Robo Home | Changes | Preferences | AllPages

Showing revision 27
Following a discussion on the Repository forums, i am thinking about a league on a very long, thin battlefield, with lots of bots. Rules i'm thinking about at the moment are:

The rankings will probably be something like 10 points for a win, 5 for 2nd, 3 for 3rd, 2 for 4th and 1 for 5th, although i may do something more conventional if i change my mind.

Is there any interest in this kind of league? It could be interesting to see what people do with SearchPatterns? and suchlike. -- Tango

Count me in with Tron. I must make some tests, but I remember version 1.5 was ready for big battlefields... -- ABC

Ok, will do. Anyone else? The first season will be run once i get 20 competitors. I'll put in some SampleBots for fun, and to boost the numbers. Please add your bot to /Entrants. -- Tango

I just tested, Tron is ready without modifications. I tried a 35 rounds battle, 5000x400, with Tron, DT, and all the sample bots, it is quite fun to watch, and Tron comes out the clear winner. ;) -- ABC

Tron did quite well in my test battles yesterday too. DT also did quite well. -- Tango

Why wait for 20 entrants? Just let the battle begin with the entrants you now have. Shouldn't take too long to run 10 rounds with 13 bots. -- PEZ

I guess i could do. I'll do it over the weekend if i get a moment to set it up. So anyone who wants to enter has until Friday evening. After that you might miss the first season. -- Tango

I never run any battles in fields with this size. But I'm glad to see that DT isn't prepared there too. nevertheless he always won in my tests. -- rozu

I just ran a 10 round battle with all the bots and it only took a few seconds. Thank you for all submitting FastBots?. :-D Therefore, i am now running a 1000 round battle so we will get a much better idea of what the good bots are in this battles. -- Tango

Ok, the battle has reached 300 rounds. It should be finished relativly early this evening. :) Although Robocode doesn't show scores midway through a battle, Tron's own survival count is showing that Tron is probably winning. The battles are quite fast, but i think i could significantly increase that speed (in later battles) by lowering the inactivity time. Most of the bots don't search for other bots if they can't see any of them, so end up sitting still or just moving aimlessly until the inactivity timer kicks in a kills them. What do people think about me lowering the time next battle? -- Tango

I've not been looking at the wiki for a day or two - hence the late entry - I've no idea how DT will do or behave. I had not considered the problem of not being able to scan your opponent. Of course if it does not win a new version will have to be developed :) -- Paul Evans

You've missed this season, of course, but i will enter you for the next one, whenever that is. Possibly tomorrow. I have been thinking of a format for the rest of the league to take. What do people think of the following:

As battles don't take too long i can probably run 3 or 4 seasons a week, so the turn around for relegated bots will be quite fast. New versions of bots will replace the old version, either in the main list, or in the same place of the old version in the waiting list. As only 10 bots will compete at a time, i am considering a one bot per person rule. I will give you at least a few days notice to choose your best bot if i am going to do that. -- Tango

Done! :-D -- Tango


I never run a battle of this kind, but I added my robot here, so I can see how it behaves :-) -- Simonech

Cool, Tron got over 50% firsts. :) Can't wait for season 2, I believe Tron has a good chance of placing 1st even with DT in the battle, it looks like you have something to work on, Paul... ;) -- ABC

Season 2 will probably happen at some time today. I think i will move the SampleBots to the bottom of the waiting list, as they are only really there to make up the numbers. I have some ideas for a bot specially made for this league, which i will start coding today. I hope other people will do similar (either in new bots, or as additions to current bots). -- Tango

Season 2 is going to start now. I'll put the results on /Season2 when it's finished. The inactive time will be 100 (default is 450). I don't know what people think of this, as no-one answered. Please let me know what you think of it. I may well put it back up when people start making bots that actually search, so they have time to find their enemies, but at the moment too many battles finish with 2 bots sitting outside radar range doing nothing, which is a waste of my cpu time. -- Tango

It doesn't seem to be running much faster... considering there are less bots, and the inactive time is lower, i think all the blame must fall on DT being a SlowBot... -- Tango

With such a big field, I don't think lowering the timeout is a good idea. I'm ok with it while very few bots search for each other, but it takes a long time for a bot to go from one side of the battlefield to the other. Tron currently uses a simple "random wander" (prefering places where it hasn't been recently) when it doesn't see any of the enemies, the default timeout seems good for when it doesn't find anyone for too long. -- ABC

As the speed increase was virtually nothing, i will put it back to default next time. Having watched a few rounds, Tron is very impatient. It starts going in one direction, and then just before it gets within scanner range of a bot it turns around. Prehaps you should set up a slightly better search pattern, random isn't really working. -- Tango

It's working good enough to win ;), as soon as an enemy is detected the antigravity kicks in, it's still a "defensive" melee bot. Assuming that the enemies also want to fight, it should work decently enough. Of course a smarter search pattern should work much better, I'll think about that after someone dethrones Tron ;). Anyway, it's great to have some melee action going on again! -- ABC

Yes, it is significantly better than having your movement in the onScannedRobot method, so as soon as you go out of radar range you stop dead, which is what most bots seem to do (DT included). Also, where bots fire is strange when they have no bots in range. DT continues to fire where it last saw a bot, despite it being miles away, and thus loses a lot of energy. So, Paul, if you add one line of code to make sure you have uptodate info on the location before you fire, i think you will do much better. Same applies to a few other bots. -- Tango

Viewed with femto-eyes WallsPoetHaiku performs pretty well. I've not looked at it in action on this strangely shaoed battle arene, but I guess it moves in a tiny square somewhere near where it started. =) -- PEZ

/Season3 is complete. Tron won again. That makes it 3 out of 3. Well done! Simone did quite well, but has a bug which caused it to be disabled in the first few rounds. I haven't looked into it much, but it looks like it marks bots as dead in the onRobotDeath? method like everyone else, but doesn't check if it knows the bot first. In normal battles it is almost impossible for a bot not to be seen before it dies, but in this league it is quite common. When that happens Simone gets a NullPointerException and is disabled. It looks like the enemies hash or whatever it uses is remembered between rounds, so once it got a complete list it operated fine for the rest of the battle, so did ok overall. It shouldn't be hard to fix, so i suggest it is done in the next release of the bot. Walls got last place, and so is removed from the league. That makes Crazy the last SampleBot left, and it is quite high in the league still. Surely we can make 10 bots that are better than the sample bots? -- Tango

Robo Home | Changes | Preferences | AllPages
Edit revision 27 of this page | View other revisions | View current revision
Edited October 25, 2003 18:58 EST by PEZ (diff)