An Interview with Roger, the author of roger,
the current Master of the GUN-TACTYX Championship

I have the unique opportunity to interview Roger, author of roger, the most effective team ever seen in the GUN-TACTYX arena. Roger's warriors are the best performers thanks to their precise aiming technique that tranforms straight walkers in targets very easy to hit.
Let's know more about Roger's trade secrets and his ideas about the future of GUN-TACTYX tournaments.

leo: Roger, the results that you achieved in the last tournament are incredible! Your team got an overall percentage of victories equal to 96.7% losing only 11 matches over 330. Can you explain us the main idea that is at the base of the behavior of your team of bots?

Roger: Hi leo! First let me thank you for this great game!
Anyway, the first thing which came on my mind when I saw it for the first time was: They don't predict! How can they hit anything?
So the first idea was to make bot which can predict. And prediction of the oponents position is the main reason my bots can win.
The other reason is that I spent lot of time with the game (two weeks, mostly evenings) so my bots have only few bugs and I think I have some know-how now.

leo: As you said, one of the strong points of your team is its advanced aiming technique. Can you tell us more about it? Of course, I don't want that you share your trade secrets with us, simply describe two or three ideas that a player should consider to improve the aim of his bots.

Roger: It seems many have tried and many have failed. In my opinion, the reason is that all this rotation problems are complex and hard to debug.
When I made the first version of my prediction, it did not work. But the problem was not in the prediction itself. Problem was that I watched and shoot when body was rotating. Another problem was that the angular variables can overflow over 2*PI. Small problem in head turning can lead in loosing target too often. Many small things. The prediction itself is the most simple, one line code, and it is only angle-based. It uses 2 consequent watches to estimate distance and relative angular speed. That's it, I'm not gonna explain it any more. I tried also more complex cartesian coordinates based prediction, which leads to quadratic equation, but it is more sensitive to small changes of target's direction and speed, so I don't use it.
So my final advice is: make only small steps and debug everything. Use console on F3 and printf. Make stupid bot which doesn't shoot and use it as a target. First aim while not moving. Don't get stucked with 'tuning' constants. And one universal: there is always one more way how to do something.

leo: In the description of your bot you claim that radio communication among team mates is not important. Do you mean it's not important in general or only in a fast paced enviroment like the Plaza level is?

Roger: I mean in Plaza, anyway I'm trying to implement some comunications for the last 'clean-up' phase of the game. And in soccer game-type it will be very important. In Plaza there is problem at the beginning. Game starts and bullets are instantly flying toward start-up positions. No time to loose, you must run run run. And kill, if you manage to. But comunications? What for?
Some bots use 'hear' to get position in start-up crowd. But (as you surelly know) that is not needed. The crowd is completely regular. The position can be computed from ID of the bot.

leo: I know, but I think that was a nice example to show how hearing works ;)
In a post on the forums you already show a certain interest in competitions in the Maze level. Have you in mind a list of possible fighting techniques that should be taken into account for good performances in the Maze?

Roger: You must have better wall detection. You must prevent your bots getting stuck into corners or other bots. You will have to pick those nice items lying next to start-up position, because your oponents will do it and full armor and full health makes difference.
I would like to try to shoot grenades over the walls and I'm playing with bots on upper floor - but this doesn't seem to be of some strategic importance, just fun.

leo: I think that roger (the bot) will not meet any serious competitor for months, so this question may be very premature, but... have you any improvement in mind for your team?

Roger: As I said, I'm working on better clean-up phase. Some smarter wall avoidance and at least basic item-picking should improve my performance in Maze while not affecting performance in Plaza.

leo: This is a question that I usually make to players, but in this context may sound amusing because it's difficult to find anything interesting is opponents that one always beats: Among the bots of the other competitors did you like one in particular? :)

Roger: I-bot, of course. He does half the things I do and some of them he does better. He was always my enemy number one. Gunny beats me often, so he is the second one in my default options.ini file. Krobot is the third, because he is the second best. But he appeared after my first repository upload (same day) so I'm not so familiar with him.

leo: Thank you for citing gunny even if it does not beat roger so "often" :)
Now a few questions about the game itself. Can you tell us where did you hear about the contest and why did you decide to partecipate?

Roger: My friend told me... he didn't even play it. I wanted to try the prediction and I also wanted to learn new programming language. I'm member of langsmiths group on yahoogroups I'm interested in programming language design.

leo: Do you know any good place where I could post news about GUN-TACTYX to find more and more players for the game?

Roger: I'm gonna post some info to langsmiths. I think there are some people which could be capable of writing bot, could be interested in it and would have time for it.

leo: Thank you very much!
What could you tell us about yourself, your hobbies, favorite books and music?

Roger: I live in Czech Republic, I'm 29, programmer, I work in GIS company. My hobbies are many, the most actual are tai-ji, aikido and quitar. I would also like to mention my Palm OS activities (www.questions.cz)

leo: Do you have any particular suggestion to make GUN-TACTYX a better game?

Roger: Replay (of matches played in tournaments), better documentation, championships in different arenas and game-types. This could be simple.
Random generated bigger mazes, bot design for more distant future.

leo: That you very much, Roger, for the interview and the suggestions. I'm sure that your notes about aiming will be very useful for poor players that lose the 96.7% of matches when there is roger around. Let's see if any player in the next tournaments will upload any teams that can compete with yours.

Roger: I hope my opponents will not be out of ideas or out of free time. I wish them GOOD LUCK!