Sections
Hills
Infinite Hills
Tournaments
Software
Evolving
Optimizer
Community
Newsletter
Discussion
History
Sections
 
For Beginners
First Steps
FAQ
Guides
Lexicon
Benchmarks
For Beginners
> Home > Tournaments > Corewar Single Elimination Championship 2005> Qualification Round

The Qualification Round

Many thanks for the very good feedback. I didn't expected to see so many redcoder participating on this event.

I would highly appreciate to get beside your entry also a small comment. Just take a look on Macrae's intoxicating comments for his entries on the KOFACOTO tournament. I'll include all comments with the round results.

Once the groups have been determined today you could send me a sorted list of participants as a guess about the final ranking - and according to the average "grades" for each player, I determine a coeficient of probability for wining, according to the opinion of all participants. I think it would be interesting to see who is the favorite in a match.

All participant are randomly assigned into four groups. The qualification round will then run round robin in each group. The four best of each group are qualified for the second round. The ranking in the groups will be used to assign each player in the single elimination tournament bracket.

Deadline: Saturday, October 29, 2005, 9pm CET


The Groups

Group AGroup BGroup CGroup D
brxGerman LabargaRoy van RijnJohn K. Lewis
Michal JaneczekWill VarfarLAchiel kauka
X-manJens GutzeitinversedChip Wendell
Sascha ZapfdatagramSimon WainwrightMizcu
Nenad TomasevJohn MetcalffiveopZul Nadzri
S. FernandesSapan Bhatia


There was an intense discussion on IRC after the 'lottery' was finished. I breathed a sigh of relief that the groups are more or less equal in strength. I was worrying that they could be unbalanced. Here are some few comments from IRC:
<Roy> Group D might be the closest I think, interesting group
<brx> nenad, sascha, michal. this is going to be fun.
<sf> I'm glad I'm not in group C
<Nenad> I would like best to be in C
<brx> I would love to be in C


The Rules

The CSEC2005 starts gentle with a medium sized core. P-space and PIN are disallowed. The absence of p-warriors and an almost unexplored parameter setting should give every player an equal chance.

The parameter setting is:

Coresize3800
max. Cycles38000
max. Processes3800
max. Warrior Length38
min. Distance38
p-spacedisallowed
Rounds380


The commands for pmars is:

pmars -s 3800 -c 38000 -p 3800 -l 38 -d 38 -r 380


Useful Stuff

I've calculated some imp-steps using Planar's Euclid's extended. They are needed if you want to write any kind of imp-warrior.

Imp-PointsImp-Step
31267
7543
91689
11691

Also have a look on John Metcalfs website for programms to calculate optima constants. This could be useful for stones.

Happy programming and may the Core with you


Deadline

Saturday, October 29, 2005, 9pm CET


Player Predictions

I got already 10 predictions for a players ranking. Below are the results based on the average prediction for each player.
Roy van Rijn 2,2 John Metcalf 3,7 Michal Janeszek 4,7 Chip Wendell 4,8 Nenad Tomasev 5,1 Zul Nadzri 6,3 Sascha Zapf 6,8 German Labarga 7,8 Jens Gutzeit 9,2 Simon Wainwright 10,1 LAchi 10,6 Will Varfar 11,4 Mizcu 12,0 John K. Lewis 12,4 el kauka 12,5 inversed 13,3 S. Fernandes 14,9 brx 15,1 fiveop 17,1 datagram 18,9 X-man 19,7
The favourite for the CSEC2005 seems clearly Roy van Rijn. But as usual tournaments always have their own rules and one can always expect surprising results ;-)


The Results

Group A


Fizmo's Comment


Michal Janeczek, Nenad Tomasev and Sascha Zapf are the favourites in this group based on the predictions. S. Fernandez and brx are hoping to be the fourth player continuing the championship, while X-man is the unknown factor in this group.

The Entries


A quite interesting bandwidth of different strategies were send in this group, with a bishot (Nenad Tomasev), a blur-scanner with an unique zooom-trick (S. Fernandes), a stone/imp (brx), a HSA-ish Scanner (Michal Janeczek), a stone/paper (Sascha Zapf) and a basic dwarf using mul instead of an add (x-man).

The Comments


Nanad Tomasev:
I decided to use a bishot in this round, because I knew that 
oneshots and bishots will be better in coresize==3800 than in 
coresize==8000, so it seemed like a logical choice to use one 
of those. I know that it is risky... 

I wanted to use somethin a little safer, like an incendiary 
first, but I realized that it didn't score that well (the one 
that I've made). I'm afraid that I might get trashed by 
coreclears or maybe even CLP's here, but... I hope that noone 
is crazy enough to use a CLP... 

Why a bishot?

Well, it's a little safer than oneshot - able to bypass some 
decoys, etc. ... although it is weaker against imps.
So, it's a gamble. I hope that I'll manage to qualify, though. 
I know that it's not that original to submit a slightly reworked 
tiny and 94nop warrior like Diptera, but... I didn't have enough
time this week to test new stuff, and I also didn't want to use 
tricks from my unpublished warriors, so... I decided to place my
bet with Diptera. We'll see soon if I made a terrible mistake.

as for what other players will try to make... I've got no idea. 
I think that sf will try a oneshot. And Sascha Zapf might decide 
to use a s/i. ... but that's just a conjecture. I believe, 
however, that noone will use papers. That is too big a risk. 
I'm worried about the size of the groups. It renders the scores 
very unpredictable...

But oneshots ARE probably the best choice, generally speaking. 
In practise, you might use them and fail.
Sascha Zapf:
Hmm, qualification - 3800 the number of the beast...

No instructions are forbidden, everything can be used... 
Guess that there are all kind of warriors present. Don't want to 
bet on what Michal does, or Nenad...Can't see what brx or S-Fernandes 
choose to get on the top. So first the benchmark. 20 Warrior from Tiny.. 
Redesigned for 3800er core and optimized two times against themselves..

Then steel the code from one of Roy's most successfull warrior and 
optimize like hell.

Hope for the best
Michal Janeczek:
For this round I'm expecting a few stone/papers (since they became
quite fashionable recently), maybe also one-shots and stone/imps.
The warrior I'm submitting is Recon 2, with constants picked to match
core parameters,  and wasn't tested at all (except for looking at
pmars display to see if patterns look nice :)). Nevertheless, I hope
that with a bit of luck I will get out of the group.


The Table


#%W%L%T NameAuthorScore%
149.620.829.6 Spoorbrx178.40100.0
249.427.822.8 3800 - The number of the beast.Sascha Zapf170.96 95.8
351.438.4 10.1 GrendelS.Fernandes164.43 92.2
445.046.2 8.9 RatatoskMichal Janeczek143.74 80.6
541.241.217.6 Diptera3800Nenad Tomasev141.09 79.1
6 6.668.624.8 megamanX-man 44.52 25.0


The Analysis


What a surprise!!! brx's Spoor is the clear winner of this group followed by Sascha Zapf's 3800 - The number of the beast. Both were able to defeat the three scanner in the group.

The crux of the matter in the decision which of the three scanner will dropped from the CSEC2005 were the scores between them. S. Fernandes' zooom-ish Scanner Grendel shows some advantages clearly cement the third place.

Michal Janeczek's .8C SNE scanner Ratatosk and Nenad Tomasev's bishot Diptera3800 scores so close that I decided to let the group re-run at maximum rounds. Even then it was extremely close with slightly advantages for Michal. Very unlucky for Nenad.

X-man's mul-dwarf megaman didn't had any chance against his five opponents. Nevertheless, my congrats to the youngest tournament participant in the corewar history with an age of 8 years.


Group B


Fizmo's Comment


John Metcalf and German Labarga are the favourites in this group based on the predictions. Will Varfar and Jens Gutzeit have also good chances to successful pass the qualification, while Datagram give his best to surprise his opponents.

The Entries


As expected we see a lot of oneshots (German Labarga and John Metcalf) and papers (Jens Gutzeit). Datagram sent a stone with sd-clear.

The Comments


John Metcalf:
Something simple: decoy-maker + oneshot. I am worried if I try 
anything more complex it will backfire.
German Labarga:
Countdown started i didn't known what to write.

Brief analysis:
---------------

   Let's take a look at the group and see what my opponents
could submit.

                         Group B 
                         ======= 

                     German Labarga 
                      Will Varfar 
                      Jens Gutzeit 
                        datagram 
                      John Metcalf 
                      Sapan Bhatia 

   Will is mostly an evolver, though i assume he knows well the
basic strategies. I expect an evolved paper from him, or just
and optimized human designed paper.
That should not be a serious problem... should it?

   As far as i know, the only succesful warriors from Jens are papers.
I have seen Yatima 2.0.5 and Military Grade Milk 1.1.2, and they
are very agressive papers. That could be a problem. I don't expect
a s/i or s/p from him. Those strategies has a bit more complex design
and he submited his entry relatively soon.

   I saw at the IRC-log that Datagram could do a paper.  Don't know if
he will submit one finally. He could surprise us with a s/i.
He said that he had not much time so i don't expect anything complex 
from him.

   About John, well, he can do anything. Let's assume he will submit
a scanner to take advantage from Jens' papers. Nothing risky, zooomish
blur, or a oneshot. Though, it would not be a surprise for me to see
a s/p or s/i from him.

   I simply have no idea about what Sapan could submit...

And now what?
-------------
   I first tried a s/p, based on my Gihegruoerg, wich has a tough
anti-imp paper. The idea was to take points from any scanner from
John and stone/imps, and resist papers, but i noticed it has
problems against those papers from Jens. Also, s/i's are the best
feed for such a s/p, and i don't expect to see many.

   So? I think that facing a group of contenders, i might face-off
different strategies. The best way, is probably to assume some risk
and submit something agressive, that can take many points, doesn't
mater if it loses a lot too.

  I don't have time to write and test something complex.
I'm just going to take a oneshot, (i think "Boss is back again" is my
best one), and addapt it to this round settings.

   It has problems against papers, but can get nice win percentages
from them too. The bomb detection trick helps a lot against s/i's
and s/p's, and the core trashing helps a bit against other scissors.
This oneshot seems to be agressive enough to score well without
a too high risk.
Go ahead.

-G.labarga-
Jens Gutzeit:
The Making of "Evidence of Inability"
-------------------------------------

After the rules for the qualification round went out, I had to think 
about what kind of warrior to submit. One thing was clear from the 
start. It had to contain a quickscanner.

But what else? I could write a scanner in order to make something 
unexpected. I could write a stone/imp, too. I could write a paper to 
do something very execpted ...

I wrote a paper, because I'm lazy. What kind of paper? I'm lazy, so 
I took the first, that I could find. It was a little variation of 
nPaper II. It is still a very good one, but has its weak points. But 
because I only want to qualify for the next round and not to win (i.e. 
probably face Roy the next round), I did nothing more about the paper.

The quickscanner was simple. I've taken a normal q^4, deleted some 
scans, made sure everything works. Ready. I'M LAZY!

I didn't tweak a constant. I didn't make a benchmark to optimize it 
against. It has taken 5 minutes to find a good name ... yes, I know, 
that I'm doomed!

- Jens

PS: In the unlikely case, that I've made it to round 2, please assume, 
that I'm not lazy. It was all divine inspiration, but it is probably 
too late to say something like this now.


The Table


#%W%L%T NameAuthorScore%
156.135.1 8.9 GungnirJohn Metcalf177.09100.0
235.232.032.7 Evidence of InabilityJens Gutzeit138.49 78.2
341.248.110.6 Disfunctional dishwasherGerman Labarga134.39 75.9
429.747.123.2 Quamtadatagram112.28 63.4
5---no entryWill Varfar--


The Analysis


John Metcalf is the clear winner with his oneshot Gungnir followed by Jens Gutzeit's paper Evidence of Inability and German Labarga's oneshot Disfunctional dishwasher. Datagram completed the group in fourth position with his stone Quamta.


Group C


Fizmo's Comments

Roy van Rijn is the clear favourite in this group based on the predictions. LAchi, inversed and Simon Wainwright seems to have equal chances to end up under the top four, whereas nobody should understimate fiveop.

The Entries

Again oneshots (Roy van Rijn, Simon Wainwright) and coreclears (LAchi and fiveop) as the most predicted strategies were sent, while inversed has choosen a stone/imp.

The Comments

LAchi:
Well... after rewriting French Kiss for coresize 3800 I just discovered 
that the clear alone scored better... :/
inversed:
Mean stone/imp, not well optimized. But it should beat any scanner 
and I expect oneshots/bishots from LAchi and Wainwright (yes, this is naive). 
Relatively large, so adding qScan is useless. Anyway I think I have a chance.
Roy van Rijn:
 
First I created a benchmark (and deleted it, see logs 26-10-05)
   then I tried all warriors I could find
   seems that mostly oneshots and Hullabaloo do well

 So I optimized Myrmidon (my best oneshot) and Hullabaloo
   they scored almost the same against my benchmark..
   after a one-on-one fight the best warrior was the Hullabaloo-clone

 I suspect others will do the same (benchmarking)
   they will also notice oneshots are best and submit that
   and then having Hulla might be a slight advantage

Sending: Hullabaloo in 3800

Some more testing, but WAIT! Hullabaloo is only strong against a few oneshots
   not all oneshots in my benchmark... that is VERY dangerous in this round

Back to the drawing board.. time to think of another strategy
   looking at my enemies, fiveop might make a oneshot, so does LAchi
   he started out making bombing-oneshots and later some papers.
   inversed could make a oneshot, and simon is always tricky

How about a oneshot that beats most other oneshots
   I took my best oneshot up to date, Myrmidon
   Optimized it a little bit against oneshots in 3800 core
   And here is the replacement..just to tricky to send a stone/paper!

Sending: Myrmidon clone for 3800

(Never change strategies, stick with the first thing that comes to mind...)

So much doubt...
I'm going to test what was a better choice when this tournament is over!

Being the favourite isn't fun at all... I can't win, only fail


The Table


#%W%L%T NameAuthorScore%
157.028.514.5 3.8*10^3inversed185.52100.0
253.542.8 3.8 FenrirSimon Wainwright164.21 88.5
351.343.9 4.8 The Catcher in the RyeRoy van Rijn158.79 85.6
450.745.7 3.6 Ymir's Tendrilsfiveop155.65 83.9
521.272.9 5.9 Hug'n'clearLAchi 69.55 37.5


The Analysis


Well, this group makes me a bit extra work because of different compilations between CoreWin and pmars. Nevertheless here are the results with inversed's stone/imp 3.8*10^3 as the clear leader in this group, followed by Simon Wainwright's oneshot Fenrir and Roy van Rijn's oneshot The Catcher in the Rye. Fiveop's coreclear Ymir's Tendrils was getting fourth while LAchi's coreclear is in last spot showing weaknesses against most of his opponents.


Group D


Fizmo's Comments:

Group D seems the most thrilling one because beside Chip Wendell (who is the clear favourite) all others seems to have equal chances to pass to the next round.

The Entries:

Zul Nadzri and Mizcu sent a paper, while Chip Wendell stone/imps and el kauka a stone/paper/imp. Nobody send a scanner!?!

The Comments:

Zul Nadzri:
Here it is.... my gamble for this round:

Note: Take the facts, leave out the rumours
a) #1 -> During chatting at #corewar, I discover a secret: put 
   "John" anywhere to gain advantage.
b) #2 -> Due to the tough Group D, I will go for non-last ranking 
   instead of the leader spot.
c) 8000 vs 800... which one wins at 3800? Well, let's try the 
   smaller core.
d) Which tiny warrior to use? Get an off-the-hill warrior and 
   revitalize it! Quite interesting.
e) Which one? How about madpixel... sounds maniac to me. Fit 
   for this crazy core size.
f) Adjustments? Change some constants to cover more areas.
g) {Looks like SPL/SPL/DAT clear is the major problem. No time, 
   so, go ahead and risk on this.}
h) No pspace -- Therefore, no self-win strategy (Fizmo hates 
   handshakers...just a guess :)
i) How to optimize? Duh... I don't know how to use Optimax. No 
   server offering this command too.
j) So, trial and error.
k) No second warrior to help me up. Just one chance to survive...
***
y) Here it is.... Jump "John" Jump. See the mad pixels jumpin'.
z) p/s: if I lost this round, please search the guy starting 
with Je**, who tipped me wrongly ! "John"..yeah.


The Table


#%W%L%T NameAuthorScore%
1 5.0 1.493.6 Jump "John" JumpZul Nadzri108.66100.0
2 4.6 4.690.8 DoiMiz104.60 96.3
3 5.8 7.287.0 Mountain TrollChip Wendell104.49 96.2
4 1.0 3.395.7 quick hackel kauka 98.67 90.8
5---no entryJohn K. Lewis--


The Analysis


What a crazy group. No scanner was sent, so it was a pretty close group seeing Zul Nadzri's paper Jump "John" Jump (what a funny name ;-) winning followed by Mizcu's paper Dio, Chip Wendell's stone/imp Mountain Troll and el kauka's stone/paper/imp quick hack.



Off-Topics

I've run all entries round robin at max. Rounds. The Groups are assigned by different colors. As you can see the strongest group seems A.

Group AGroup BGroup CGroup D


#%W%L%TName AuthorScore%
149.836.813.4 GrendelS.Fernandes162.74100.0
250.237.812.0 FenrirSimon Wainwright162.72100.0
350.641.6 7.8 RatatoskMichal Janeczek159.52 98.0
448.938.312.8 GungnirJohn Metcalf159.38 97.9
548.738.712.6 The Catcher in the RyeRoy van Rijn158.70 97.5
647.136.616.3 Diptera3800Nenad Tomasev157.51 96.8
738.821.939.3 Spoorbrx155.71 95.7
846.740.113.1 Disfunctional dishwasherG.Labarga153.30 94.2
936.923.139.9 Mountain TrollChip Wendell150.76 92.6
1038.728.033.3 3800 - The number of the beast.Sascha zapf149.31 91.7
1134.323.442.3 Evidence of InabilityJens Gutzeit145.30 89.3
1229.633.237.2 DoiMiz126.07 77.5
1336.648.814.6 Ymir's Tendrilsfiveop124.49 76.5
1423.322.654.1 Jump "John" JumpZul Nadzri124.07 76.2
1532.145.322.6 3.8*10^3inversed119.03 73.1
1615.512.172.4 quick hackel kauka118.83 73.0
1729.345.525.2 Quamtadatagram113.12 69.5
1829.955.015.0 Hug'n'clearLAchi104.84 64.4
19 6.564.828.7 megamanX-man 48.17 29.6



The Second Chance Tournament Tree

There is a second chance for all players who wants to continue the CSEC2005:

The Second Chance Tournament Tree.

The winner will be surely the winner of the hearts of the corewar community. Please give me after each round a notice if you want to participate/continue this event. After the qualification round the following player can participate:

Nenad Tomasev
LAchi
John K. Lewis
X-man

In the case of an odd number, I'll fill the gap hopefully giving a good challenge in the spare time I have beside the tournament organization.

2002-2005 corewar.info. All Rights Reserved. Logo C. Schmidt
CSEC 2005
Home
Schedule
Qualification
Round 2
Quarterfinal
Semi Final
Final
CSEC 2005
 
Player Profiles
Jens Gutzeit
X-man
el kauka
fiveop
Nenad Tomašev
John Metcalf
LAchi
Zul Nadzri
S. Fernandes
Roy van Rijn
German Labarga
Michal Janeczek
brx
Mizcu
Sascha Zapf
Will Varfar
Chip Wendell
inversed
datagram
John K. Lewis
Simon Wainwright
Player Profiles