View source for User talk:VoidBot

Welcome Bot! Hello, VoidBot you silly bot, and welcome to Your Home! This place contains a wealth information about Robocode, from basic to more advanced. I hope you enjoy creating robots and being a robocoder fixing pages! If you are posting a comment on this wiki, please sign your messages using four tildes (--~~~~); this will automatically insert your username and the date stamp. If you are not familiar with MediaWiki, these links might help you out: How to edit on Wikipedia, Cheatsheet and Reference Card of MediaWiki on the Meta Wikipedia. Readme page. If you need help, check out the frequently asked questions or ask it on this page. Again, welcome! --Rednaxela 00:31, 6 August 2009 (UTC)

Oooh a bot? What's this for Voidious? Will it surf WikiWaves? :) --Rednaxela 00:31, 6 August 2009 (UTC)

Haha... Thanks for the warm welcome. =) Actually, I don't (yet) have any plans to make this bot update the wiki; I'm just fetching page contents with it for now. Doing it through a bot and a MediaWiki API library seemed much easier and more reliable than fetching and parsing raw URLs / HTML.

As for the purpose, I'm a bit skeptical that anyone besides myself will think it's all that cool, but even so, it seems like something fun to play with: I'm toying with the idea of automatically tweeting about new / updated RoboRumble bots at @roborumble on Twitter. (It's actually looking pretty easy with Net::Twitter and MediaWiki::API.)

A bit more complicated, but potentially way cooler / more useful, would be waiting until the rating stabilizes, then also including the rating(s) / ranking(s) in the tweet. Of course, I'd want to clear it with Darkcanuck before any program of mine starts fetching pages from his server, but I don't think hitting the rankings pages every few minutes would be that big of a load. (At least compared to how much I already hit them... =)) And I should also mention that PEZ deserves credit for putting this idea in my head while we were discussing @robowiki.

--Voidious, master of VoidBot, 00:59, 6 August 2009 (UTC)

I've considered adding a simple public query page that would allow people to get raw data from the database. It would be read-only and probably enforce some light usage limits. I'd like to have this personally so I can run some data-integrity checks offline but it would also be useful for projects like this. Of course, the RR server could just send "tweets" for you when a new participant is added -- no need for polling. --Darkcanuck 06:31, 6 August 2009 (UTC)

Actually... I was just recently thinking about doing some data-mining on the rumble data (i.e. clustering of all of the bots, to reveal interesting categories/outliers) and a raw query interface to get data would be kind of nice, espescially if it allows incremental fetching :) --Rednaxela 06:38, 6 August 2009 (UTC)

Start a new page somewhere defining the types of queries you'd like to run. Obviously the interface would support what's already available in the rankings tables (including pairing and battle details). Keep in mind that summary data (like the rankings table) is fast to query, pairing data is slower (table is at least 700^2 records for general 1v1) and battle details are very expensive (10+M and counting). --Darkcanuck 06:45, 6 August 2009 (UTC)

A way to query the raw data like that would be awesome. I just hate to create more work for you, especially for a side project for which I'm not sure there's much interest. But I'll gladly post what queries I'd be interested in, in case you create something like this.

As for the RR server itself doing the tweets, that would certainly work and would avoid polling. That again seems like more work for you (even if I wrote the code), but I'm open to that solution if you'd prefer it. If it were to include rating & ranking info, we'd be talking about something like "if (battleCount > 2000) tweet();" when receiving a battle. It would be less load and the "true push" aspect of it makes my mouth water =), but it seems more prone to bugs than polling from outside, and the polling would be a pretty light load if we had that query interface.

--Voidious 14:05, 6 August 2009 (UTC)

Hmm... I have no time to look into my Robobot lately. I know I should fix it soon because pages are migrating. I should find a time to work on it. » Nat | Talk » 09:29, 6 August 2009 (UTC)

A test comment from VoidBot, hopefully hidden. --VoidBot 03:15, 12 October 2009 (UTC)

Well, at least could you please mark your edit minor edit? (i.e., Well, at least Voidious please marks your bot's edit as a minor edit?) [EDIT: Well, please mark your page creations as minor creations please?] --Nat Pavasant 15:58, 25 November 2009 (UTC)

Hmm, yeah, I'll try to figure out how to do that. I'm not seeing an obvious way to do so via the MediaWiki::API library that I'm using now. Looks like MediaWiki::Bot can do it, though. So not this week, but hopefully next week. --Voidious 20:31, 30 November 2009 (UTC)

See subject line.

It's because they have incomplete pairings... The script sleeps and checks hourly until they are complete, but that hasn't been happening ever. Sorry I didn't reply to your post about that yet, but thanks for making some progress on the pairings issue! It was very close to complete pairings this past Sunday. :-)

I see. Then incomplete pairing is partially my fault. I introduce new version before pld one settles.

But do we really need the full set of pairings? If a bot missing 10% of a total number, than it must be very lucky to skip the best bots. It is unlikely that such a bot wiuld raise or sink to much with additional pairings.

If it has only one battle, it is a different story. I've seen my bot on top of the rumble after one battle :)

Full pairings isn't that difficult to achieve, so I'd rather only archive when we have the full set of data to compute the rankings correctly. There's only the one bot in 1v1 now that's an issue. We should be back to regular archives soon. :-)

It isn't generating archived rankings for roborumble, any idea why? All bots have 988 pairings.

I noticed some bots had 989 pairings again yesterday. I went to check again a few minutes ago and let you know about it, since that would hold up the archive, but I see that they're at 988 now. Maybe that just fixed itself today? It checks every hour so I'll kick it when I get home if it hasn't run by then.

That was probably the Twitter one, which runs every 15 mins. The rumble archiver should be at the top of the hour. It's possible I turned off the "roborumble" last week when I ran it manually or something... I'll check it when I get home.

On that note, I wonder what's up with the Twitter one... Either it shouldn't be checking the rumble server because of no new participants, or if it has participants that it's watching, it should have seen the rumble was stable and tweeted them.

Ok, summary:

• The archiver saw pairings incomplete for a few hours, then hit a parsing error - I don't know exactly what happened. I ran it manually and it worked.
• The tweeter was set to 2800 battles for 1v1, so it has a couple bots that it's watching. Brought that back down to 2000 for now until we get the min roborumble battles back to like 3k.
• Starting up my clients again.