November 21, 2003 - I have been continuing to re-work Jekyl's gun in an effort to develope a system that I think is fast enough to go without pre-loaded data on enemy bots. As such I have been doing a lot of tests using the TargetingChallenge. I think I now have a system that works relatively well. It scores relatively close without data over 35 rounds to what it scores with data collected over 500 rounds. Funny thing is it has done that almost from the start. Yet another subtle clue that Robocode is more about good to great movement. In any event, I have returned to using a RollingAverage in my data system. I have started to write a system to intelligently decide if it is neccessary to store data on a given opponent. Once this has been committed to CVS, I will once again be back int he movement labratory. ABC's new Shadow has wall smoothing done fairly well and it seems to make a pretty big difference in it's performance so I think the time has come to finally integrate the system that I have developed into Jekyl. I think I will delay the release of a new Jekyl after the repository comes back a few days to allow all the clients to catch up with the flurry of new bots that I am sure we will see. Hope it comes back soon. I am starting to suffer withdrawl :)
18 November, 2003 - Well I have a Jekyl .62.1 release packaged and ready for upload but the repository is down for a server switchover. Very frustrating. The immediacy of the whole RR@H architecture makes you start to take these things for granted. For me this was to be a very radical approach. This version of Jekyl had only been tested againt one other bot: Jekyl! I have got to the point where I am confortable movement against the previous release version. All this bot had was about 5000 rounds. I was even going to re-use the data file from the previous release. Please come back soon repository.
13 November, 2003 - 0.62 development has progressed nicely. I have been running some focused tests vs the RoboRumble top 11 (minus VertiLeach PEZ as I know the answer there =^P) and have been getting some suprising results. Jekyl consistently finishes in the top half of that grouping (averaging 4th). This suprised me some as I have only made one very subtle change to the targeting and one small survival related change at this point. I think I am going to hold off on integrating a buch of new guns into Jekyl at this point in favor of releasing a new version to test these results. Look for the release of a new Jekyl as early as tomorrow sometime. Curiousity is killing me on this one.
10 November, 2003 - Jekyl 0.62 development has started in earnest today. I am refactoring the guns completely and am going to try to implement a VirtualGun? array. I got a pretty good hint at an implementation that I think will work so I am even off to a good start. I am also in the process of implementing a wall smoothing algorithm that I hope will cut down on the wall bounce that some of the top bots seem to pick up on. With any luck we will see a new Jekyl before the week is out.
26 Oct, 2003 - Jekyl 0.60 released. This version features a refactored gun which now uses Waves. With a little help from a friend I was able to finally track down a bug in my targeting code that was making my gun score in the very high 60's to low 70's in the TargetingChallenge. As I posted over there, I am now getting into the mid-80's. I also added some quick and dirty fixes for some lurking NullPointerExceptions? that were waiting for me in the LongThinLeague. Here's hoping.
25 Oct, 2003 - I was running tests on the development of Jekyl's new gun I decided to run this one just for fun.
Rank Robot Score Survival Last surv. b. Bullet dmg. Bullet dmg. b. 2*Ram dmg. Ram dmg. b. #1st #2nd #3rd 1 Jekyl 8121 2500 500 4265 853 2 0 50 0 0 2 SnippetBot 1.0 530 0 0 530 0 0 0 0 50 0For me this was just a nostalgic moment as burried underneath all of the cruft and hair brained addiditions that I have built into Jekyl, there is a SnippetBot crying at the abuse it has been subjected to. On the development front I have the Wave system working (I think) and in testing now. If this pans out I am going to try and modify it for use as a melee gun (it just another segment after all ;)) and maybe work on a decent melee strategy for once.
23 Oct, 2003 - Jekyl's gun is undergoing a major re-write. I am getting close to finalizing the details. Still looking for one or two more bugs that are holding down it's performance. This new gun will be a Wave based system with melee awareness. I will finally be able to throw away my old Circular Targeting gun I use in melee :)
2 Oct, 2003 - Jekyl .55 is in final testing. It should hit the streets in the next few hours. This version has had the movement system reworked to produce a very flat movement profile as seen by just about everyone but GloomyDark. Even Gloomy's is a bit flatter. Look for it around 2:00PM EST if you are interested.
18 Sept, 2003 - Jekyl .53 hits the streets.
17 Sept, 2003 - The next release of Jekyl may be delayed a little bit as I have discovered a bug in his stats processing code.
11 Sept, 2003 - Jekyl is back in the movement laboratory. I am not happy with his performance in the RoboRumble so it looks like it is time to make a change. To add insult to injury, I found an error in my data saving routine which garuntees that every battle I start, I begin with essentially n rounds of data where n equals the number of matches between me and my target. This was sub-optimal to say the least. Look for a new Jekyl some time within the next couple of days.
1 Sept, 2003 - Well at long last the rumble has started again and I have a fire in my belly to improve. Jekyl .50 is in final testing. I have found a few errors in the gun and corrected those. I have revamped the movement system with a much simplified system tested in BlackPearl. I can now finally beat all the PatternMathcers? in my immediate vicinity in the Rumble rankings. I have revamped the AntiGravity movement system used in melee and added a melee radar dwell option to hopefully improve the targeting. Melee still has not recieved the attention it deserves and I think that will be the focus of the next release. I have also done a little housekeeping and removed a lot of old Classes that I was no longer using and refactored some of the ones that made the final cut. Look for the new Jekyl some time later today.
4 July, 2003 - I will be on vacation in California for the next 17 days. I tried to rush out an update this week but was not really able to test it with everything I had going on. I am disappointed as I wanted to get it out so badly. I have made a bunch of changes, some of which are actually for the better. Perhaps the time away will help clarify some of the mistakes I am making and I can release a new Jekyl after I get back. In the mean time I wish you all luck. Some one please save a copy of the rumble page for me to see should some one some how manage to knock DT off the throne.
29 June, 2003 - Well the update of the rumble seems to have helped me slightly in melee. The removal of so many classic bots saddens me but if it means faster updates I suppose I can live with it. In development news I continue to struggle to beat decent pattern matchers. I have tried varying my velocity, direction, random stops, closer average target distance, farther average target distance, every variation on bullet power I can imagine, sacrificial offerings, and anything else that has a remote chance of helping. All of this to no avail. I can manipulate my movement profile to a pretty flat offering (not that this will help vs a pattern matcher) but in the end I think it is a good thing that I have vacation coming up in a week. I am so frustrated it is not funny. My trusty ThinkPad? 600 has bourne the brunt of my frustration. I have deleted my whole code base several times in removed robocode from my computer only to re-install it again the next day. I want very badly to get over this next hump of bots (to about position 15 - 18 or so). There are a few bots below me that I would not have to face which are dragging me down. I would also get to face a few of the top 15 bots more often, some of which I fare better against.
15 June, 2003 - Just a quick update to let everyone know that Jekyl is still alive and kicking despite the fact that there has been no new release in a while. I have totally ripped Jekyl's movement apart. Quite frankly I am suprised that he made it as far as he did with the movement he had. There where a couple of glaring errors in some of the logic and distance calculations that were simply wrong. Evidently I had programmed around all of them and correcting them broke it's movement pretty badly. I have fixed the WallAvoidance mechanism, straight line distance to wall calculations, and all of the anti-cornering logic. I am now in the process of returning to 0.41 movement functionality as well as some curve flattening. I think I will also put in an overhaul on Jekyl's Antigravity movement system to return Jekyl to the top 64 in melee. Time permitting I may also add some factors to my guns and maybe put in a VirtualGuns system.
11 June, 2003 - Jekyl slid back a bit this week in the Rumble. This was expected as many bots at this level save data and not too many of then have seen me yet. Jeykl's melee rating took a big hit with the scoring correction too. Not all that good of a week for Jekyl. In development news, I have been working on flattening my movement profile quite a bit. I have finally suceeded in getting it quite flat. There is still some room for improvement but I am going to set that aside for a bit as I have another thing I would like to work on: beating a PatternMatcher.
31 May, 2003 - Jekyl continues it's 1-v-1 rise through the Eternal Rumble with the 0.41 release. As expected it's melee performance dropped off slightly. It also took 1st place honors in the semi-pro division at the Outpost competition as well. I think I am going to take a break from jekly to work on a mini with a Poison movement system. I would like ot extend a huge THANK YOU to PEZ. With out the wiki this all would have been so much more difficult. I think you can start to put yourself up there with the giants of Robocode. Your contributions to this obsession are truly amazing and very welcomed. Thank you very much.
24 May, 2003 - The 0.40 version seemed to be a success. I was able to actually qualify for the Champions League. I also promptly got splattered. At least I did not have high expectations. Thanks to all of you who have helped me to progress this far. A special thanks to PEZ for this excellent wiki and to kawigi who patiently answers my never ending robocode questions. I hope to return the favor to someone else some day --jim
18 May, 2003 - I uploaded Version 0.40. This one has a slight change in 1-v-1 movement. I completely refactored the Antigravity movement for melee. I am hoping with this change to jump the remaining 10 spots to get into the Champions League. I plan on promptly getting slaughtered there but but who knows.