0 can beat [[Shadow]], [[Aleph]] a lot of times (~40% of times beats one of them), [[Portia]] every now and then and [[Fermat]] every time under that circumstances, but I knew that in reality battles don't such a high concentration of top bots so results were kind of irrelevant as rumble performance meant. Thanks. --[[User:Zyx|zyx]] 21:03, 19 August 2009 (UTC) | ||
+ | |||
+ | : Well, I struggle with this too, but my attitude is to try to emulate the same mixes of bots that you'll face in the rumble. So you'd want the same range of skill levels, some battles with all weak bots, some with mid-range bots, and some with stronger bots. Right now my test bed is 27 bots, split into 6 groups (with each bot appearing twice), and I'll run ~20 seasons. For many versions of [[Diamond]] in a row, this was accurately predicting a rating increase or not. Then I released a version without much testing, its rumble rating went up, but I ran the test afterward and the score went down. :-/ So I'm still figuring it out myself, really. If you're curious about my test bed, it's here: [http://www.dijitari.com/void/robocode/dibed6.rrc dibed6.rrc]. Good luck, and please let us know if you make any Melee test bed breakthroughs. =) --[[User:Voidious|Voidious]] 21:30, 19 August 2009 (UTC) |
Revision as of 22:30, 19 August 2009
- Archived Talks:
- 2009/05/06
Contents
Wiki Help
I'm trying to learn how to use the wiki, so probably will be asking more questions. But to start I would like to know how to add those nice content tables to the pages, I've been going trough the pages that has them and see nothing special in the code :-S. And as a more general second question, is there somewhere I can learn how to write wiki stuff? Particularly this one of course. --zyx 01:11, 7 May 2009 (UTC)
Here is all about mediawiki syntax. MediaWiki automatically add the TOC once the page have more than 4 headers (inclusive) at before the first header. If you want to add the TOC manually, add __TOC__ to the page (like this page) » Nat | Talk » 01:22, 7 May 2009 (UTC)
Another useful link I bookmarked recently is this MediaWiki cheatsheet. I tend to check mediawiki.org, or just Google (eg, mediawiki namespaces) when I'm looking for something specific. There's a lot of good info out there because so many people use MediaWiki. But I've learned a lot of what I know from people like Nat and AaronR doing stuff on this wiki. =) --Voidious 01:30, 7 May 2009 (UTC)
- Hey, you won't believe me that this is the first wiki (beside the old one) I've working on. I learn the thing from what AaronR and Nfwu used to done actually. » Nat | Talk » 01:52, 7 May 2009 (UTC)
Thanks guys, I'm too dinosaur-ish for web related stuff. This is the first wiki I've been, and have been moderator of two forums, but there I'm just as dumb as I am here :-P. --zyx 01:56, 7 May 2009 (UTC)
Handle
I always realized that your handle was the last 3 letters in the alphabet, but it only just occurred to me that this is like the opposite of "ABC". I'm curious if that is intentional or just coincidence? (To be clear, I mean if it's inspired by ABC the Robocoder or not.) Kind of funny that two of our PL kings have such similar handles. =) --Voidious 18:16, 4 August 2009 (UTC)
- I think it is coincidence, because his email is zyx###@gmail.com (### replace with number, I don't remember which) » Nat | Talk » 08:49, 5 August 2009 (UTC)
- It is a coincidence, I've been using zyx since nintendo times :). --zyx 11:56, 5 August 2009 (UTC)
Melee Testing
How do you guys test melee bots? Not tool-wise, I know Voidious patched RoboResearch which I will probably use to test the melee bot as well soon. But I mean battle-wise, I was testing with the top 6 bots from different authors and 2 random middle bots, the other two spots were taken by two versions of my bots and ran 100 round battles and compare the results of my two bots to decide which one is better. But I knew that testing wasn't suitable, YersiniaPestis 3.0 can beat Shadow, Aleph a lot of times (~40% of times beats one of them), Portia every now and then and Fermat every time under that circumstances, but I knew that in reality battles don't such a high concentration of top bots so results were kind of irrelevant as rumble performance meant. Thanks. --zyx 21:03, 19 August 2009 (UTC)
- Well, I struggle with this too, but my attitude is to try to emulate the same mixes of bots that you'll face in the rumble. So you'd want the same range of skill levels, some battles with all weak bots, some with mid-range bots, and some with stronger bots. Right now my test bed is 27 bots, split into 6 groups (with each bot appearing twice), and I'll run ~20 seasons. For many versions of Diamond in a row, this was accurately predicting a rating increase or not. Then I released a version without much testing, its rumble rating went up, but I ran the test afterward and the score went down. :-/ So I'm still figuring it out myself, really. If you're curious about my test bed, it's here: dibed6.rrc. Good luck, and please let us know if you make any Melee test bed breakthroughs. =) --Voidious 21:30, 19 August 2009 (UTC)