I just remembered it got confused battles/ruounds at first. Obviously it was me confusing it. =) -- PEZ
How can I add my national flag to the ranking list? I'm from Italy -- Simonech
You tell PEZ and he will put it on, so consider that done. And while you're there, PEZ, can i have a UK flag please? Thanks. -- Tango
I've added a temporary italian flag. But it isn't the same size as the rest of the flags (which I ripped of amarook's site). If anyone has a complete and consistent set of world flags in a usable format and size, please send me them. -- PEZ
That was quick! Thanks. -- Tango
As long as it's one of the flags already in the list it's as simple as making a link to that flag. -- PEZ
While we're adding flags that the ER doesn't have, I'm pretty sure shinh.Entangled is from Japan, and Marcelo.Alpha.DevilFISH is from China. -- Kawigi
Yes, I think so. I could always install it. I'm on a Debian system you know. =) -- PEZ
will ask here if this makes sense. I think that there should be an admin page that allows a Robocoder to add his/her bot to the RoboRumble. I would think that on this admin page you would be able to specify all the parameters on the participants page plus some other things (flas association for instance). If this is a sane approach, then I would suggest that this information be placed in a table in the databse. The participants file would be buuilt dynamically on the fly and downloaded as needed. It would also help cut down on some of the errors that are waiting to happen as people continue to hand edit that page. While I am on a jsp kick, I would think that the results page should be built dynamincally on the fly. In order to accomadate this, I would suspect that the Upload Servelet would have to be modified to push all of the results into a database. This would allow us to store raw numbers and free our selves from the current display dependancy we have. Now, if all of this information is pushed into a database, there seems to me that we should now allow ourselves a table which specifies league parameters. If you want a Minibot challenge, then we would need to create a jsp that asked the database for bots which met the requirements and a display page that would properly display the results. We would need some scoring module to plug in that described your scoring system. We would potentially need to allow clients to specify which league they were interested in helping along. I am willing to work this task list over the next week or so but I can assure you it will most certainly not have pretty graphics or style.
PEZ: To do this I would need Tomcat installed, mySQL or some other SQL based DB, JDBC connector installed on the system. I can help you with this as I get closer to actually deploying the system.
Albert: I need you to not changed the format of the files (participants and posted results) for a bit without telling me. I actually had started this once but I could not keep up with you so I put it on hold a bit.
The question is: Does this have merit and would it be helpful? -- jim
yes, i think it's a good idea. -- Tango
Most of this seems good. But I don't think it's a good idea to build the results page dynamically. All CPU issues aside I think a given standing should stick a while. Maybe a day, maybe a few days. Then it could as well be built batch-wise. Though I guess it could still benefit from the decoupling of data and display I think you are aiming at. Tomcat is already involved in the current setup and mySQL should be installed if I read the dpkg output correctly. A JDBC connector for this could be including in the war/jar deplyoment files of your efforts I guess? -- PEZ
(stupid editing conflicts) It's generally a good idea. I'll nitpick on a few things. First the dynamically building the page on demand thing has caused a nightmare for amarok and ET, the load times are horrendous (as I'm sure you've experienced). Second it would be more useful to run a seperate 'game' (something Albert seems to have included accomodations for) for stuff like Minibot leages, because you're measuring how they perform against each other, not how effective they are in a codesize-free environment. That's details, though. -- Kuuran
I think some of the performance issues you experience on ER is due to that that server is behind a 56k modem. At the moment it takes less than 0.5 seconds to build that wiki page. And that's likely to get faster once we have limited the detailed history data. But's it's unecessary anyways. And I kind of like to have it as a wiki page too. =) -- PEZ
I like the idea of using a DB (even if I'w be of little help there, because I have no idea about it). But it guess it's not the moment. As theory says, first build the model, then add the technology, and clearly it's faster to make changes in this "quick and dirty" environment we have. I agree that calculation pages dinamically would be horrendous, not because the ammout of data, but because the number of calculations involved. I'w upload (I hope tonight) a new Server, that keeps sumarized ratings (so we don't need to read all files to create the ranking), calculates the rankings using a rolling estimate of %score, and reuses previous matches (with new ratings) to speed up convergence (it will also allow us to decrease change constant). The space needed by this system is small (I guess less than 15K per bot), but of course it will take a little bit more CPU when results are uploaded (because of recalculations), but I hope it will be offset by the smaller file sizes.
About participants and upload results format, I'm not changing it, since they are client related (and I don't expect to change it for a while). If I do I'w post it first. -- Albert How about allowing bots with different URLs? Since they are run with the robocode security manager in place there isn't too much chance of malicious ones, and I'd like to enter a bot into the competition without having to wait for the robocode repository to come back online. :-p --David Alves
Just upload it somewhere else, ask people to download it manually, and add it to the participents list. I'm sure people will be willing. -- Tango
When we're not a small group of people that know each other doing this, it may get tedious... -- Kawigi
Hopefully the repository won't go down again. -- Tango
A simple sugestion: Bot author rankings. Or maybe just a filter that displays only the best bot of each author. -- ABC
Replace "author" with "package" and it would be simple. They are usually the same thing, but not always. eg the wiki package. -- Tango
Well, simplify it to outer-most package name maybe. Then you'd get all the kawigi.* robots, not seperate the kawigi.nano and the kawigi.sbf and the kawigi.f and so on. If people are into national pride, we could also filter to the best bot representing each country ;-) -- Kawigi
Yeah, that's what i meant. I wasn't very clear, sorry. -- Tango
Hi. Just a few thoughts about the big response times on EternalRumble. I don't think it's a 56k connection. I would say it's my bad style of programming :) All results are stored in one big file (now about 5 MB of size). I don't really have fast algorithms to build the grouping tables. That's why i decided to move to database access. Currently i try to find a clean and reusable concept for handling those big amounts of data. I'm going to use mySQL. Later i try to run my battles on the server (if my provider doesn't kill me :) At the end there should be a fully automatic web site where you can enter your bots, update, and all battles run automatically (am i dreaming ?). I think i will give more attention to the melee battles (and maybe team matches) as you have now a cool tool to run lots of battles. Also ChampionsLeagues? and other competitions (nano, micro, mini leagues) are worth to think about. cu -- Gerald aka amarok
Imagine that we ever thought you had quitted Gerald! =) -- PEZ
I'm testing an aothor ranking page now. Let me know what you think. -- PEZ
Nice one. May be we could use only the package name? When you look at it it gives you the impression you seeing the bots rankings page. -- Albert
The numbers in the rankings by author are not consecutive (ie. they jump from 3 to 5). -- Albert
Rapid feedback I must say. =) -- PEZ
Shouldn't we eliminate the old Rankigs pages (or make them point to the correct ones)? -- Albert
I don't understand the meaning of the new raking pages:
what does "% score" means?
and I think that the old details page is more interesting that the new one... it had the list of battles, the "White wales" and so on. -- Simonech
We had to sumarize the battles in order to reduce file sizes and bandwith. About WhiteWhales, I plan to reindroduce them in short (we just needed to fix the ratings and that was not priority #1). Also, all battles are stored in a big "raw data" file, that will be shared when we manage to mirror it. The percentage of the score is the percentage of the score you get against a given enemy. Let's say, if you win 200-100, then you percentage is 66,6%. A percentage greater than 50% means you are able to beat the enemy. A percentage of less than 50% means he beats you. In order to deal with multiple battles per enemy, the system smoots the results using an "exponential smooth". -- Albert
Edit-conflict-but-I-dont-care: % score means how big a share of the total score a bot has grabbed. For instance if botA beats botB 6500-3500 in score the share would be 65%. I think the current page is more interesting than all those battle details. But I guess we could make the servlet take arguments that will make it display just about anything we have data on. I think we will see the ProblemBot (was WhiteWhale) featuremake a reappearance. It was way cool! -- PEZ
Thank you: now that I understand the meaning of "% score" I admit that it's more interesting that the battle list!!!
Anyway, Good job guy!!!! -- Simonech
Rather than a specific page for problem bots the Details page for each bot could have two extra columns added - one would be the 'expected score' (based on the present rateing difference, the other the difference between the actual score and the expected score (negative scores are bots dragging down your rating). The page could have an option to sort by name (case insensitve), expected score, and score difference. -- Paul Evans
Cool. I'w like that (may be the sort functionality will take a little bit more).
Now, I'm running also my first battles. even I can do it, great. I've two questions and I think I post them here: the rating page shows the number of battles. is this the number #rounds/35 (just because if someone changes ROUNDS in the properties-file, the number showed would be false)? is 1200x1200 the default size for team-battles? (if yes, I should test my team with this size set in the future) -- rozu
I don't know how it's calculated, but battles that aren't 35 rounds are ignored by the server when you try to upload them, so it shouldn't be a problem. I have no idea about the second point, someone else will have to tackle that one. I'm not sure team rules have ever really been fleshed out. -- Kuuran
The server stores the number of battles, not the number of rounds (anyway, the server filters any battle not conforming the specifications, so ust don't worry). About teams, I would say yes (10 rounds, 1200x1200 battle field). These values were the ones used for the Face2Face Teams competition (which was the only team competition ever existed, and lasted only for a few weeks, so there is not much experience about it). -- Albert
Last weekend I went to the seaside and I left my pc at home running with roborumble@home on it.
My PC computed 110 iterations of the battles...(running 50 battles for iteration) in 3 days of running.... and it worked without problems!!! -- Simonech
Today I'd prefer be sailing my boat like I did yesterday, but I'm stuck here in my office, but I don't want to do the silly web-project that are our clients at the moment:
So just thinking about robocode... what do you think about a melee championship handled by rr@home? I always preferred melee on one-vs-one, and also my bot performs better in melee than in 1vs1 :-) -- Simonech
looking at team ranking. Your right Albert 1200x1200 is great :) -- rozu
Good job, rozu! -- David Alves
Ditto, but don't count your chickens - the ranking will take much longer to settle down because there are fewer opponents - it's too early to say who will end up on top. -- Paul Evans
thx, and I know. as far as I can see your team will take the first place soon. I wish I could run team-battles too but I can't. DT is eating my memory :). I just want to ask if we could add some other classic teams like TroodonPack or Phalanx? -- rozu
Sorry about the memory thing - if you want to test locally create a a team of 5 DT 1.21's (it will be similar to 1.71 performance) - In the next release of DT I will create a new team package. -- Paul Evans
Something came to me while I was looking at an old version of Predator and considering adding it to the rumble. Alot of these older bots retired on top, is it right to drag them out and ruin the prestige they went out with by beating the hell out of them a year (or years) after they're relevant? -- Kuuran
Yes, it's right. It's not like with athletes who stop being able to perform out of injuries or age reasons. These bots still perform at the peak of their abilities. If other bots have raised above their level it's not more than right that this shows in the rankings. Let's beat the crap out of those abandoned bots! -- PEZ
*cringe* --Duelist
In a sense it is because their technologies have dated, but I guess I'll add et.Predator 1.8 to the participants list. Note this robot's design is soon to be two years old. -- Kuuran
What do you think of adding sample.Walls (and others?) to the participants list? It would be interesting to see how far we are from the FIRST bots. -- Albert
et.Predator was the king of melee for a long time when the Gladiatorial League was the main robocode competition. If we ever get a melee roborumble I'm sure it will still be in the top half of the table. -- ABC
Indeed, that's why I added it, I ran some tests and et.Predator was still able to score 2/3 of what each Tron did in a field full of Tron 2.02s. It's a monster considering it's age. I'm curious why ER didn't include it. -- Kuuran
I have two suggestions maybe already requested. 1) giving battle pariorties to new bots so they have equal match numbers with the older ones. 2) a mechanism to check the new vesion of the bot exists in the repository so that the score doses not start from 0 every time when there is an update to a bot. SSO
The first request will come, probably, with the next client release. About the second one, the new server classes (that should be installed soon) deal with that, and bots start from the ranking of its previous version :-) Albert
Greate work guys, its getting better and better. SSO
You only get large numbers of downloads when someone downloads roborumble for the first time, not just an update. The robot jars are the same for any version. I doubt there will be any more problems until the release of the first real version, when people not involved in the wiki download it, and then the bots. -- Tango
Maybe we can get an accout from geosites or any other free hosts? Than the rumble gets the latest bots form this address. SSO
The repository is hardly the most stable site on the net. It's down quite often, for reasons unknown. I think we should opt for not relying on it at all, not even with caches in between. I have had a version of GloomyDark ready (to PEZ standards) for rumble a few days now, but haven't been able to upload it. I'd say it's time for a RoboRepository@Home. -- PEZ
Should we all just run a webserver? :-p -- Kawigi
That might do the trick why not discuss the various approaches on the RoboRepositoryAtHome page. -- PEZ
It's little bit confusing now which version to download which client is the last one and competible with server ect. I think the old versions should be saperated, archived and explaned very briefly. SSO.
I'm running it, and in fact today there have been more than 10000 battles, so people is also running it. The problem is that new bots take some time to climb the rankings. It was not evident when we created the list (because all bots were evolving at the same time) but now ... Anyway, I'm working on the new client that will priorize the bots with a low number of battles fought, so it should correct it and make a bot stabilize its ranking in just a few hours. -- Albert
I'm running it too on a spare machine I have tomporary. =) It will be pretty cool when a new or updated bot can get its ranking that fast. I'm the impatient type, as you all know. =) -- PEZ
I hadn't run many battles lately, so i've been running quiet a few today. The current Total battles = 113881.5, so i think albert missed a 0. BTW, how can you have half a battle? Is that a rounding error, or is it something to do with people changing the battle settings (i thought changed battles were ignored, but...)? -- Tango
It could be a concurrent update problem. See the /RankingsChat? page for some random babbling on that issue. -- PEZ
Is anything happening with the idea of a list of who's run the most battles? I think more people would run more battles if there was. People generally like to be credited for work they do. I'd say the people running battles at the moment are the kind that do it because they want to, and they know they and others will benifit, but more people would run battles if it meant they could be a "Top 10 Uploader" or whatever. -- Tango
I think we all think that's a great feature. But noone has started working on it yet. I'm not sure we have the data necessary to create this list. Have we, Albert? -- PEZ
The data used to be on the details page, but is probably lost now with the summerising. Could the server just keep a simple int array of client names, and increment the right one every upload? The information about what battle they uploaded can be lost, although whether it was general, team, nano, etc. might be good to keep. -- Tango
Yes. There is a log file that has the details of all battles fought, including who uploaded the results. It is only necessary to read all lines and count how many times the names are repeated. -- Albert
Wow. This RoboRumble@home idea is the coolest thing ever. In fact on saturday I was looking at a friend's computer running SETI@home and was wondering what it was doing; when she briefly explained how it worked I understood then what SETI@home is and then what RR@home is so today I downloaded it and am now participating. Since I live in residence, I should be able to basically leave my computer on all day and let it run battles while I'm in class. I have about 6 hours of class a day and I'm AFK most of the time I'm here anyway, so I think I'll just let it run all the time, whenever I'm not programming (Robocode has this way of slowing the crap out of NetBeans, as if it wasn't slow enough already). Anyway, although it seems as though the project is basically complete, I'd be interesting in helping. I'm busy for the next few days (perhaps week or two), but I'll catch up on the development of the project some time soon and see if I can lend a hand. -- Vuen
It is not finished yet. There are some things to be done to consider it completed: (a) Make the client smarter to speed up new bots ranking (I'm working on it) (b) Correctly dealing with versions (now they all appear in rankings) (c) Deal with concurrency issues when uploading data to the server. Your help is welcome :-) Albert
Might be a problem or not... but I discovered that (as maybe you also discovered) that sometimes the method for storing data to a file fails under Java 1.4.2. This can be a problem because people will not sava data among rounds and battles and when the bot tries to save data and the operation fails the robot get disabled. -- Simonech
The settings are in a file called roborumble.txt into the \robocode\robots\roborumble\ folder. In any case, I don't recommend to change anything at all, except the user name (note that the server will refuse any result not accomplishing with field size-number of rounds). About runing RoboRumble while using the computer for some other tasks, I do, but I guess it deppends on your computer: the main issue will be that it slows down your computer. -- Albert
The other thing to change in the file would be if you want to run minis only or something. -- Kawigi
Thanks for responses, if you put the priority of the roborumble processus to low, it don't slow too much the computer, but there's big difference of allocated CPU every time you do somthing (I hope that was clear), do you think this can change the results ? --Synnalagma
I think it can, unless you manually change the delay time allowed to bots... -- Kawigi
That will just change the skipped turns, won't it? I thought it was generally decided that skipped turns don't have much effect in small quantities. Anyway, i thought the time allowed to each bot was dependant on how fast robocode was running anyway... -- Tango
It's dependent on how fast it is running at start-up. -- PEZ
With the iteration system it starts up quiet often, so it shouldn't be too much of a problem. -- Tango
I agree. I usually run it at work while I work with the computer and doesn't seem to deliver strange results. -- Albert
Hey, I am new to robocode and I love it. I wanna ask you a question. How do you guys feel about clones? The fastest path for newbie here is to start with some open source bot (there is good number of great bots to learn from) and hopefully improve on it. I *think* I improved Sedan a lot, to the point my version can possibly top mini list... what do ya think - should I go on and enter roborumble? -- Andre?
It should not be a problem, as long as you get the permision of the author of the bots (for Sedan they are Iiley and Kawigi). Just send them a mail, or post the question in Sedan page. For sure they will answer. -- Albert
If you are new to robocode and can improve Sedan, that's pretty impressing in itself. We never have had a clone before that I am aware of. I'd say that you should send your suggested changes to the authors of Sedan and if it gives the extra edge they'll include it and you'll be one of Sedan's contributors. -- PEZ
If it's one of Kawigi's you'd better read the KawigiPublicLicense, i think it says you can't use his bots to make better bots. -- Tango
Also iiley's open source bots ask you not to just use the code as is and tweak it. But I can't see why they would object to you improving the bot and they can incorporate it into a new and stronger Sedan. That's part of the reason to open source stuff to begin with. -- PEZ
I agree, i expect they would be grateful of suggestions to improve, and will give credit where it is due. -- Tango
Thanks for the input. I posted the question cause I don't like the idea of tweaking someones code anyway. Just imagine horror of tens of Smoke/Flood?/etc clones entering the competition. So now I go and write mine from scratch. Sedan code is too bloated for mini anyway :) Credit for teaching me (and I bet many others) how to do good statistical gun still goes to Kawigi.
I suspect that Sedan still can probably beat just about any MiniBot given time, and really close to any MegaBot. But its main purpose was to beat SandboxDT 2.01. So if you have an improvement that makes it beat SandboxDT 2.11, that would interest me ;-), I've searched for such an improvement for some time. -- Kawigi
Just downloaded 0.61 version, but I'm looking at something quite strange:
apv.MicroAspid 1.7 is the only one playing, against every other robot... I understand that this can be the "smart battle" algoritm goin on, but I suppose that this way all other bots will not run battles untill the newer bots have a number of battle similar to the other ones. And also, it seems to me that the rrh is not checking rating files after each iteration, so if I start the engine in the morning I would run battles just for the new robot because I'll never know if it reached a "good" numbero of battles.
Don't know if this is the right place for this discussion, if not, I'm sorry -- Simonech
The other bots will run matches still. About half as many as when all bots have an equal battle count if I understand it correctly. The engine should look for battle counts each iteration though I guess. -- PEZ
The smart algorithm works as follows:
It should speed up the updating of new bots. Next step will be to make the clients run the battles bots have pending against a given enemy. -- Albert
Check question: If a bot belongs to the, say, mini category. And it gets its 500 rounds in the general game. Let's say 100 of these are in the mini game. Then it seems like that bot would get 400 priorty mini battles. Can't that tilt the general rating some? Wouldn't it be better to give a bot priority in an inverse linear proportion to it's battle count share of the total number of battles? Then new / updated bots would get lots of priority battles while the mix of general/mini/micro/nano battles would be kept intact. As it is now Griffon (a mini) quickly got 1000+ battles while Jekyl (a mega) which was updated a few days before Griffon only has 500+ battles. And, I think, an unproportional share of those 1000+ battles of Griffon are against minis. For a nano this would of course be even more pronounced. -- PEZ
How about allowing bots with different URLs? Since they are run with the robocode security manager in place there isn't too much chance of malicious ones, and I'd like to enter a bot into the competition without having to wait for the robocode repository to come back online. :-p --David Alves
Just upload it somewhere else, ask people to download it manually, and add it to the participents list. I'm sure people will be willing. -- Tango
When we're not a small group of people that know each other doing this, it may get tedious... -- Kawigi
Hopefully the repository won't go down again. -- Tango
A simple sugestion: Bot author rankings. Or maybe just a filter that displays only the best bot of each author. -- ABC
Replace "author" with "package" and it would be simple. They are usually the same thing, but not always. eg the wiki package. -- Tango
Well, simplify it to outer-most package name maybe. Then you'd get all the kawigi.* robots, not seperate the kawigi.nano and the kawigi.sbf and the kawigi.f and so on. If people are into national pride, we could also filter to the best bot representing each country ;-) -- Kawigi
Yeah, that's what i meant. I wasn't very clear, sorry. -- Tango
Hi. Just a few thoughts about the big response times on EternalRumble. I don't think it's a 56k connection. I would say it's my bad style of programming :) All results are stored in one big file (now about 5 MB of size). I don't really have fast algorithms to build the grouping tables. That's why i decided to move to database access. Currently i try to find a clean and reusable concept for handling those big amounts of data. I'm going to use mySQL. Later i try to run my battles on the server (if my provider doesn't kill me :) At the end there should be a fully automatic web site where you can enter your bots, update, and all battles run automatically (am i dreaming ?). I think i will give more attention to the melee battles (and maybe team matches) as you have now a cool tool to run lots of battles. Also ChampionsLeagues? and other competitions (nano, micro, mini leagues) are worth to think about. cu -- Gerald aka amarok
Imagine that we ever thought you had quitted Gerald! =) -- PEZ
I'm testing an aothor ranking page now. Let me know what you think. -- PEZ
Nice one. May be we could use only the package name? When you look at it it gives you the impression you seeing the bots rankings page. -- Albert
The numbers in the rankings by author are not consecutive (ie. they jump from 3 to 5). -- Albert
Rapid feedback I must say. =) -- PEZ
Shouldn't we eliminate the old Rankigs pages (or make them point to the correct ones)? -- Albert
I don't understand the meaning of the new raking pages:
what does "% score" means?
and I think that the old details page is more interesting that the new one... it had the list of battles, the "White wales" and so on. -- Simonech
We had to sumarize the battles in order to reduce file sizes and bandwith. About WhiteWhales, I plan to reindroduce them in short (we just needed to fix the ratings and that was not priority #1). Also, all battles are stored in a big "raw data" file, that will be shared when we manage to mirror it. The percentage of the score is the percentage of the score you get against a given enemy. Let's say, if you win 200-100, then you percentage is 66,6%. A percentage greater than 50% means you are able to beat the enemy. A percentage of less than 50% means he beats you. In order to deal with multiple battles per enemy, the system smoots the results using an "exponential smooth". -- Albert
Edit-conflict-but-I-dont-care: % score means how big a share of the total score a bot has grabbed. For instance if botA beats botB 6500-3500 in score the share would be 65%. I think the current page is more interesting than all those battle details. But I guess we could make the servlet take arguments that will make it display just about anything we have data on. I think we will see the ProblemBot (was WhiteWhale) featuremake a reappearance. It was way cool! -- PEZ
Thank you: now that I understand the meaning of "% score" I admit that it's more interesting that the battle list!!!
Anyway, Good job guy!!!! -- Simonech
Rather than a specific page for problem bots the Details page for each bot could have two extra columns added - one would be the 'expected score' (based on the present rateing difference, the other the difference between the actual score and the expected score (negative scores are bots dragging down your rating). The page could have an option to sort by name (case insensitve), expected score, and score difference. -- Paul Evans
Cool. I'w like that (may be the sort functionality will take a little bit more).
Now, I'm running also my first battles. even I can do it, great. I've two questions and I think I post them here: the rating page shows the number of battles. is this the number #rounds/35 (just because if someone changes ROUNDS in the properties-file, the number showed would be false)? is 1200x1200 the default size for team-battles? (if yes, I should test my team with this size set in the future) -- rozu
I don't know how it's calculated, but battles that aren't 35 rounds are ignored by the server when you try to upload them, so it shouldn't be a problem. I have no idea about the second point, someone else will have to tackle that one. I'm not sure team rules have ever really been fleshed out. -- Kuuran
The server stores the number of battles, not the number of rounds (anyway, the server filters any battle not conforming the specifications, so ust don't worry). About teams, I would say yes (10 rounds, 1200x1200 battle field). These values were the ones used for the Face2Face Teams competition (which was the only team competition ever existed, and lasted only for a few weeks, so there is not much experience about it). -- Albert
Last weekend I went to the seaside and I left my pc at home running with roborumble@home on it.
My PC computed 110 iterations of the battles...(running 50 battles for iteration) in 3 days of running.... and it worked without problems!!! -- Simonech
Today I'd prefer be sailing my boat like I did yesterday, but I'm stuck here in my office, but I don't want to do the silly web-project that are our clients at the moment:
So just thinking about robocode... what do you think about a melee championship handled by rr@home? I always preferred melee on one-vs-one, and also my bot performs better in melee than in 1vs1 :-) -- Simonech
looking at team ranking. Your right Albert 1200x1200 is great :) -- rozu
Good job, rozu! -- David Alves
Ditto, but don't count your chickens - the ranking will take much longer to settle down because there are fewer opponents - it's too early to say who will end up on top. -- Paul Evans
thx, and I know. as far as I can see your team will take the first place soon. I wish I could run team-battles too but I can't. DT is eating my memory :). I just want to ask if we could add some other classic teams like TroodonPack or Phalanx? -- rozu
Sorry about the memory thing - if you want to test locally create a a team of 5 DT 1.21's (it will be similar to 1.71 performance) - In the next release of DT I will create a new team package. -- Paul Evans
Something came to me while I was looking at an old version of Predator and considering adding it to the rumble. Alot of these older bots retired on top, is it right to drag them out and ruin the prestige they went out with by beating the hell out of them a year (or years) after they're relevant? -- Kuuran
Yes, it's right. It's not like with athletes who stop being able to perform out of injuries or age reasons. These bots still perform at the peak of their abilities. If other bots have raised above their level it's not more than right that this shows in the rankings. Let's beat the crap out of those abandoned bots! -- PEZ
*cringe* --Duelist
In a sense it is because their technologies have dated, but I guess I'll add et.Predator 1.8 to the participants list. Note this robot's design is soon to be two years old. -- Kuuran
What do you think of adding sample.Walls (and others?) to the participants list? It would be interesting to see how far we are from the FIRST bots. -- Albert
et.Predator was the king of melee for a long time when the Gladiatorial League was the main robocode competition. If we ever get a melee roborumble I'm sure it will still be in the top half of the table. -- ABC
Indeed, that's why I added it, I ran some tests and et.Predator was still able to score 2/3 of what each Tron did in a field full of Tron 2.02s. It's a monster considering it's age. I'm curious why ER didn't include it. -- Kuuran
I have two suggestions maybe already requested. 1) giving battle pariorties to new bots so they have equal match numbers with the older ones. 2) a mechanism to check the new vesion of the bot exists in the repository so that the score doses not start from 0 every time when there is an update to a bot. SSO
The first request will come, probably, with the next client release. About the second one, the new server classes (that should be installed soon) deal with that, and bots start from the ranking of its previous version :-) Albert
Greate work guys, its getting better and better. SSO
You only get large numbers of downloads when someone downloads roborumble for the first time, not just an update. The robot jars are the same for any version. I doubt there will be any more problems until the release of the first real version, when people not involved in the wiki download it, and then the bots. -- Tango
Maybe we can get an accout from geosites or any other free hosts? Than the rumble gets the latest bots form this address. SSO
The repository is hardly the most stable site on the net. It's down quite often, for reasons unknown. I think we should opt for not relying on it at all, not even with caches in between. I have had a version of GloomyDark ready (to PEZ standards) for rumble a few days now, but haven't been able to upload it. I'd say it's time for a RoboRepository@Home. -- PEZ
Should we all just run a webserver? :-p -- Kawigi
That might do the trick why not discuss the various approaches on the RoboRepositoryAtHome page. -- PEZ
Are you guys not running the client any longer? Please do. -- PEZ
It's little bit confusing now which version to download which client is the last one and competible with server ect. I think the old versions should be saperated, archived and explaned very briefly. SSO.
I'm running it, and in fact today there have been more than 10000 battles, so people is also running it. The problem is that new bots take some time to climb the rankings. It was not evident when we created the list (because all bots were evolving at the same time) but now ... Anyway, I'm working on the new client that will priorize the bots with a low number of battles fought, so it should correct it and make a bot stabilize its ranking in just a few hours. -- Albert
I'm running it too on a spare machine I have tomporary. =) It will be pretty cool when a new or updated bot can get its ranking that fast. I'm the impatient type, as you all know. =) -- PEZ
I hadn't run many battles lately, so i've been running quiet a few today. The current Total battles = 113881.5, so i think albert missed a 0. BTW, how can you have half a battle? Is that a rounding error, or is it something to do with people changing the battle settings (i thought changed battles were ignored, but...)? -- Tango
It could be a concurrent update problem. See the /RankingsChat? page for some random babbling on that issue. -- PEZ
Is anything happening with the idea of a list of who's run the most battles? I think more people would run more battles if there was. People generally like to be credited for work they do. I'd say the people running battles at the moment are the kind that do it because they want to, and they know they and others will benifit, but more people would run battles if it meant they could be a "Top 10 Uploader" or whatever. -- Tango
I think we all think that's a great feature. But noone has started working on it yet. I'm not sure we have the data necessary to create this list. Have we, Albert? -- PEZ
The data used to be on the details page, but is probably lost now with the summerising. Could the server just keep a simple int array of client names, and increment the right one every upload? The information about what battle they uploaded can be lost, although whether it was general, team, nano, etc. might be good to keep. -- Tango
Yes. There is a log file that has the details of all battles fought, including who uploaded the results. It is only necessary to read all lines and count how many times the names are repeated. -- Albert
Wow. This RoboRumble@home idea is the coolest thing ever. In fact on saturday I was looking at a friend's computer running SETI@home and was wondering what it was doing; when she briefly explained how it worked I understood then what SETI@home is and then what RR@home is so today I downloaded it and am now participating. Since I live in residence, I should be able to basically leave my computer on all day and let it run battles while I'm in class. I have about 6 hours of class a day and I'm AFK most of the time I'm here anyway, so I think I'll just let it run all the time, whenever I'm not programming (Robocode has this way of slowing the crap out of NetBeans, as if it wasn't slow enough already). Anyway, although it seems as though the project is basically complete, I'd be interesting in helping. I'm busy for the next few days (perhaps week or two), but I'll catch up on the development of the project some time soon and see if I can lend a hand. -- Vuen
It is not finished yet. There are some things to be done to consider it completed: (a) Make the client smarter to speed up new bots ranking (I'm working on it) (b) Correctly dealing with versions (now they all appear in rankings) (c) Deal with concurrency issues when uploading data to the server. Your help is welcome :-) Albert
Might be a problem or not... but I discovered that (as maybe you also discovered) that sometimes the method for storing data to a file fails under Java 1.4.2. This can be a problem because people will not sava data among rounds and battles and when the bot tries to save data and the operation fails the robot get disabled. -- Simonech
The settings are in a file called roborumble.txt into the \robocode\robots\roborumble\ folder. In any case, I don't recommend to change anything at all, except the user name (note that the server will refuse any result not accomplishing with field size-number of rounds). About runing RoboRumble while using the computer for some other tasks, I do, but I guess it deppends on your computer: the main issue will be that it slows down your computer. -- Albert
The other thing to change in the file would be if you want to run minis only or something. -- Kawigi
Thanks for responses, if you put the priority of the roborumble processus to low, it don't slow too much the computer, but there's big difference of allocated CPU every time you do somthing (I hope that was clear), do you think this can change the results ? --Synnalagma
I think it can, unless you manually change the delay time allowed to bots... -- Kawigi
That will just change the skipped turns, won't it? I thought it was generally decided that skipped turns don't have much effect in small quantities. Anyway, i thought the time allowed to each bot was dependant on how fast robocode was running anyway... -- Tango
It's dependent on how fast it is running at start-up. -- PEZ
With the iteration system it starts up quiet often, so it shouldn't be too much of a problem. -- Tango
I agree. I usually run it at work while I work with the computer and doesn't seem to deliver strange results. -- Albert
Hey, I am new to robocode and I love it. I wanna ask you a question. How do you guys feel about clones? The fastest path for newbie here is to start with some open source bot (there is good number of great bots to learn from) and hopefully improve on it. I *think* I improved Sedan a lot, to the point my version can possibly top mini list... what do ya think - should I go on and enter roborumble? -- Andre?
It should not be a problem, as long as you get the permision of the author of the bots (for Sedan they are Iiley and Kawigi). Just send them a mail, or post the question in Sedan page. For sure they will answer. -- Albert
If you are new to robocode and can improve Sedan, that's pretty impressing in itself. We never have had a clone before that I am aware of. I'd say that you should send your suggested changes to the authors of Sedan and if it gives the extra edge they'll include it and you'll be one of Sedan's contributors. -- PEZ
If it's one of Kawigi's you'd better read the KawigiPublicLicense, i think it says you can't use his bots to make better bots. -- Tango
Also iiley's open source bots ask you not to just use the code as is and tweak it. But I can't see why they would object to you improving the bot and they can incorporate it into a new and stronger Sedan. That's part of the reason to open source stuff to begin with. -- PEZ
I agree, i expect they would be grateful of suggestions to improve, and will give credit where it is due. -- Tango
Thanks for the input. I posted the question cause I don't like the idea of tweaking someones code anyway. Just imagine horror of tens of Smoke/Flood?/etc clones entering the competition. So now I go and write mine from scratch. Sedan code is too bloated for mini anyway :) Credit for teaching me (and I bet many others) how to do good statistical gun still goes to Kawigi.
I suspect that Sedan still can probably beat just about any MiniBot given time, and really close to any MegaBot. But its main purpose was to beat SandboxDT 2.01. So if you have an improvement that makes it beat SandboxDT 2.11, that would interest me ;-), I've searched for such an improvement for some time. -- Kawigi
Just downloaded 0.61 version, but I'm looking at something quite strange:
apv.MicroAspid 1.7 is the only one playing, against every other robot... I understand that this can be the "smart battle" algoritm goin on, but I suppose that this way all other bots will not run battles untill the newer bots have a number of battle similar to the other ones. And also, it seems to me that the rrh is not checking rating files after each iteration, so if I start the engine in the morning I would run battles just for the new robot because I'll never know if it reached a "good" numbero of battles.
Don't know if this is the right place for this discussion, if not, I'm sorry -- Simonech
The other bots will run matches still. About half as many as when all bots have an equal battle count if I understand it correctly. The engine should look for battle counts each iteration though I guess. -- PEZ
The smart algorithm works as follows:
It should speed up the updating of new bots. Next step will be to make the clients run the battles bots have pending against a given enemy. -- Albert
Check question: If a bot belongs to the, say, mini category. And it gets its 500 rounds in the general game. Let's say 100 of these are in the mini game. Then it seems like that bot would get 400 priorty mini battles. Can't that tilt the general rating some? Wouldn't it be better to give a bot priority in an inverse linear proportion to it's battle count share of the total number of battles? Then new / updated bots would get lots of priority battles while the mix of general/mini/micro/nano battles would be kept intact. As it is now Griffon (a mini) quickly got 1000+ battles while Jekyl (a mega) which was updated a few days before Griffon only has 500+ battles. And, I think, an unproportional share of those 1000+ battles of Griffon are against minis. For a nano this would of course be even more pronounced. -- PEZ
It should not, because at most it will update a %score that was updated before. The main inconvenient is that other bots fight less battles. -- Albert
I've just strated running the rr@home client again - even after a bot had got 500 rounds in all it's weight catagories it continued to be selected. Does rr@home use the ratings file to do the smart selection? If so it looks like the ratings file is only downloaded once - before iteration 0. --Paul Evans
Rating files are updated every two hours, but current number of battles is downloaded everytime you upload results. I'w check to see if it is bug in the client. -- Albert
If there are 2 bots with less than 500 battles, will they just fight eachother lots of times, or does the priority only apply to the first bot in a battle with the 2nd being random? -- Tango
I'm noticing some strange behavour in result uploading:
This is a piece of the screen dump of result uploading
OK. dmp.nano.Eve 3.41 vs. ratosh.Debo 1.33 received OK. dmp.nano.Eve 3.41 vs. ratosh.Debo 1.33 received OK. dmp.nano.Eve 3.41 vs. ratosh.Debo 1.33 received OK. dmp.nano.Eve 3.41 vs. ratosh.Debo 1.33 received OK. dmp.nano.Eve 3.41 vs. ratosh.Debo 1.33 received OK. dmp.nano.Eve 3.41 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. kawigi.nano.FunkyChicken 1.1 vs. ratosh.Debo 1.33 received OK. kawigi.nano.FunkyChicken 1.1 vs. ratosh.Debo 1.33 received OK. kawigi.nano.FunkyChicken 1.1 vs. ratosh.Debo 1.33 received OK. kawigi.nano.FunkyChicken 1.1 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. davidalves.net.DuelistNano 1.0 vs. ratosh.Debo 1.33 received OK. ratosh.Debo 1.33 vs. myl.nano.Kakuru 1.20 receivedas you many notice there are some results sent more than 4 times.
OK. I'w consider it for the next release. Anyway, the fact a bot is not updated for more than one day should correct itself during the next days. Right now, what happens it that all the delayed battles are fought in a row. When all bots currently in the RR@H have 500 battles, only new entrants will have some priority, so it won't take long. -- Albert
Is there a place we can see how many rounds have been contributed by each person? If not maybe there should be one - it would be a good incentive for people to run more battles. --David Alves
I suggested that a while back, and everyone was positive about it. It just needs someone to find the time to implement it. -- Tango
It shoudn't be difficult from the battles_*.txt (ie. battles_roborumble.txt) files, using some script. Just totalize number of records per user name. -- Albert
Although the default would have to be changed, if it hasn't already (i haven't checked), because Albert would win by a long way if it isn't. ;-) I recommend NO_NAME or something similar. -- Tango
Or ANONYMOUS_COWARD like Slashdot. :-) --David Alves
That would also work. It might be better to just not allow matches to be run without a name. People could, of course, manually set their name to something like ANONYMOUS_COWARD, but that doesn't really matter. -- Tango
Hey, but what's the problem? Nobody seems to run rr@h any more, this afternoon the results were not updated from 3 pm to 8 pm. -- Simonech
Don't know. In my case, I usually run the client only at night (from 12PM to 8AM). -- Albert
There is currently a bug in the ranking server that ignores matches older than 5 days, Albert is working on it. See RoboRumble/RankingChat. -- ABC
![[Home]](/images/RoboWiki.png)
Unexpected Signal : EXCEPTION_ACCESS_VIOLATION occurred at PC=0x6D3DE6BE
Function=JVM_FindSignal+0x17BF7
Library=C:\Programmi\Java\j2re1.4.1_05\bin\client\jvm.dll
Current Java thread:
at robocode.peer.robot.EventManager.processEvents(EventManager.java:732)
****************
Another exception has been detected while we were handling last error.
Dumping information about last error:
ERROR REPORT FILE = (N/A)
PC = 0x6D3DE6BE
SIGNAL = -1073741819
FUNCTION NAME = JVM_FindSignal
OFFSET = 0x17BF7
LIBRARY NAME = C:\Programmi\Java\j2re1.4.1_05\bin\client\jvm.dll
Please check ERROR REPORT FILE for further information, if there is any.
Good bye.
It has never happened to me before...
These are the battles that were running when the JVM crashed
synnalagma.test.MiniNeural 1.1,dmp.nano.Eve 3.41
synnalagma.test.MiniNeural 1.1,sgp.SleepingGoat? 1.1
synnalagma.test.MiniNeural 1.1,bons.NanoStalker 1.2
synnalagma.test.MiniNeural 1.1,arthord.NanoSatan Kappa
-- Simonech
Are you using a special robocode install for RR@H? If so it could be that the jar files of MiniNeural got corrupted in the download (this happens at times). -- PEZ
I don't think that's the problem because RR@H is running battles just for MiniNeural (it's a new robot I suppose) -- Simonech
I add this problem (exactly same error and another check SynnalagmaQuestion) while coding an engine for robocode. I think it's unsynchronized thing that cause the problem. (It's a new version of MiniNeural). -- Synnalagma
Question: what is needed (special Jdk-version? I use 1.1.8 at work) to view the graph in the RoboRumble results? I can see an applet is loaded, but is doesn't display anything --Loki
Hmmm... I tried to make the LRP graph completely work with version 1.1. I'll look into it. For what it's worth, you have to be using the new domain name for it to work (if you're looking at roborumble.robowiki.dyndns.org, it won't work. I suppose I could still make it work, though). -- Kawigi
My old Wiki address is http://robowiki.dyndns.org/perl/robowiki, but the graph is requested from the new address http://rumble.robowiki.net/lrp?game=roborumble&name=bvh.fnr.Fenrir%200.25. So i guess thats correct, but to make sure i will take a look when i get at home -- Loki
Can you try with another browser and see if the problem might be related to Opera's security settings? -- PEZ
--Loki
Make sure you're using the Java Plug-in for IE, and not the standard MS JVM, which may have problems. -- nano
The applet will work just fine with the MS JVM, I put effort into making java applets that way (which made me less familiar with newer Java features for some time). The difference in domain is the only problem, and it's a problem by definition (restrictions on Applets according to Sun) - not just a Microsoft-invented one. -- Kawigi
How many rounds are in a RoboRumble 1-v-1 match? and a melee match? I've looked in most RR pages, but could not find a definitive answer. Maybe a RR Rules page would be handy for newcomers. -- Vic
Yes. Welcome back by the way. It's 35 rounds in both 1v1 and melee. But I think there's a majority for 100 rounds in melee, so it might end up there some day. Problem is that noone but Albert puts any work into the servlets. And he's not part of that majority. =) -- PEZ
Thanx. it sure has been a while :-) -- Vic
Caveat. The servlets deadlock easily if too many clients deliver results simultanously. So don't go ahead all of you starting your clients. Try ynchronizing it on a page of it's own. Why not /ClientRunnerSchedule?.
-- PEZ
I don't know if it was asked before ( and i didn't relalize it :D ) but what about this file: priority1v1.txt in robots\roborumble\temp . Can i used it to make RR run a certain bot more oftenly than others (on my computer) and how can this be done ??? --deathcon
I expect you could, but it would be a bad idea. The current prioritisation works reasonably well, and people trying to change it just confuses everything. You can experiment by downloading the server files, and running your own RR@H, but keep the client completely separate from the one you use for the real thing. -- Tango
There's no way currently. But it could be implemented without too much work I guess. There is an often-asked-for feature of getting stats on results uploaders. These two features would fit in the same servlet I think. -- PEZ
There is a manual way to see it. Just take a look into the log file and check how many people are uploading results. Statistics could be built from the log file also. -- Albert
Hi, Sorry for this question but i don't now were to post it ;-) is it a Wiki of yourself or if not were can i get it, it is simply very good ... -- Franck
You mean the wiki program? It's a UseModWiki. There are lots of good wiki scripts, but usemod is indeed one of the best. Check it out on http://www.usemod.com/cgi-bin/wiki.pl -- PEZ
Everything is kept on the server. Anything in particular you're looking for? -- PEZ
No but i thought that i could write a little application that compares RankingDetails? of two bots (e.g to versions of a bot ). This would help me to find out weaknesses and goid points of special versions.If old data would be deleted i had to store old RankingDetails? ... --deathcon
OK. Yes, ranking details for old versions of any bot are kept and just as available as for the latest version. Just edit the URL and you can fetch it in your browser. This goes for the LRP graph too. -- PEZ
Yes, I think a data saving bot executed at the same time in two different clients might risk getting corrupted data and possibly crash. But you can install two separate robocode with an RR client in each. They will slow down each other though. If you have a multiprocessor machine you could assign each client to a different CPU. -- PEZ
The rating for a new bot is 1600, and for a new version of a bot it is the existing rating for the previous version. But, when you enter a new version, it can vary wildly at the begining, because the changes are calculared only on battles fought by the new version (pe. if it fights first 10 hard enemies, then its rating will go down quick, till it gets more battles and the victories make it recover). -- Albert
That is 10 problem enemies. It can fight 10 hard enemies, beat them bad and get an insanely high ranking too. =) -- PEZ
@Albert , i thought it would work like this, but could it be that something is wrong in the piece of code, that finds the previous version, because every time i enter a new version of shiva, its starts somewere around 1840 points no matter which opponents it meets first. --deathcon
I think there is a check to prevent updating the rating up or down until the bot has fought 10 battles. After that, RRH starts updating the ratings (also using, of course, the first 10 battles). -- Albert
I won't give up my 2nd place without a fight, that I can assure you ;). I've been dedicating some time to Tron's melee movement, made some melee experiments with Shadow too, both gun and movement. -- ABC
Thanks, good luck with your gun tweaking. Nice to see you back in then fight for the top, btw. Not long ago you and PEZ were saying something about never being able to reach Shadow's performance (1950 points at that time), times change... ;) -- ABC
IsnŽt time to raise the limit number of battles priorized? Everybody knows that with 500 battles the bot in ranking isnt stable yet. -- Axe
Nice to hear you are considering coming back. Robocode is still going strong, lots of competitive minds working hard as ever. :) -- ABC
Something wrong with SmallDevil? 21000+ battles fought and I get an error message when I try to look at the details. -- Alcatraz
Yeah, it's ratings file got corrupted in a crash recently. I thought I had fixed it. I'v edited it again and now at least the rating details can be read again. -- PEZ
-Dsun.io.useCanonCaches=false statement in my batchfile. Any ideas ? --GrubbmGait
Its the same on my machine. I seem to remember that its written in java 1.5, and since i'm using 1.4 there might be some compatibility problems. Anyway, battles that end with one bot getting 0 points are ignored by the server, and there is at least one client that is able to run the bot (Archers last battle was uploaded today). So its no serious problem i think. --Mue
That is not my intention, I have "RUNONLY=SERVER" in roborumble.txt, maybe it's a problem with codesize.jar, I'll check. -- ABC
Is it possible (or will it be possible) to run RoboRumble as a screen saver? (like seti@home) -- F4kill
I doubt it is possible today. But I am hoping that the open sourcing of Robocode will focus some of its attention to making it more RR@H aware and your idea here is great. -- PEZ
Edit conflict: It is currently not I'm afraid but I've been thinking about implementing that and discussing the idea with PEZ. But lack of time (and if I have some I currently prefer robot-coding!). But I think it would be great and it is possible to do. I would love to do it one day but I'll not even look into it until robocode is open sourced for now, but that is going to happen any day now though. Shouldn't be too hard (java screen saver API along with dumping a roborumble structure somewhere, the GUI can limited to start with). -- Pulsar
Who added the sample bots to the RoboRumble? Does it make sense to have them there? As a point of reference, I just watched Musashi shut out SittingDuck. If I recall correctly this means the Rumble is going to keep trying that pairing until it gets a score. Thats a lot of re-run pairings looking for a result vs. a bot that will never shoot. -- jim
It's me ... I saw that they were in the melee rumble competition and I was curious to see over how many bots they are able to win. Maybe it was an error, in this case > I'm sorry :( -- F4kill
No worries. I was not sure enough of myself to remove them EITHER=^< -- jim
The samplebots are probably entered in melee to have some more bots, otherwise you have only 18 (nano) and 30 (micro) bots left in melee. We just have to make more meleebots ! (or make them melee-aware) -- GrubbmGait
I added mine .. even if it is not made for melee -- F4kill
There is a big difference between 'not made for melee' and 'melee-unable'. Your bot performs roughly the same in one-on-one and melee, so it is at least melee-aware. --GrubbmGait
fabiolmt.nano.FireRodCk 1.0,2495 from the participantslist because it fails to correctly download. There is a mismatch between expected and real name. --GrubbmGait
There was something like the RR Uploaders Competition, but it seems not active anymore. Ofcourse you can use a stopwatch ;-) Just for reference, I run the RR@H client on a 400MHz, 128Mb XP-laptop and I get between 40 and 60 battles per hour. -- GrubbmGait
By the way, QuickHack 0.3 can not be found on the Repository -- GrubbmGait
Ouch, I uploaded it as a source ;) Nevertheless this doesn't matter, as I'm going to upload a new version today (yes, QuickHack has a daily release schedule, each time with a new algorithm). The funny thing is that I generated 682 battles for it on my own ;) --lRem
I'm afraid it wasn't just you. The power supply to my switch gave up on me it seems. A temporary solution is in place until I can get a new one. -- Pulsar
Thanks, looks like I finally managed to steal that 3rd place form you ;). Any chance of bringing back the LRP graphs? -- ABC
Yeah. I think we forgot about the LRP graphs in the move... -- PEZ
The roborumble server will be unreachable during the day today (European time). If I don't have a permanent solution in place late this evening when I'm back (depends on me having time to go buy some new stuff) it will at least be reachable during the night. I'm sorry for the inconvenience. --Pulsar
Nevermind that, a workaround to the temporary solution is i place. I didn't want to disappoint anyone addicted to robocode out there. That could've meant people actually got something important done! -- Pulsar
It should indeed correct itself. As soon as a corretly configured client is run it will note that there are particiapants that shouldn't be there and remove them, the sooner the better (less impact on scores). On scores it will have a small effect, if any that will diminish over time (rather quickly in this case). I'm more concerned about the structure of roborumble which makes it so vulnerble. -- Pulsar
I see what you mean, it looks like someone has run melee-teambattles also (must try that sometime :-D ) There certainly is need for an extra check on the battle-configuration. I'll try to fire up my client this evening. -- GrubbmGait
I hope im not the one -.- Is there a way to check who does it? --Krabb
Not entirely sure this is the best place...but...Just to let you know that I have added Vapour v150 but not yet removed v124 - it should be functionally
identical but the amount of rewritten code makes me nervous so a good test that I haven't broken anything that I haven't noticed yet will be nice :) Will
remove one or the other soo depending on the results... Shrubbery
You could also take the old version out and still be able to compare later (just edit the URLs) - the data on old bots isn't thrown away, it just isn't used
in the ranking list. -- Kawigi
Ditto, just for your convenience, here is the url for the comparison. -- GrubbmGait
http://rumble.pulsaronline.com:8080/rumble/servlet/RatingDetailsComparison?game=roborumble&name1=shrub.Vapour%20v150&name2=shrub.Vapour%20src124
Ah, that's pretty cool, thanks. v124 now removed. -- Shrubbery
Is there a problem with melee? No battles for several days... -- Shrubbery
Everybody runs melee battles for his own testing ;] And I have a little problem downloading amk bots... Apart from that, meleerumble works OK. --lRem
No problems, just the number of clients has been low lately. And this weekend my ADSL-modem broke down :-( -- GrubbmGait
(in this way it was a bit like RoboLeague, but without UI and with the possibility to test a number of your robots at once). I think the utility was written
by Mue.
Who has this utility (and can make it available)? And could it be used to test multiple versions of the same robot (if i remember correctly you could only
enter the robot-name, not the version number. is that right) ? thanks, --Loki
You can find is on Ascendant/CommentsTill200502. It only supports the bot's name, without version number. -- ABC
Hi, i'm reading the wiki every now and then. I remember making the utility aware of the bots version for my runs of the TargetingChallenge. Probably some
other things got changed too, i dont remember right now (especially not the uploaded version). So here it is:
http://mue.sonar-echo.de/runbattles.zip
Please feel free to do anything you want with the code. --Mue
main files, one for 1-vs-1 battles, the other for melee battles. You can test a number of versions of your own robot against a list of opponents. I have also
changed the logging of messages that are written to standard output a bit. It's a great tool for 'stress testing'. Mue, thanks a lot for this utility! --
close is this to how its actually done? Can someone explain the exact formula's for calculating rankings.
I think the formula for projecting rating difference based on relative score is exactly what is described there, with a base rating of 1600. In order to
show stable ratings faster, they do some iteration on past results until the rating is a bit more stable. Try /RankingSystem, which could be linked from
the RoboRumble page, too. -- Kawigi
Complete newbie here - added my bot to the RoboRumble yesterday, and didn't realise that I'd put an underscore in the name rather than a space... Anyway, how
long does it usually take a bot to get ranked? -- DarienPhoenix
Usually not long. It should appear in the rankings within a a few hours at most; however it can take a day or two before your ranking stabilizes. Anyway,
good luck in the Rumble! -wcsv
Woah, that reply was fast. :) I was wondering why it hadn't appeared, but obviously it was due to the underscore in the participants page. -- DarienPhoenix
I have just launched RR@Home here and it is runnning your bot so the first results should be here in 20 minutes or so. --Florent
Bots with under 500 battles get dibs on the next battles processed. The Roborumble client generally processes 25 duels at a time, and then the client check
the server for a new list / string of battles to process. Assuming at least one person is processing battles at the time, you will see yourself on the
rankings pretty quickly. I just run the Roborumble and Meleerumble, usually overnight or during the workday. If you run it yourself you can guarantee you
will see results quickly. Assuming you figgure out the installation, which threw me for a good hour or so. I don't follow instructions well. - [[Martin
Alan Pedersen ]]
Well, currently 282nd, which I figure isn't bad for a bot that was just made to beat SpinBot. ;) I'm impressed with how well this thing is run - entering a
tournament and receiving results within hours... Anyway, I'll have to play around with some more advanced Robocoding after I finish my exams. --
point me to this site? thanks! --Loki
Hm, maybe you're looking for Robocode/SimilarGames? --mue
Thanks! That's the page i was looking for! B.t.w. are you still active? Can we expect a 2000-topping melee-bot from you soon? ;) --Loki
No, i'm not active at the moment, i just check the changes in the wiki regularly. I have to admit though that i've been considering to start experimenting
with melee. But either way: i dont expect to enter the upper half of the melee ranking anytime soon :-). --mue
movement :) --Loki
Greetings, I'm trying to spider the wiki for offline reading. unfortunately it isn't working. I've tried several spiders (HTTrack, FDM): HTTrack is working,
but the performance is very low (1Mb/hour), FDM is telling me that I have to Login to the Wiki?? - Tried with/without using robot.txt, I'm connected via T1
and forced-proxy. Any Idea's or has someone a working copy as ZIP/RAR? Any hint/help will be appreciated. BJD
Before Robocode can process any battles, it extracts new bots into the .robotcache directory and rebuilds its database.
This takes a while, especially when running the rumble for the first time. If you have Robocode installed in the same directory, try running it first and
watch the spash screen, which tells you what robot it is presently extracting. When the client comes up you know it is done, and you can close it. Next run
the Roborumble again and you should see it start right up (since it will use the same cache and database). -- Martin Alan Pedersen
Ok thanks. I'll try I already started roborumble a long time ago and I didn't remember this (maybe I wasn't in front of my pc at this time) -- F4?
If you have the disk space, it's generally a good thing to keep your RoboRumble client directory separate from your Robocode development directory. For
instance, some robots have .properties files that you can use to turn off the gun or the movement, and you can easily end up getting bogus results in the RR
if you forget to change them back before running RR@Home again. -- Voidious
yep don't worry they aren't in the same directory ;) -- F4?
I even have separate directories for RR@Home-one-one-one, RR@Home-melee and RR@Home-teams. This keeps the startup of the battles as fast as possible. --
processing. -- Martin
have filled in in the xxxrumble.txt file. In my RR@Home-team\robots directory I have only 24 entries. Once in a while I also remove older and obsolete
versions from the robots- en .robotcache directory. -- GrubbmGait There is an indexOutOfBounds? issue allegedly caused by a corrupt download of a robot's jar file, and I suspect that a leading cause of it is actually running
two Rumbles at once against the same cache/database. If they are both trying to extract and rebuild at the same time, I can see it causing a problem. I
have one installation of the rumble on all the machines that run it, and on my work computer I'll run it overnight to process about 3000 battles without the
issue, but it errors out on occasion on my home machines. On my primary machine I run duels and melee at the same time (each assigned to one of two cpu's),
but with my present setup I cannot run two duels or two melee rumbles at once, but it's not really a problem. -- Martin
Hey guys, I have my first bot uploaded and on RR. Sobieski. As of 4/29/06, 229 in General, 93 in Mini, 53 in Micro. Anyway, could you check it out and
suggest improvements? -- PFVICM?
When thinking of Team formations, I usually envision ways in which four drones protect the one with the radar. It is my understanding that power earned by
landing shots against the enemy go directly to the non-drone robots (I don't know how / if they are divided if you have more than one non-drone), meaning the
drones never get recharged. This has led me to consider having the radar robot acting as a ram bot and setting up easy shots for the drones targeting the
same opponent which has few movement options. If the radar bot doesn't get too ganged up on, the drones should be able to keep it fed with energy, and not
much is going to help the opponent facing ram damage and 5 cannons. The tricky part would be leading your rammer enough so that you don't hit it, while not
too far that you miss too much. Assuming the ram victim tries travelling directly away from the rammer his motion should be pretty predictible, since
anything else will cause him to be hit. You could try the same tactic with a drone but you would be more likely to lose the drone, since it is not being fed
energy. You'd definitely want to rotate through drones after each kill to balance the damage out. -- Martin
Unless it has changed recently, it has been that the robot that had a bullet hit another robot regained the energy. But it is still a good idea to try to
protect the one (leader) with the radar. -- Kinsen
I was wondering if there are archives of this year's rumble rankings because I wanted to see how my bots were since I have taken them out. -- Kinsen
The rankings are not archived anymore, but the results of your bots are not lost. If you select the details of any bot, and then manually edit the url to
reflect your bots name, you can see their rating. You could also (temporarily) enter them again. Are you planning a new bot? --GrubbmGait
Yes. I was wondering about how well my various targeting/movement attempts worked because they are an easy group of robots to test with since I know exactly
how they are supposed to work. -- Kinsen
continiuosly used. Java IO is huge. Is logging still turned on in the uploaded versions of these bots? --Loki
Sorry, normally Slarti is slow but does not print that much, but there is something wrong with V. 0.5. IŽll remove debug messanges completely in the next
version. If Slarti is to slow you can remove him from the rumble. I start running melee rumble battles now, so everything shoul be faster now. --Krabb
David, is that nevermind because you figured out what was wrong? I've had my 2 clients running all day, and I'd like to know if something is wrong with one
of them. -- Voidious
I am having trouble with two robots:
Both of them are reported as corrupted. -- Kinsen
I'm still having trouble with nanoStep1.0:
Could not download bot nanoStep1.0.jar Ignoring nanoStep1.0.jar: .\nanoStep1.0.jar (The system cannot find the file specified) Ignoring nanoStep1.0.jar: .\nanoStep1.0.jar (The system cannot find the file specified)--Kinsen
The new version of CassiusClay is having trouble:
Iteration number 10 Preparing battles list ... Using smart battles is true Preventing unknown thread Application Thread from access: (java.io.FilePermission .\pez.rumble.CassiusClay_2pi.05.jar read)It only happens sometimes but it has this error sometimes after starting a new iteration. -- Kinsen
Uh, strange. Could this have to do with CC trying to fetch data from the disk. That should only happen at the beginning of the first round though and this
seems to happen before that. Anyone have a clue? -- PEZ
Doesn't sound like CC reading data, but I don't really know what it is. I can tell you that I've been running 2 RoboRumble clients for a couple of hours and
haven't seen that error yet, though. I've been running another remotely for a while now, and while I don't have the scrollback there to verify it, I haven't
noticed it on that one either. What OS and Java version are you using, Kinsen? -- Voidious
I've also been running 2 clients from two different locations for several hours. I am using Windows XP and Java 1.5.0_08. I started noticing the error on the
3rd iteration. -- Kinsen
The only thing that comes to mind is if both clients were accessing the same .jar, but you said they're in different locations... Hopefully someone else
running RR@Home on Windows can shed some light, because mine are running on Mac and Linux. -- Voidious
Is it okay if I run the client again because I think that the problem was the missing Dsun.io.useCanonCaches=false. -- Kinsen
It's fine by me. I seem to recall one of the questionable things being identified as a buggy bot. -- Martin
Yeah, I'd say give it a shot. I'm sorry if I blamed you for something that turned out to just be a buggy bot, dude. -- Voidious
Though initially melee battles were my favorite part of Robocode, I've really lost a lot of interest in them. I think the main reason is the scoring. It
heavily favors waiting out your opponents rather than trying to harm them. While that may be a realistic strategy when applied to a real world brawl, is it
fun?
I suggest trying a new score adjustment in the robocode client that gives (damage * number of opponents remaining) points for bullet and ram damage, and a
similar boost to the bonuses for finishing someone off.
I realize this will have a big impact on the standings, but it seems to me that there isn't much interest in the competition there right now anyway. Maybe
it will spark some interest. I know it would renew mine, not having to turn Ugluk into a pacifist in order to get ahead. -- Martin
A pacifist will never score well in a melee battle! It's not possible to survive in a melee without constantly hitting others to leech energy. No movement
strategy can make you dodge all the bullets flying around. -- ABC
And yet SittingDuck, the last place bot who fires no bullets and doesn't even move, has an average of 12.67% of the points between himself and each of his
opponents. It's hard to see bot improvements reflected in your rating when so much of your opponents' scores are out of your hands. Granting more weight to
damage, something you have more control over, would improve the relationship between your bot's personal performance and his rating, in my opinion. -- Martin
Well, the score distribution in the melee rumble is inherently different than in 1v1, too, with the top bots in the 1700's. Clearly SittingDuck's pacifist
style still lands him in last place, no? As for scoring changes, I think it might be better to consider them for Robocode 2 (or whatever Fnl calls
RC's successor), as there are already so many melee bots that were designed with the current scoring in mind. -- Voidious
Although I personally like the more agressive style (like DoctorBob or Gruwel), a meleefight should reward the last bot standing. If that is only
possible with evasive manouvering, so be it. Rewarding only 50% of the participants (5 in melee and 1 in one-on-one) could be considered for RC's successor.
If you really like agressiveness, and can bring yourself to write a bot with C-style code (like me) to meet the code-restriction, consider the TwinDuel
competition, where being agressive does pay off. -- GrubbmGait