Sections
Home
Hills
Infinite Hills
Tournaments
Software
Evolving
Optimizer
Community
Newsletter
Discussion
History
Sections
 
For Beginners
First Steps
FAQ
Guides
Lexicon
Benchmarks
For Beginners
> Home > The Corewar Newsletters > Core Warrior > Issue #1

Issue 46                                                   October 7, 1996
______________________________________________________________________________
Core Warrior is a weekly newsletter promoting the game of corewar.  Emphasis
is placed on the most active hills--currently the '94 draft hill and the
beginner hill.  Coverage will follow where ever the action is.  If you have
no clue what I'm talking about then check out these five-star internet locals
for more information:

FAQs are available by anonymous FTP from rtfm.mit.edu as
pub/usenet/news.answers/games/corewar-faq.Z
FTP site is: ftp.csua.berkeley.edu /pub/corewar
Web pages are at: (Please note new Stormking's address)
http://www.koth.org/                            ;Stormking
http://www.ecst.csuchico.edu/~pizza/koth        ;Pizza
http://pauillac.inria.fr/~doligez/corewar/      ;Planar

Newbies should check the stormking page for the FAQ, language specification,
guides, and tutorials.  Post questions to rec.games.corewar.  All new players
are infinitely welcome!

If ftp.csua.berkeley.edu is unreachable, you can download pMARS at:
Terry's web page--http://www.infi.net/~wtnewton/corewar/
Planar ftp site--ftp://ftp.inria.fr/INRIA/Projects/para/doligez/cw/pmars
Fechter ftp site--ftp://members.aol.com/ofechner/corewar

A collection of Bezzi's hints in the first issues is available at:
ftp://ftp.volftp.vol.it/pub/pc/msdos/games/solutions/bbhints.zip
______________________________________________________________________________
Greetings.

There was a lot of activity on the 94 hill this week mainly due to testing -
the claim is that 100 rounds isn't enough to measure the performance of a
warrior - there will be greater fluctuations in test scores. Perhaps testing
should go for 200 rounds.

The Core War Standards committee has not been discussed much. Is it needed
right now? Maybe we just need some players to volunteer to complete the
'94 draft documentation and other details. To get more people to play the
game someone could arrange an archive for DOS (a popular platform) which
contains PMARS, documentation, tutorials, Internet information, etc. and then
distribute the archive. If you do plan to do either of these things, let
everyone else know beforehand. Just some thoughts :-)

A reminder: Beppe has organised a tournament. Those who are participating
should have received information from him.

--Anton Marsden
______________________________________________________________________________
Current Status of the Internet Pizza Server ICWS '94 Draft Hill:

Hill Specs:
   coresize: 8000
   max. processes: 8000
   duration: after 80,000 cycles, a tie is declared.
   max. entry length: 100
   minimum distance: 100
   rounds fought: 200
   instruction set: ICWS '94 Draft

The current ICWS '94 Draft hill:
 #   %W /  %L /  %T                       Name               Author Score  Age
 1  47.0/ 36.7/ 16.3                    Blur 2        Anton Marsden 157.3  183
 2  31.3/  9.6/ 59.1      Return Of The Jedimp             John K W 152.9    6
 3  43.3/ 34.3/ 22.4       Damage Incorporated        Anton Marsden 152.4   47
 4  35.3/ 20.8/ 43.9                    Gigolo   Core Warrior staff 149.8   60
 5  34.4/ 20.5/ 45.1           unrequited love                kafka 148.3  104
 6  30.7/ 14.3/ 54.9               Impish v0.2          Ian Oversby 147.1   45
 7  45.8/ 44.7/  9.5                  Q^2 Miro         Anders Ivner 146.9  234
 8  40.4/ 36.5/ 23.1          Instant Wolf 3.4                Edgar 144.3   67
 9  41.0/ 38.0/ 21.0                     Probe        Anton Marsden 144.1  184
10  40.8/ 39.5/ 19.7                 Goldfinch              P.Kline 142.1  125
11  29.1/ 18.3/ 52.6                   Rosebud                Beppe 140.0  857
12  38.3/ 38.8/ 22.9                  mrb-test           m r bremer 137.9   90
13  32.5/ 27.4/ 40.1               Falcon v0.3          Ian Oversby 137.6   89
14  34.2/ 30.8/ 35.0              CC Paper 3.3                Franz 137.5    2
15  35.5/ 34.2/ 30.3                     Frogz                Franz 136.9   64
16  36.0/ 35.2/ 28.8                T.N.T. pro    Maurizio Vittuari 136.9  902
17  31.8/ 28.5/ 39.7              Simple v0.4b          Ian Oversby 135.1  135
18  23.5/ 12.0/ 64.5                    ompega       Steven Morrell 135.0  299
19  36.2/ 38.7/ 25.0               Inferno 2.4       Philip Kendall 133.8   62
20  38.0/ 43.7/ 18.2              Taking Names              P.Kline 132.3   49
21  29.3/ 29.3/ 41.5        Jack in the box II          Beppe Bezzi 129.2  541
22  33.5/ 39.3/ 27.2                    mayhem             harleyQ2 127.7    3
23  24.9/ 35.4/ 39.7             Barbarian 3.1                Franz 114.4    5
24   7.3/ 20.0/ 72.6                       sin             harleyQ2  94.6    1
25   4.0/  0.0/  0.0                      Test        Anton Marsden  12.0   94

Weekly age: 46 ( 13 last issue, 15 the week before )
New warriors: 6 (2 of which won't last long)  Turnover/age rate 13%
Average age: 169 ( 140 last issue, 146 the week before )
Average score: 133 ( 140 last issue, 142 the week before )
The top 25 warriors are represented by 15 authors: Marsden with 4; Franz and
Oversby with 3; Beppe, Kline and harleyQ2 with 2; everyone else with 1.

King Report: Gigolo was on top for the majority of the challenges but Return
of the Jedimp and Blur 2 managed to get reach 1st place for several
challenges.

Q^2 scans are still doing very well and will probably continue to do so.
They turn draws into wins and this has a big advantage when combined with
paper/imp types.

There was a lot of testing this week, mainly by JKW and Franz.
______________________________________________________________________________
94 - What's New

 #   %W /  %L /  %T                       Name               Author Score  Age
 5  28.3/ 15.9/ 55.9               Impish v0.2          Ian Oversby 140.6    1
 1  27.7/ 10.4/ 61.9      Return Of The Jedimp             John K W 145.0    1
24  22.3/ 36.5/ 41.2             Barbarian 3.1                Franz 108.1    1
19  34.0/ 40.1/ 25.9                    mayhem             harleyQ2 127.9    1
14  31.8/ 33.0/ 35.2              CC Paper 3.3                Franz 130.5    1
25   3.3/ 21.1/ 75.6                       sin             harleyQ2  85.6    1

Return of the Jedimp enters the hill in 1st place and Impish v0.2 also does
well to make 5th.
______________________________________________________________________________
94 - What's No More

 #   %W /  %L /  %T                       Name               Author Score  Age
26  31.4/ 40.6/ 28.0                 Yogi Bear              P.Kline 122.3  341
26  35.8/ 47.7/ 16.4           Earthquake v0.2         Bjoern & Ian 123.9   95
26  33.6/ 45.2/ 21.2                 myVamp5.5             Paulsson 122.1   59
26   1.2/ 35.3/ 63.6               IMP Factory                Franz  67.1    2

The most significant loss this week was Yogi Bear (a pspacer).
______________________________________________________________________________
94 - What's Old

 #   %W /  %L /  %T                       Name               Author Score  Age
16  36.0/ 35.2/ 28.8                T.N.T. pro    Maurizio Vittuari 136.9  902
11  29.1/ 18.3/ 52.6                   Rosebud                Beppe 140.0  857
21  29.3/ 29.3/ 41.5        Jack in the box II          Beppe Bezzi 129.2  541

No new entries here, but Yogi Bear is now gone. None of the older warriors are
in the top ten, indicating that the newer warriors are of a higher quality.
______________________________________________________________________________
HALL OF FAME
* means the warrior is still active.

Pos    Name                  Author          Age     Strategy
 1  Thermite II            Robert Macrae     2262    Qscan -> bomber
 2  Impfinity v4g1         Planar            1993    Stone/ imp
 3  Jack in the box        Beppe Bezzi       1620    P-warrior
 4  Tornado 3.0            Beppe Bezzi       1567    Bomber
 5  Torch t18              P.Kline           1539    Bomber
 6  Chameleon              Myer R Bremer     1437    P-warrior
 7  Frontwards v2          Steven Morrell    1420    One shot scanner
 8  Evol Cap 6.6           John Wilkinson    1299    Imp / stone
 9  quiz                   Schitzo           1262    Scanner/ bomber
10  T.N.T.                 Maurizio Vittuari 1204    Bomber
11  Grilled Octopus v0.5   David Boeren      1154    P-warrior
12  Hazy Shade II          John Wilkinson    1102    P-warrior
13  Stepping Stone         Kurt Franke       1049    Qscan -> Vampire
14  Iron Gate 1.5          Wayne Sheppard     926    CMP scanner
15  Agony II               Stefan Strack      912    CMP scanner
16  T.N.T. pro             Maurizio Vittuari  902 *  Bomber
17  Barrage                Anton Marsden      876    Qscan -> replicator
18  Blue Funk              Steven Morrell     869    Stone/ imp
19  Rosebud                Beppe Bezzi        857 *  Stone/ imp
20  Flurry                 Anton Marsden      835    Qscan -> pwarrior
21  Thermite 1.0           Robert Macrae      802    Qscan -> bomber
22  Blue Funk 3            Steven Morrell     766    Stone/ imp
23  Night Train            Karl Lewin         755    Replicator
24  Mirage 1.5             Anton Marsden      736    Scanner/ bomber
25  Blizzard               Anton Marsden      713    Qscan -> replicator


Both T.N.T. pro and Rosebud move into better positions.
______________________________________________________________________________
Current Status of the Internet Pizza Server Beginner's Hill:

Hill Specs:
   coresize: 8000
  max. processes: 8000
   duration: after 80,000 cycles, a tie is declared.
max. entry length: 100
 minimum distance: 100
    maximum age: At age 100, warriors are retired.
   rounds fought: 200
  instruction set: ICWS '94 Draft

The current Beginner hill:
 #   %W /  %L /  %T                       Name               Author Score  Age
 1  49.6/ 39.4/ 11.1                 Throw Rug             J E Long 159.8   78
 2  43.2/ 38.8/ 18.0          Velveeta Shift-I                 shar 147.6   54
 3  41.9/ 36.2/ 22.0                    mayhem             harleyQ2 147.5    4
 4  43.5/ 40.5/ 16.0          nSplit Lite v1.2                 shar 146.6   42
 5  43.9/ 41.4/ 14.6                  Hawk 1.0       Philip Kendall 146.4   75
 6  40.1/ 34.2/ 25.7           Versatility 1.6                 Ross 146.1   90
 7  33.7/ 21.3/ 45.0                 Gravel 11           Justin Kao 146.0   15
 8  42.9/ 39.9/ 17.2             Flamberge 9.0        Matt Lewinski 145.8   14
 9  42.5/ 40.0/ 17.5       Violent Micro v0.4d             basehead 145.1   92
10  39.0/ 33.1/ 27.8               Inferno 2.3       Philip Kendall 145.0   79
11  33.0/ 27.9/ 39.1              FastScan 1.0       Philip Kendall 138.2   34
12  36.5/ 35.0/ 28.5                Black v0.1           Justin Kao 137.9   93
13  31.2/ 24.7/ 44.0                  Sapphire       Philip Kendall 137.7   13
14  35.7/ 37.1/ 27.2                  murder.p             harleyQ2 134.3   36
15  29.9/ 27.4/ 42.8                Papyrus 13           Justin Kao 132.4   16
16  37.2/ 42.5/ 20.2         Microsoft NT v2.5           Justin Kao 131.9    5
17  37.3/ 42.7/ 20.1                  Eagle.b1       Philip Kendall 131.9   19
18  35.1/ 38.4/ 26.5         Microsoft NT test           Justin Kao 131.8    1
19  37.8/ 45.7/ 16.5   Neverending Brutality 7                Franz 129.9   69
20  38.0/ 46.5/ 15.5               Conquest II           Zul Nadzri 129.6   35
21  35.3/ 41.5/ 23.1             Vampirism 1.4       Philip Kendall 129.2   96
22  30.3/ 32.6/ 37.1                      Frog                Franz 128.0   56
23  32.3/ 37.5/ 30.2                       RSB             harleyQ2 127.2   63
24  23.7/ 22.4/ 54.0           Sand Storm v0.1           Justin Kao 125.0   10
25  20.5/ 37.5/ 42.0            Hammermill 4.0        Matt Lewinski 103.5    2

The top score has fallen substantially from what it was last week and the
warrior scores are more evenly spread indicating that the "beginners" are
getting better - it might be time for some of you to start submitting warriors
to the 94x hill - there are some old but good warriors on there.
______________________________________________________________________________
The Hint
Mutation - improving your end game
by Anton Marsden

There has been little development in the basic warrior types recently,
perhaps because further optimisation cannot be obtained or is difficult to
obtain. In this hint I will discuss mutation of warriors - a potentially
powerful method of optimisation.

The most common use of mutation is to bomb your code with a SPL instruction to
start a core clear. This technique is usually used in incendiary bombers (see
CW #15 for more information on this). The whole idea behind mutation is to
reduce the code size (or the number of sequential instructions).

To illustrate the possibilities ...

This is one of my old warriors (Mirage 1.5) with a "standard" DAT clear. It
is 11 lines long.

ORG s

step EQU 3039
gate EQU data
btm EQU (clr+1)

data: spl.a  #0,>step
pos:  add.ab #step,#2*step
kill: mov.i  bomb,>data
s:    jmz.f  @0,@pos
c:    slt.b  pos,#btm-data+2
      mov.b  pos,data
      jmn.b  @0,pos

bomb: spl.a  #0,#0
      mov.i  clr,>gate
      djn.f  -1,{gate
clr:  dat.f  0,btm-gate+1


I wanted to add a spiral clear to this (see CW #17), but adding the clear
made the warrior too big to get anywhere on the hill. Solution: mutation.
Amazingly, I could mutate the scan engine into a forward DAT clear and spiral
scan. Here is how it turned out (you really have to watch it to see what it
does):

ORG s

step EQU 98 ; mod-2 - give the spiral clear enough time to go through the core
btm EQU (bomb+1)

data: spl.a  #0,step
pos:  add.ab #step,#2*step
kill: mov.i  bomb,>data
s:    jmz.f  @0,@pos
c:    slt.b  pos,#btm-data+2
chg:  mov.f  @s,@kill
      jmn.b  pos,*0
      mov.f  next,<chg ; pos: add.ab #381,#btm-data
                       ; chg: mov.f @s,@pos
      djn.f  >-2,{chg  ; Stop jmn from falling through again
                       ; chg: mov.f @kill,@pos -> mov.f next,@pos
                       ; kill: mov.i next,>data-1

next: dat.f  381,btm-data
bomb: spl.a  #0,#0

This warrior is also 11 lines long, a great improvement! However, it took me
hours to get it working properly. Notice the use of indirection and decrements
to mutate the code (the chg line is the key to the whole process). The entire
warrior (with name, boot, decoy and a few other changes) is at the end of this
issue. It still performs poorly against fast scanners/bombers but is much
better against imp types. There is a tradeoff between robustness and
functionality - the SPL in Mirage stopped it from terminating if the code
below the SPL was bombed whereas the new warrior's instructions must be
completely intact to avoid termination.

Mutations like this are very hard to accomplish so I will now suggest a
general approach to designing these warriors.

Take the basic structure of the first stage of your warrior and try and
base the second stage around it, ie. structure it so that the second stage
can be obtained with little change to the instruction sequence.

You need to trigger your mutation somehow. The best approach may be to fall
through a DJN/JMN check. If you are dealing with a self-splitting bomber DJN
is one to use. Also, self-splitting has the advantage that you don't have to
jump back into the main loop after a mutation.

The next step is to make the mutation happen properly. Use as many lines of
code as is necessary. Also remember to modify the code so that the mutation
is not retriggered later.

The final step is the most creative. You should aim to reduce the number of
lines required for the mutation. This can be done by:

- Using indirection and decrements (DJN is very useful).
- Swapping instructions.
- Thinking very hard.

You will need a lot of patience. Sometimes you may find that optimisation is
just not possible. Then it's time to take another look at the structure or the
trigger... or just give up :-)

I hope this information will help someone out there...

Have fun!
______________________________________________________________________________
Extra Extra
Eggbeater
by Anton Marsden

The warrior boots and immediately starts overwriting it's own boot code,
starting with the destination pointer.

;redcode-94
;name Eggbeater
;author Anton Marsden
;strategy Mirage 1.5 with spiral clear
;assert CORESIZE==8000

N FOR 7
    spl.i #0,#0
    spl.a #N+(1*8),#N+(1*8)
    spl.f #0,#0
    spl.x #N+(2*8),#N+(2*8)
    spl.b #0,#0
    spl.ab #N+(3*8),#N+(3*8)
    spl.ba #0,#0
    spl.b #N+(4*8),#N+(4*8)
    spl.ab #0,#0
    spl.f #N+(5*8),#N+(5*8)
ROF

FOR 3
    spl.x #0,#0
    spl.a #1,#1
ROF

;-----------------

destdiv2 EQU 2000 ; not really this
dest EQU (2*destdiv2+1) ; odd number :-)

boot: mov.i  bomb,dest
FOR 9
      mov.i  {boot,<boot
ROF
      spl @boot,}2001
      mov.i  {boot,<boot
      dat.f  $0,$0

;-----------------

step EQU 98 ; mod-2
btm EQU (bomb+1)

data: spl.a  #0,>-dest+bomb
pos:  add.ab #step,#0
kill: mov.i  bomb,>data
s:    jmz.f  @0,@pos
c:    slt.b  pos,#btm-data+2
chg:  mov.i  @s,@kill
      jmn.b  pos,*0
      mov.f  #381,<chg
      djn.f  <-2,{chg

next: dat.f  0,btm-data
bomb: spl.a  #0,#0

END boot
______________________________________________________________________________
Questions?  Concerns?  Comments?  Complaints?  Mail them to people who care.
authors: Beppe Bezzi <bezzi@nemo.it> or Myer Bremer <bremermr@ecn.purdue.edu>
or Anton Marsden <amarsden@comp.vuw.ac.nz>
2002-2005 corewar.info. Logo C. Schmidt