September 29th 2001
This month no more veterans were thrown out of the arena. There is only the fail of the young Dragon (a never refined robot of mine that followed an interesting path) and the youger Pirhana (in spite of its name, its performance is far from the one of the strong Fish and the weak Sharks). A more thought version of Pirhana will get better results, if its author doesn't surrend so soon.
There are two surprises in the higher positions of the list: Two brand-new robots appeared in the first six placements with nice percentages. They performs well in only one type of play. In single matches Schmamageddon shows a good 63% but is very weak in other types of play because its author haven't implemented yet a friend identification algorithm and his robots fire each other. In team matches OneEyedWilly got an incredible 54% (only 3.5% away from KillerBees and better than Myst - the weaker clone of KillerBees) and it gained the second place of the list. With a lucky tournament OneEyedWilly could replace KillerBees without any change in its code: We are waiting for such an event so long :)
A bad news for me is that bdj1 (a clone of the old Platoon) performs better than my new Platoon2. This latter robot runs at 100% speed all around the arena as Platoon do, but it adopts an aim correction algorithm to foresee the positions of the enemies when the missile arrives on the target; since bdj1 performs better, I think that this kind of correction (or at least my implementation) is not so useful in an arena where the majority of robots follows an irregural path :(
The robots fought a total amount of 10434 battles, a respectable number for a tournament of only three weeks. The reason for such a result is the large number of people that visited the Battle Arena page in the last weeks. I've recently discovered a game similar to Jrobots sponsored by IBM: this game is Robocode and it is has a nice graphical interface but still no on-line tournaments. I've posted news about Jrobots on their server and a lot of players wanted to make the comparison but none of them sent any robot yet :(
Robocode is turn-based instead of thread-based like Jrobots, but it seems that even that game suffers of some kind of machine dependency; in fact, there is a onSkippedTurn() method to warn robots that the execution of their main cycle takes too much time and they are losing turns :((
To see the results read the Challenges page.
Here is an abstract:
|Single Winner||KillerBees||by MontyFish||92.75%|
|Double Winner||KillerBees||by MontyFish||91.19%|
|Team Winner||KillerBees||by MontyFish||57.61%|
Now you can upload your robots to win the October 2001 Jrobots Challenge (October 6th-27th 2001) aka The Raise of the Novices.
September 8th 2001
The September 2001 Jrobots Challenge starts now.
Remember that you can send your robots even in the course of the challenge, so you're not too late. The latest robots are stored in the file jjrobots_challengers.zip. You can download it to test off-line the behaviour of your robot against the other challengers.
To read the results in real-time simply visit the Battle Applet page
Good luck to all the players!