--- In javajousters@yahoogroups.com, "Leonardo Boselli" <boselli@...>
wrote:

> I tried comparisons a lot of years ago and the behaviors of threads
were
> very different. Now they could be more uniform.

Possibly.  They still seem to be different in various ways which make
using thread priorities very difficult, if not impossible.  see
http://cs-people.bu.edu/sowmya/htdocs/threadsTest.html for some
interesting graphs.  Sometimes the lower priority threads actually get
more cpu time!

In fact, I see that the threading model of the JVM doesn't guarantee
that the thread priorities actually have any effect, and states that
they shouldn't be relied on for any algorithmic use.  Thus any
solution to this problem shouldn't use them.

Also there is no guarantee of the thread implementation using
timeslicing, so it really shouldn't be relied on either.  Thus the use
of yield() in the current implementation is correct and necessary.


>Anyway, because of time
> constraints, I think that, for the moment, it's better to remove the
jtobots
> that slow down the simulation and ask to their authors to fix their
code.

OK.  I see that this has been done already, and that TTL has been fixed.


> In the meantime, if you have modified the code of the simulator
(using API
> compatible with older versions of the JDK), feel free to send it to me
> (please, mark clearly the changes, thank you in advance!)

I don't have anything at the moment, but I may do in the future.  Thanks.


Cheers, Tim.

From: Tim Foden
Sent: Friday, June 01, 2007 6:38 PM
Subject: [javajousters] Re: A couple of possible bugs in the arena applet.
>
> Well, I'd be a shame to lose TTL, as it is quite strong now,
> and has been recently added.

However, its author can upload it again, after fixing the problem about dead
robots.
The number of matches is very important for the significance of monthly
tournament, so it's better to remove any cause that reduce it.

> There may be a simpler solution for TTL's problem though, as it only
> seems to be the dead bots that are causing the problem... if the
> simulation reduced the thread priority of dead bots...
> Does anyone know of any bot that still actually _needs_ to run
> after it is dead? :)

There are only way to stop a thread: put in the code of the thread a flag
that make the execution return; use stop(). The latter method was deprecated
just after I used it in the Jrobots applet. The first method is not
applicable because there is no way to control the player code in this game.
Reducing the thread priority, probably, does not work, just because the
control of threads was (or better was) not reliable in Java. Unluckily I
have no time to try  :(

> In fact, some of the bots do actually exit (return from their run()
> function) once they realise that they are dead, but others don't.
> I guess it wouldn't be safe to kill (using Thread.stop()) the ones that
> don't stop by themselves, as they may leave some shared structure
> in an unknown state...

That is what I do. Anyway, it seems that stop() is not effective.

>  Perhaps it'd be good to add a hint to the rules for
> some future date, that once a bot detects that it's
> dead, it should return from its run loop.

The problem is that such a rule can't be verified, because the sources are
closed. One can discover such a bad habit only when the dead robots enters
some short infinite loop.

> [BTW, I've tried reducing the priority of robot threads that use over
> 80% cpu... I guess we'd need to try this (reducing thread priority)
> on some other platforms too to see if it works on them.]

I tried comparisons a lot of years ago and the behaviors of threads were
very different. Now they could be more uniform. Anyway, because of time
constraints, I think that, for the moment, it's better to remove the jtobots
that slow down the simulation and ask to their authors to fix their code.

In the meantime, if you have modified the code of the simulator (using API
compatible with older versions of the JDK), feel free to send it to me
(please, mark clearly the changes, thank you in advance!)

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net

--- In javajousters@yahoogroups.com, "Leonardo Boselli" <boselli@...>
wrote:
> For the moment, I think that I could simply remove Kraft and TTL
because
> their behavior eats too much simulation time.
> If no other player, is going to complain, I'll remove them before the
> beginning of the next challenge.

Well, I'd be a shame to lose TTL, as it is quite strong now, and has
been recently added.

Kraft though, hasn't been improved since May 2004, so it wouldn't be a
great loss, IMO.

There may be a simpler solution for TTL's problem though, as it only
seems to be the dead bots that are causing the problem... if the
simulation reduced the thread priority of dead bots (which doesn't
need JDK 1.5), then this may be an improvement.  Does anyone know of
any bot that still actually _needs_ to run after it is dead?  :)

In fact, some of the bots do actually exit (return from their run()
function) once they realise that they are dead, but others don't.  I
guess it wouldn't be safe to kill (using Thread.stop()) the ones that
don't stop by themselves, as they may leave some shared structure in
an unknown state.  Perhaps it'd be good to add a hint to the rules for
some future date, that once a bot detects that it's dead, it should
return from its run loop.

[BTW, I've tried reducing the priority of robot threads that use over
80% cpu, and this makes the problem liveable with on Windows.  The
simulation gets a bit jerky, but the other bots don't seem to be
affected, and the bout now continues, albeit at a slower pace (as the
lowerer priority threads still use 20% or so of the CPU).  I guess
we'd need to try this (reducing thread priority) on some other
platforms too to see if it works on them.]

Cheers, Tim.

From: Tim Foden
To: javajousters@yahoogroups.com
Sent: Thursday, May 31, 2007 7:11 PM
Subject: [javajousters] Re: A couple of possible bugs in the arena applet.
>
> Perhaps we could give people the option of having the old, but
> 1.1/1.2 compatible version, or the newer (with thread usage
> monitoring) version that requires 1.5?

For the moment, I think that I could simply remove Kraft and TTL because
their behavior eats too much simulation time.
If no other player, is going to complain, I'll remove them before the
beginning of the next challenge.

Of course, that is going to happen also to other jrobots that show such an
expensive behavior in the future.

I'm waiting for any comment. Silence will be interpreted as: "Yes, go
forward and remove them"  :)

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net

--- In javajousters@yahoogroups.com, <boselli@...> wrote:
> Anyway it's important to keep the compatibility with older JDKs,
> so I don't think it's a good idea to use an API available only
> from 1.5

Well, I'm afraid I haven't found any way to get the thread cpu usage
data from any JDK version before 1.5 (apart from using JNI, but that
gets silly).

I'd argue that keeping compatibility with old JREs (particularly
Microsoft's one... where support is ending this year, and they
encourage people to move to other solutions -- quote "The MSJVM will
reach its end of life on December 31, 2007" -- see
http://www.microsoft.com/mscorp/java/) is not as important as it once was.

Perhaps we could give people the option of having the old, but 1.1/1.2
compatible version, or the newer (with thread usage monitoring)
version that requires 1.5?

Cheers, Tim.

--- In javajousters@yahoogroups.com, "Alan Lund" <alan.lund@...> wrote:
>
> One possibility that you might want to keep in mind when
investigating this
> problem and its possible solutions is that the slowdown caused by some
> robots may be due to excessive creation of temporary objects
(JJVector's)
> and the correspondingly high load that places on the garbage
collector.  If
> garbage collection takes place in a separate thread, measuring the
CPU time
> per robot thread will not help.

OK.  Good point ... goes away and adds a CPU usage window to arena
applet ... time passes  :) ....

Well, this itself doesn't look like a problem.  I've trapped it
happening with both Kraft and TTL.  With Kraft I've seen it in Singles
and Doubles, with a bot that is still alive, and with TTL I've seen it
in Team mode, with a bot that is dead (damage >= 100).  When it occurs
the robot thread concerned uses 100% of the cpu.

Cheers, Tim.

From: Alan Lund
Sent: Wednesday, May 30, 2007 3:15 PM
Subject: Re: [javajousters] Re: A couple of possible bugs in the arena
applet.
>

Hi Alan! It's a long time since your last message.
Have you improved your IonStorm recently?

> Since there is no way to "sleep" in a robot thread, all robot threads
> are trying to get as much CPU time as possible, and I do not think
> there is any way that a single thread (or the group of threads of a
> single team) can hog the CPU such as to slow down the
> simulation... So, I think perhaps you should do some more
> research into the cause before considering solutions..

Anyway the problem does not seem too serious now, but I run very few matches
each time that I connect to the site.

I'll wait and see what's going to happen in the next challenge.

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* APOCALYX 3D Engine: http://apocalyx.sf.net

To read the complete results visit the Hall of Fame page at
http://jrobots.sf.net

During the May challenge, the number of played matches was good enough in
both the leagues.
Three main events happened.

The first is a good news: A new strong jrobot appeared among the cadets. Its
name is Spike and achieved a triple victory with a very good 82%-82%-87%.
We'll see if its valor is going to be confirmed among in the Veterans'
League.

The other news may be good or bad according to the point of view: Tango
failed its usual triple victory, in fact it won only in Double category,
while StrangeMatter got both Single and Team victories. Sometimes it happens
and testifies that, in some situations, StrangeMatter and Tango are very
close, but usually Tango prevails.

The final news is a bad one, at least for me: Platoon failed to survive.
Platoon was the oldest of my jrobots, the one that you know because included
in the samples, so I'm not so happy to see it removed. Anyway the time has
gone for these linear jrobots that run in a straight line at constant speed.

The results of the Veterans' League are:
Survivors: GulleFjun, RoboOlm, KillerBees (for their results in Team),
Hellfire, Bohan (for their results in Double), Earthworm (for its result in
Single), Mouth (for its result in Double and Team), Headhunter (for its
result in Single and Double), Stately (for its result in Single and Team),
Doom, TTL (for their results in Single, Double and Team), IonStorm (4-3-3),
Starkle (3-4-3), Tango (2-1-2), StrangeMatter (1-2-1).
Demoted jrobots: Vampire, Kraft, Ladybug, Errata, DirtyCoward, FlameBall.

The results of the Cadets' League are:
Promoted jrobots: 845bytes, AronsRobot02, DarkSoldier, Berserk, Jimbo,
Snutteplutt, Yoda, Bosozoku, Dragon, ELECTRON, Janeway, Hal9000 (5-6-11),
Teddybear (3-2-2), Spike (82%-82%-87%).
Survivors: NeoBot, Genuine, GenuineA, Phalanx, Ender, Battlemage, Touche,
Stinger, AronsRobot01, Shark, MarukoV4, Ciaky, RauaRoobert, DeathIncarnate,
DeadDuck, CounterStrike, Shrike, Bizarro.
Removed jrobots: Platoon.

Now you can upload your jrobots to win the "June 2007 JROBOTS Challenge"
(June 2nd-30th 2007) aka "Let's Spike Tango's Gun!" alias "A Farewell to
Platoon!"

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* APOCALYX 3D Engine: http://apocalyx.sf.net

From: "Tim Foden" tim_yahoo@...
Date: Wed, 30 May 2007 07:33:32 +0200
To: javajousters@yahoogroups.com
Subject: [javajousters] Re: A couple of possible bugs in the arena applet.
>
> Looking at this a bit further, it may be possible to fix this by
> monitoring the amount of cpu time each thread has been using...

It's a good idea. If you, or someone else wants to implement this rule in the
code of the simulator, I'll be very happy to add it to the official applet.
Anyway it's important to keep the compatibility with older JDKs, so I don't
think it's a good idea to use an API available only from 1.5

> we could change it's priority to be lower, until it's back in the fair range.

I'm not sure that priorities work very well on native threads (at least they
work very badly at the old  times of JDK 1.2, but things may be better now).

Finally, this afternoon (or tomorrow morning) I'll publish the official results
of the May tournament. Then another challenge it's going to start: The June 2K7
JROBOTS Challenge, aka "Let's Spike Tango's Gun!"

--
Ciao, leo
JROBOTS Java Programming Battles
http://jrobots.sourceforge.net

--- In javajousters@yahoogroups.com, "timfoden" <tim_yahoo@...> wrote:
>
> But I guess it would be nice to sort it out in the applet.  Maybe by
> timing the matches, and if one takes too long (say > 10 mins), it
> could stop the match and try another.

Looking at this a bit further, it may be possible to fix this by
monitoring the amount of cpu time each thread has been using.  But the
facility to find out how much time a thread has used (in ThreadMXBean)
is only available with java 1.5 onwards.

It would be nice though, as we could enforce fair use of cpu time by
the bots, without changing them at all.  If we found a thread using
too much time (compared to the others, or compared to some limit per
virtual clock tick, or by some other method, etc.), we could change
it's priority to be lower, until it's back in the fair range.

Cheers, Tim.

--- In javajousters@yahoogroups.com, "Leonardo Boselli" <boselli@...>
wrote:
>
> Welcome, timfoden!
>
> Your Spike seems very promising with its 82%-82%-87% in the Cadets'
League.

Thanks.


> I'm curious to see it at work among the veterans. Tango has been
waiting for
> another serious challenger for a long time.

Well, I'm not sure about beating Tango. :)  But I'll see what I can
come up with.


> > Thirdly, some of the bots seem to "lock up" the simulation
>
> Do you think that those jrobots use the
> infinite loop technique to slow down the simulations they are going
to lose?
> In that case I can ban them, but in my tests does not seem so.

I don't think any of them are doing it on purpose.  However, it's a
pain when it happens.  If I run the Veterans League CPU contribution
at the moment, I rarely get 20 matches before it "locks up".  Leaving
it overnight (about 8 hours) it only gets to about 150 matches played.

I was just pointing it out in case there are any solutions.  At least
TTL is still being developed, so in it's case the problem could be
cured.  The problem hardly ever occurs with Starkle, so I think in
it's case the problem could be ignored.  I guess the biggest problem
would be with Kraft.

But I guess it would be nice to sort it out in the applet.  Maybe by
timing the matches, and if one takes too long (say > 10 mins), it
could stop the match and try another.

Cheers, Tim.

>
> Thank you for your great Spike! I'm curious to see its performances
in the
> next tournament.

Me too.  :)


Cheers, Tim.

From: timfoden
Sent: Friday, May 25, 2007 2:41 PM
Subject: [javajousters] A couple of possible bugs in the arena applet.
>
> I'm new here, but I've been mucking around with
> JRobots for a couple of months or so now
> (I'm the author of Spike BTW)

Welcome, timfoden!

Your Spike seems very promising with its 82%-82%-87% in the Cadets' League.
I'm curious to see it at work among the veterans. Tango has been waiting for
another serious challenger for a long time.

> and I think I've found a couple of bugs.
> First, in the JJGraph class. At line 114 the maximum number
> of games value is calculated... but there's a bug in it:
> Here's the line (JJGraph.java, 114):

I agree. I'll update the sources as soon as possible. In the meantime,
players can fix the sources themselves . Thank you!

> Second, in the advanced applet only, the initial speed
> value is set to 2 initially.
> Here's the line (JJArena.java, 83):

I agree also here. The same as above applies.

I wrote neither JJGraph nor the advances applet... but I wrote very badly
the simulator, so I have no excuses  :)

> Thirdly, some of the bots seem to "lock up" the simulation
> sometimes. By "lock up" I mean that they make it run very slowly.
> I guess they go into some loop that doesn't cause a
> Thread.yield() to be called.

This may happen when a jrobot enters an infinite loop without calling any of
the API for a long time. Probably Kraft, TTL and Starkle experience a
similar problem.

Usually malicious jrobots that use infinite loops are easily detectable
because they can't finish any match. Do you think that those jrobots use the
infinite loop technique to slow down the simulations they are going to lose?
In that case I can ban them, but in my tests does not seem so.

Thank you for your great Spike! I'm curious to see its performances in the
next tournament.

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net

Hi All.

I'm new here, but I've been mucking around with JRobots for a couple
of months or so now (I'm the author of Spike BTW), and I think I've
found a couple of bugs.



First, in the JJGraph class.  At line 114 the maximum number of games
value is calculated... but there's a bug in it:

Here's the line (JJGraph.java, 114):
	   max = Math.max(Math.max(graph[number][0].length()/2,
graph[number][1].length()/2), graph[number][2].length())/2;

At the end of the line it reads "))/2;" but it should be ")/2);".
It's subtle, but it means that if you've only got singles or doubles
results, then you only get to see half of the graph!


Second, in the advanced applet only, the initial speed value is set to
2 initially.

Here's the line (JJArena.java, 83):
private double speed = 2;

This isn't too much of a problem I guess, except the default setting
in the combo box later on is still left as "Official", and not changed
to "x2".



Thirdly, some of the bots seem to "lock up" the simulation sometimes.
  By "lock up" I mean that they make it run very slowly.  I guess they
go into some loop that doesn't cause a Thread.yield() to be called.  I
measured it as about 1 simulation second in about 10 real minutes on
my machine.  Perhaps this is a possible reason for the low number of
results that get generated sometimes.  I've observed the problem now
with 3 bots:

1.  Kraft.

2.  TLL, but only the most recent version.

3.  Starkle, but it happens only very rarely with this bot.  I think
this is livable with.

Cheers, Tim.

To read the complete results, visit http://jrobots.sf.net

Finally the March-April tournament has reached its end. The extended
challenge was motivated by the fact that during the first month the number
of played matches was not enough, but after another full month of matches
that number became acceptable.

Among the veterans, the list confirms the usual results, except in the
unexpected event that Tango failed to achieve another triple victory.
Sometimes happens, but I'm sure that Tango is going to repeat its positive
sequences very soon.

The survivors are: Mouth, Errata, HellFire, Headhunter, for their results in
single, DOOM Bohan, KillerBees, Stately, for their good results in all the
three categories, IonStorm (4-4-3), Starkle (2-3-4), StrangeMatter (3-2-1),
Tango (1-1-2). Compliments to StrangeMatter for its 64% in team category.

The demoted jrobots are: Berserk, MarukoV4, 845bytes, Shark, Jimbo,
DarkSoldier, Bosozoku, Hal9000, Janeway, ELECTRON, Teddybear.

The cadets' tournament demostrates a time more that good results in single
category are not enough. Good performances in single are not awarded because
of very bad performances in the other categories: It's very important to
distinguish enemies from friends, so read the tutorials and learn how to
avoid friend fire.

The promoted cadets are: TTL, RoboOlm, Earthworm, FlameBall, Ladybug,
DirtyCoward, Vampire, Kraft, GulleFjun. TTL is going to test again its valor
agaist the long-aged veterans. I hope that it's going to behave better.

Survivors among the cadets are: Shrike, Bizarro, CounterStrike, DeadDuck,
Yoda, RauaRoobert, Dragon, Snutteplutt, AronsRobot01, Ciaky, Phalanx, Ender,
DeathIncarnate, BattleMage, Touche, Platoon, GenuineA, Genuine, Stinger,
NeoBot.

Only PaulBot was removed from the arena because of its poor performances.

Now you can upload your jrobots to win the "May 2007 JROBOTS Challenge"
(May 5th-26th 2007) aka "TTL is Back!"

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* JROBOTS Java Battles: http://jrobots.sf.net

February 2007 JROBOTS Challenge
To read the complete results visit
the "Hall of Fame" page at
http://jrobots.sourceforge.net

The tournament of February shows a reduced number of played matches in
comparison with the last challenge. As usual the low number of matches gives
unreliable results, but this time I consider the percentages good enough to
apply the tournament rules for promotions and demotions.

Among the veterans, the results confirm the usual relative values. The
survivors are: Headhunter (for its results in single), KillerBees (for its
result in team), Hellfire, Stately (for their results in single and double),
Mouth, Bohan (for their results in double and team), DOOM (for its results
in all the categories), Starkle (3-4-4), IonStorm (4-3-3), StrangeMatter
(2-2-2), Tango (1-1-1).

The demoted jrobots are: Vampire, Ladybug, Kraft, GullFjun, FlameBall,
RoboOlm, Earthworm and TTL, a new jrobot with good results among cadets that
suffered the competition of the long-aged veterans.

The cadets' tournament suffered of the low number of matches, thus we can't
fully trust the final results. Anyway, I decided to remove, for their poor
percentages, Leech and Ajax, below the 25%-25%-12.5% in two categories. 20
jrobots survived and the following 12 jrobots were promoted: Janeway, Jimbo,
845bytes, ELECTRON, Bosozoku, MarukoV4, Teddybear, Errata, DarkSoldier,
Berserk, Shark and Hal9000.
A special mention to the latter jrobot: Hal9000 comes from the long
tradition of Crobots, the game of which Jrobots is a clone. Let's see if the
Crobots' tactics can put new ideas in a quite different game (in timings and
programming) like Jrobots has become since the beginning of the century.

Now you can upload your jrobots to win the "March 2007 JROBOTS Challenge"
(March 3rd-31th 2007) aka "The Odyssey Continues!"

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it

The complete results are available at
http://jrobots.sourceforge.net

A long time has passed since a cadet was promoted with good results in all
the categories at its first attempt, but finally it happened again: TTL
succeded in the enterprise of being promoted with a 5th-4th-3rd place and
high percentages of victory. The next month it's going to show its value
among the long-aged veterans.

The other promoted jrobots are: Stately (1-1-2), Headhunter (2-2-30), Bohan
(3-3-1), Earthworm, Vampire, Ladybug, Kraft.

The survivors are: NeoBot, Leech, Shark, Genuine, GenuineA, Touche, Ajax,
Platoon, Battlemage, 845bytes, AronsRobot01, Phalanx, Stinger, DirtyCoward,
RauaRoobert, Ender, Yoda, Ciaky, CounterStrike, DeadDuck, Shrike, Bizarro.

Finally the removed jrobots are: Jedi, Hal9000, Sniper. They are plain
copies of robots written for Crobots (the game that is the main source of
inspiration for JROBOTS). Their performances are not satisfactory because of
the differences in the simulators of the two games, but I'm sure that the
player who upload them is working on a version written specifically for the
next tournament.

On the Veterans' side the best positions are always occupied by Tango
(another triple victory), StrangeMatter (2-2-2), Starkle (3-4-4), IonStorm
(4-3-3). Then follow DOOM, Hellfire, RoboOlm, KillerBees, Mouth, GulleFjun,
FlameBall.
All the others were demoted: MarukoV4, Berserk, DarkSoldier, Jimbo,
Snutteplutt, ELECTRON, Dragon, Bosozoku, Errata, Janeway, Teddybear. It's
now their turn to make a bit of selection among the cadets.

Now you can upload your jrobots to win the February 2007 JROBOTS Challenge
(February 3rd-24th 2007) aka Crobots Rules! http://jrobots.sf.net

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it

The complete results are available at
http://jrobots.sourceforge.net

Finally the cadets' league played enough matches and the results are
acceptable. Because of the large number of veterans demoted in the last
tournament, we had a very crowded cadets' league with 50 jrobots.

Instead the veteran's league had only 7 partecipants and the veterans had a
hard fight to survive. At last only 4 of them deserved to remain veterans:
IonStorm (thanks to its third place in double), StrangeMatter
(third-second-second), Starkle (second-fourth-third) and Tango with another
triple victory.

The jrobots, that suffered the close competition, are: Bohan, Stately,
Headhunter. Now its their turn to decimate the cadets.

Among the cadets, the veterans demoted at the end of the previous tournament
had an easy play. These jrobots achieved the requested percentages for a
promotion: DarkSoldier, Snutteplutt, MarukoV4, Berserk, Dragon, Errata,
ELECTRON, Jimbo, GulleFjun, Bosozoku, Mouth, Janeway, RoboOlm, FlameBall,
Teddybear, KillerBees, DOOM, Hellfire.

These ones still remain cadets: Stinger, Leech, Genuine, GenuineA, Shark,
Phalanx, NeoBot, Ender, Platoon, Touche, Vampire, Ajax, Battlemage,
DirtyCoward, AronsRobot01, Ciaky, Kraft, Ladybug, 845bytes, RauaRoobert,
Yoda, DeadDuck, Shrike, Earthworm, CounterStrike, Bizarro. TTL was uploaded
too late to have its results omologated.

Finally, these jrobots were removed because of their poor results: Trofion,
Sprate, TroSniper, Quake3ArenaBot, Gore2. Some of them were very young, so a
renewed version will be more effective in the next tournaments.

Now you can upload your jrobots to win the "January 2007 JROBOTS Challenge"
(January 5th-27th 2007) aka "JROBOTS 2K7!"

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* JROBOTS Java Battles: http://jrobots.sf.net

The complete results are available at http://jrobots.sourceforge.net

It happened again! The number of matches in the Cadets' League is very low,
so no cadet was promoted or removed. I hope the next next tournament is
going to be more interesting.

Anyway, the Veterans' League performed a great struggle. The played matches
ware enough, especially because of the low number of veterans, to achieve a
great accuracy is estabilishing the relative valor of jrobots.
After so a great struggle, only 7 veterans survived. They are: Bohan,
Stately (for their results in team category), HeadHunter (for its result in
single and double category), IonStorm (which got a 4-3-3 placement), Starkle
(with a 3-4-4), StrangeMatter (with 2-2-2) and finally Tango (with another
triple victory). The next tournament will be even more difficult with only 7
of the strongest Veterans ever seen in the arena.
The jrobots demoted are: Errata, GulleFjun, FlameBall, RoboOlm, KillerBees,
Mouth, Earthworm, Hellfire, DOOM. They are going to meet some weak cadets
(for a total amount of 48 jrobots in the Cadets' League). I hope they are
going to do a bit of selection there ;->

Now you can upload your jrobots to win the "December 2006 JROBOTS Challenge"
(December 2nd-30th 2006) aka "Apocalypse Now!"

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net

--- In javajousters@yahoogroups.com, "cosiniri" <cosiniri@...> wrote:
>
> Even though this month wasn't specially good for it, i'm surprised for
> the exceptional results of StrangeMatter in all categories. It presents
> very solid results in the three categories, with neat movements,
> successful shooting and well developed strategies.
> I would like to know more about this great bot and it's author!
>
> Saludos
> Caos

As the writer of StrangeMatter, I'm happy to see that my bot is being
noticed :)

I went through several different periods when writing StrangeMatter.
At first, I was just writing a new bot for the Cadet's league. Once I
got the hang of JRobots and found some simple but effective strategies
like random movement, linear shooting, and not shooting your friends,
I found it surprisingly easy
to write a bot that achieved a triple victory in the Cadet's league.

But once that had happened I realized that StrangeMatter didn't stand
a chance against the Veterans, which all improved on the simple
techniques I had used. I worked on improving my bot and getting it
ready to do combat with the veterans, and released the new version at
the beginning of StrangeMatter's second month in the JRobots arena.
Its performance was mediocre, but it stayed a veteran. I was out of
ideas, so StrangeMatter sat for many months with only occasional
checkups from its author.

Every once in a while, though, I would get an idea of how
StrangeMatter could be improved. In March I had enough of these ideas
that I decided to take the time to rewrite StrangeMatter and see how
the improved version would perform. I scrapped and rewrote a lot of
old code, and included many new ideas. Some of the breakthroughs I had
were StrangeMatter's new aggressive 700 meter strategy, its current
team strategy, and a new shooting system that I talked about in the
past on this list
(http://games.groups.yahoo.com/group/javajousters/message/766). I also
improved the movement strategy. StrangeMatter's new movement within
700 meters is, in fact, based on Tango's moves. When I was out of
ideas for how to improve my bot, I released the new version into the
arena.

With all its new code, StrangeMatter has been pretty a pretty steady
second-place bot with some triple-second-places. Just today, though, I
got an idea for a way to improve StrangeMatter's estimation of its
opponent's position and movement. Perhaps I'll have another look at my
code and see what could be made better -- if so, Tango should watch
out! :)

--- In javajousters@yahoogroups.com, "hats303" <1hatch@...> wrote:
>
>
> How interesting, my bot won at singles! It can only be because there
> were so many opponents in the veteran's league this month. Starkle may
> be slightly better than Tango against weaker opponents. Certainly
> Starkle cannot hope to beat Tango head to head.  I'm not sure what to
> think of that, except that maybe you don't necessarily have to be able
> to beat the best bot to win.
>

I agree with you. Itīs not necessary to beat a specific robot to win
the game, and that is specially true when the amount of opponents (or
games) minimizes the possibility of head to head battles.
Congratulations for the victory, and for this strong (and surprisingly
small) piece of software.

Even though this month wasnīt specially good for it, iīm surprised for
the exceptional results of StrangeMatter in all categories. It presents
very solid results in the three categories, with neat movements,
successful shooting and well developed strategies.
I would like to know more about this great bot and itīs author!

Saludos
Caos

How interesting, my bot won at singles! It can only be because there
were so many opponents in the veteran's league this month. Starkle may
be slightly better than Tango against weaker opponents. Certainly
Starkle cannot hope to beat Tango head to head.  I'm not sure what to
think of that, except that maybe you don't necessarily have to be able
to beat the best bot to win.

You can read the complete results ib the "Hall of Fame" page at
http://jrobots.sourceforge.net

After two interesting competitions among the cadets, this month no new
jrobots were uploaded and consequently the number of matches is very low.
This is the reason why no cadet was promoted or removed. Let's see if the
next tournament is going to be more skirmished.

On the veterans' league the matches were more varied. The number of fights
reached more than the requested limit to be significative and we can observe
an unusual mixture of positions. In fact, Tango won double and team
categories, while Starkle won in single. Then StrangeMatter interleaved its
results with IonStorm.
The good news is that recently promoted jrobots behaved in a convincing way:
Earthworm and the brand-new Errata survived in the close competition, thus
they are going to struggle in a even closer competition during this month.
Less convincing was Quake3ArenaBot: Its main problem is the poor behavior in
double and team, categories in which the friend fire is deadly.
In conclusion, the survivors among the veterans: GulleFjun, FlameBall,
RoboOlm, Errata, Mouth, Hellfire, DOOM, Earthworm, Bohan, Stately,
KillerBees, Headhunter, IonStorm, StrangeMatter, Tango, Starkle. The demoted
jrobots are: Stinger, GenuineA, Phalanx, AronsRobot01, Shark, RauaRoobert,
Ender, MarukoV4, Ciaky, Quake3ArenaBot, Dragon.

Now you can upload your jrobots to win the "November 2006 JROBOTS Challenge"
(November 4th-25th 2006) aka "Errata Corrige!"

--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* APOCALYX 3D Engine: http://apocalyx.sf.net

LATEST NEWS      October 5th 2006
http://jrobots.sf.net/jjr_contests.shtml

This month we had another interesting competition among the jrobots uploaded
in the two leagues. Again CPU time contributors did their job very well and
the number of matches reached the requested level to be statistically
significative. Unluckily, during the last days of the tournament we had
again two results' files corruption events that, probably, had to do with
too much CPU time contribution or some kind of interference between
contributions and jrobots' upload. Too much CPU time contribution is an
exceptional event for JROBOTS and I have to investigate on the bug causing
the interference. Anyway, thanks to all the contributors.

The veterans' tournament confirmed the relative valor of the jrobots. The
strongest competitors are always the same, but I'm sure to see soon
brand-new effective fighters because the cadets are very active.
The survivors are: GulleFjun, FlameBall, RoboOlm (as usual for their result
in team category) Bohan, Hellfire, Mouth, KillerBees, Stately, DOOM,
Headhunter, IonStorm, Starkle (the last two share the - virtual - bronze
medal), StrangeMatter (silver medal), Tango (gold medal and triple victory).
The list of demoted jrobots: Vampire, DarkSoldies, Berserk, Jimbo,
Snutteplutt, Ladybug, ELECTRON, DeadDuck, Kraft, Bosozoku, Janeway,
Teddybear.

Among the cadets, we assist to the usual change of the guard, in fact 12
veterans were demoted and 12 cadets were promoted. They are: Stinger,
Phalanx, GenuineA, Shark, RauaRoobert, Ender, AronsRobot01, Ciaky, Dragon,
MarukoV4, Earthworm, Errata (the latter is one of the newest jrobots
uploaded).
The survivors are: Leech, 845bytes, Genuine, NeoBot, Ajax, Touche,
Battlemage, Sprate, Platoon, Goro2, DirtyCoward, Yoda, Shrike,
CounterStrike, Bizarro.
Finally, the removed jrobots are: NewComer, Thomas, Counter, Igor. Try again
and you'll be more lucky.

Now you can upload your jrobots to win the "October 2006 JROBOTS Challenge"
(October 7th-28th 2006) aka "The Decimation Continues!"

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* JROBOTS Java Battles: http://jrobots.sf.net

From: "Leonardo Boselli" <boselli@...>
Sent: Thursday, August 31, 2006 3:41 PM
Subject: Version 0.8.9 and DemoPack1
>
> Message for all the subscribers of the
> APOCALYX 3D Engine mailing-list...

Oops! I sent the message to the wrong mailing-list, sorry!

To be in-topic again, I simply remeber that in only two days the
"September 2006 JROBOTS Challenge" (September 2nd-30th 2006)
aka "Decimation!" is going to start.

Remember that you can send your robots even during the challenge,
so you're never too late. To read the results in real-time simply visit
the Veterans'  Battle Applet page or the Cadets' Battle Applet page
at http://jrobots.sourcefirge.net
--
Ciao, leo boselli@...
* JROBOTS Java Battles: http://jrobots.sf.net

Message for all the subscribers of the
APOCALYX 3D Engine mailing-list
http://apocalyx.sourceforge.net

The new release of the engine (0.8.9) does not contain yet all the features
recently requested by users, but anyway includes new interesting advanced
functions. Most of them are demostrated as usual in DemoPack1 (available
at the Demos page), in fact these new simple examples were added:

* "3DAccelInfo" lists information about the 3D hardware accelerator
necessary to select the best code for GLSL shaders;
* "Speak" shows the capabilities of the built-in speech synthesizer;
* "Joystick" describes the control interface that manages joysticks;
* "AudioCd" describes the control interface for audio CD playback
on CD-Rom devices;
* "BallsOnTrimesh" demostrates the use of the ODE inner collision detector;
* "Font" is a very simple example that shows how to manage bitmap
and textured fonts;
* "ToLuaDemo" demostrates the use of the tolua++ API to interface
LUA and TinyC code;
* "ObjectsHierarchy" is a very simple example that shows how to
concatenate the transformations of several objects.

The new features of this release were already listed in the last news (see
the post of August 30th). More interesting is to list the next to come
features that will make APOCALYX jump to its major version 0.9.

First of all, I was greatly inspired by the preview of Half-Life 2 Portal
and I absolutely want those "portals" in my engine! My current
implementation resembles more a floating sphere in space than a hole
in the wall and can't support hall-of-mirrors effects yet, but I'm working
on it  :)

Other more traditional features will be:
* render scene and AVI to texture (to implement control monitors)
* render reflected scene to texture (to implement a realistic ocean surface)
* objects picking (to simplify the creation of in-engine editors)
* generation of ODE tri-meshes from heightfields, patches and bsp levels
* a sandbox for LUA scripts (to control execution step by step and memory
occupation, just like actual SMALL scripts)
* X, MS3D and MDL format loaders and animators
* script interface to Crazy Eddie GUI
* realistic sky renderer thanks to Stellarium
* and more... I wait suggestions :)

You can download the runtime of the engine and the demo packages from the
Downloads page http://apocalyx.sourceforge.net/downloads.php

Feel free to ask more about the engine on the forums or by e-mail.

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net

To read the complete results visit http://jrobots.sourceforge.net

Unluckily, we had another unfortunate cadets' tournament this month, in fact
almost at the end of the challenge the cadets' result file became corrupted and
most of the outcomes of the matches was lost. I'm sorry for the problem, but it
was a long time since the last result file corruption that it should be a very
uncommon event.

Probably the even happened because of the large number of cadets in the league.
Since most of then are very weak and survive month after month only for a few
percentage points in a single category, I think it's better to change the rules
and apply them even to this months results. In practice, only the cadets that
achieve a 50%,50%,25% at least in two categories are promoted, and only the
cadets that are below the 25%,25%,12.5% in at least two categories are removed,
except when the result in the other category is above the 50%,50%,25%.

Thank to this new rules a lot of very weak cadets were finally removed and only
the most effective were promoted to veterans. Thus the final results are as
follows.
Promoted cadets: Touche, Stinger, Kraft, NeoBot, Phalanx, DirtyCoward, Ender,
GenuineA, Genuine, AronsRobot01, DeadDuck, MarukoV4, Snutteplutt, Ladybug,
Ciaky, Berserk, Bosozoku, Vampire, Janeway, Teddybear, RoboOlm.
Survived cadets: ELECTRON, DarkSoldier, Leech, 845bytes, DarkDestroyer, Sprate,
sprayer, Platoon, Yoda, DirkGently, Jwpazz, Gip, Ajax, Battlemage, Worm, Shark,
FoxSierra, MaudDib, RauaRoobert, NumeroOtto, CounterStrike, Dragon, Shrike,
KillerBees, Bizarro, Earthworm.
Finally, removed cadets: Mine, Rabbit, Debil, Dafnetyc, Dafneltyc, Hamilton,
Halo, Shank, DarkDestroyerV04, KennyBlankenship, Joyal, GiovyVbis, GipStrict,
ImprobDrive, Jwpazz2, GipLarge, AlessioV3.

During the next challenge we'll see if a more clean cadets' arena will improve
the number of matches and the reliability of the results.

On the veterans' side, we had the usual fair and interesting tournament. This
time the arena counted only 12 jrobots and the competition was very hard. At the
end only 5 jrobots achieved the requested percentages to survive. They are:
HeadHunter (for its result in single category), Starkle, IonStorm (good third
place), StrangeMatter (good second place) and Tango (another triple victory).
The other were demoted and I hope they are going to do a little more selection
among cadets. They are: GulleFjun, Mouth, FlameBall, Bohan, Hellfire, DOOM,
Stately.

Now you can upload your jrobots to win the August 2006 JROBOTS Challenge (August
5th-26th 2006) aka Gimme Five!

--
Ciao, leo boselli@...
* APOCALYX 3D Engine: http://apocalyx.sf.net
* GUN-TACTYX Game: http://guntactyx.gameprog.it
* JROBOTS Java Battles: http://jrobots.sf.net