Issue 20 March 11, 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:
http://www.stormking.com/~koth ;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 still out of service, you can download Pmars at
Terry's web page--http://www.infi.net/~wtnewton/corewar/
______________________________________________________________________________
Greetings.
Sorry for the delay--what's new with my irresponsability? Give me a break,
I'm just a kid! Kurt Franke is kind enough to provide us with the hint this
week. Thanks goes out to everyone who has contributed to Core Warrior.
--M R Bremer
______________________________________________________________________________
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
# %W/ %L/ %T Name Author Score Age
1 45/ 32/ 23 Grilled Octopus v0.5 David Boeren 158 19
2 44/ 35/ 21 Stepping Stone Kurt Franke 152 144
3 39/ 31/ 30 Tornado 3.0 Beppe 148 604
4 43/ 40/ 16 quiz Schitzo 146 1202
5 40/ 34/ 26 Chameleon M R Bremer 146 274
6 40/ 34/ 26 Thermite II Robert Macrae 145 803
7 36/ 28/ 36 Barrage Anton Marsden 144 276
8 42/ 41/ 17 Frontwards v2 Steven Morrell 143 1055
9 42/ 42/ 16 C I A Anders Ivner 143 62
10 44/ 45/ 11 Memories Beppe Bezzi 142 196
11 29/ 16/ 55 Hazy Shade Of Winter John K W 142 431
12 43/ 44/ 14 Hipshot V2 J E Long 142 14
13 37/ 32/ 31 T.N.T. Maurizio Vittuari 142 528
14 43/ 44/ 14 Test J E Long 142 2
15 31/ 21/ 48 Impfinity v4g1 Planar 141 1002
16 36/ 32/ 32 Torch t18 P.Kline 140 40
17 39/ 38/ 23 myConfuser Paulsson 139 16
18 29/ 21/ 49 Blizzard Anton Marsden 137 522
19 29/ 22/ 49 Gem of the Ocean P.Kline 136 1
20 25/ 14/ 61 Evol Cap 6.6 John Wilkinson 135 817
21 41/ 47/ 12 Wild Card M R Bremer 135 356
22 34/ 33/ 34 Jack in the box Beppe Bezzi 134 1569
23 40/ 45/ 16 Mirage 1.5 Anton Marsden 134 716
24 22/ 11/ 67 Q.S.P. John K W 132 117
25 4/ 0/ 0 Lithium And Water John K Wilkinson 12 12
Weekly age: 114
New warriors 8 Turnover/age rate 7%
Average age: 431 ( 414 last week, 333 the week before )
Average score: 142 ( 135 last week, 139 the week before )
The top 25 warriors are represented by 15 authors.
These rankings are more current than the What's New/What's No More listings.
Sorry for the confusion but I had to kill my 200+ pizza messages rather than
sift through them.
Paul Kline shattered the monotony when he _reset_ Torch t18! Jack in the
Box now is the all time age leader at 1569. Impfinity and Frontwards both
have cleared the 1000 mark. Now FOUR programs are over 1K and two more are
alrady over 800.
Boeren has refused to relinquish his strangle hold on the top spot. Once
Grilled Octopus started dropping, he patched some new code together and
voila: Grilled Octopus v0.5. Stepping Stone has been as low as 10th place
but usually resides in the top 5. His secret handshake used in this warrior
is revealed in the issue's hint section. Age leader Jack in the box has been
residing in the bottom of the pack all week and may soon fall to p-warriors
and brainwashing programs.
______________________________________________________________________________
94 - What's New
# %W/ %L/ %T Name Author Score Age
5 42/ 40/ 18 C I A Anders Ivner 143 1
25 6/ 51/ 43 Happy Paper Christopher Gorski 61 1
20 30/ 33/ 37 Torch t18 P.Kline 126 1
1 42/ 35/ 23 Grilled Octopus v0.5 David Boeren 149 1
25 16/ 16/ 68 Obvious Thief John K. Lewis 115 1
9 38/ 40/ 22 myConfuser Paulsson 137 1
14 40/ 44/ 16 Hipshot V2 J E Long 135 1
19 29/ 22/ 49 Gem of the Ocean P.Kline 136 1
______________________________________________________________________________
94 - What's No More.
# %W/ %L/ %T Name Author Score Age
26 1/ 3/ 0 test cia Anders Ivner 2 3
26 1/ 2/ 0 Clisson P.Kline 4 18
26 2/ 1/ 1 C I A Anders Ivner 7 47
26 29/ 36/ 35 Breeder V3.1 J E Long 123 60
26 3/ 0/ 0 Hipshot J E Long 10 94
26 30/ 39/ 31 Torch t18 P.Kline 121 1539
26 35/ 47/ 17 S.E.T.I. II John K W 124 163
26 2/ 2/ 0 Grilled Octopus v0.3 David Boeren 6 130
______________________________________________________________________________
94 - What's Old
# %W/ %L/ %T Name Author Score Age
22 34/ 33/ 34 Jack in the box Beppe Bezzi 134 1569
4 43/ 40/ 16 quiz Schitzo 146 1202
8 42/ 41/ 17 Frontwards v2 Steven Morrell 143 1055
15 31/ 21/ 48 Impfinity v4g1 Planar 141 1002
20 25/ 14/ 61 Evol Cap 6.6 John Wilkinson 135 817
6 40/ 34/ 26 Thermite II Robert Macrae 145 803
23 40/ 45/ 16 Mirage 1.5 Anton Marsden 134 716
3 39/ 31/ 30 Tornado 3.0 Beppe 148 604
13 37/ 32/ 31 T.N.T. Maurizio Vittuari 142 528
18 29/ 21/ 49 Blizzard Anton Marsden 137 522
11 29/ 16/ 55 Hazy Shade Of Winter John K W 142 431
21 41/ 47/ 12 Wild Card M R Bremer 135 356
7 36/ 28/ 36 Barrage Anton Marsden 144 276
5 40/ 34/ 26 Chameleon M R Bremer 146 274
Ten of these 14 warriors are already in the Hall!
______________________________________________________________________________
HALL OF FAME
* means the warrior is still running.
Pos Name Author Age Strategy
1 Jack in the box Beppe Bezzi 1569 * P-warrior
2 Torch t18 P.Kline 1539 Bomber
3 quiz Schitzo 1202 * Scanner/ bomber
4 Frontwards v2 Steven Morrell 1055 * One shot scanner
5 Impfinity v4g1 Planar 1002 * Stone/ imp
6 Iron Gate 1.5 Wayne Sheppard 926 CMP scanner
7 Agony II Stefan Strack 912 CMP scanner
8 Blue Funk Steven Morrell 869 Stone/ imp
9 Evol Cap 6.6 John Wilkinson 817 * Replicator
10 Thermite II Robert Macrae 803 * Bomber
11 Thermite 1.0 Robert Macrae 802 Qscan -> bomber
12 Blue Funk 3 Steven Morrell 766 Stone/ imp
13 Night Train Karl Lewin 755 Replicator
14 Mirage 1.5 Anton Marsden 716 * Scanner/ bomber
15 HeremScimitar A.Ivner,P.Kline 666 Bomber
16 La Bomba Beppe Bezzi 650 Qscan -> replicator
17 myVamp v3.7 Paulsson 643 Vampire
18 Armory - A5 Wilkinson 609 P-warrior
19 juliet and paper Bremer & Bezzi 607 P-warrior
20 Tornado 3.0 Beppe Bezzi 604 * Bomber
21 Phq Maurizio Vittuari 589 Qscan -> replicator
22 Hector 2 Kurt Franke 561 Qscan -> replicator
23 T.N.T. Maurizio Vittuari 528 * Bomber
24 Blizzard Anton Marsden 522 * Qscan -> replicator
25 Boombastic Maurizio Vittuari 521 Bomber
Marsden gets yet another active warrior on the Hall of Fame listings.
Blizzard replaces B-Panama X--a stone/replicator--by Steven Morrell.
______________________________________________________________________________
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
Last challenge: Mon Mar 4 21:42:59 PST 1996
# %W/ %L/ %T Name Author Score Age
1 49/ 31/ 20 Test Ian Oversby 168 29
2 48/ 31/ 21 Auntie v1.0 Ian Oversby 165 60
3 40/ 16/ 44 Breeder V3 J E Long 164 75
4 45/ 27/ 28 Evil Elf V.8 Andy Nevermind 163 64
5 40/ 18/ 42 NNAA John K. Lewis 162 25
6 39/ 19/ 42 Pretentious v0.2 Ian Oversby 160 30
7 30/ 11/ 59 silkworm v3.0 Brian Haskin 148 76
8 41/ 34/ 25 Judge Nutmeg Scott Manley 148 79
9 39/ 35/ 27 Paralyze v0.1 Ian Oversby 142 59
10 29/ 16/ 55 AB Scanner 2.9.4 Chris Hodson 142 90
11 40/ 40/ 19 Maniac v0.3 Ian Oversby 141 53
12 30/ 24/ 47 Hyakutake C/1996 B2 + Scott Manley 136 36
13 22/ 11/ 67 Well Dressed Imp Robert J. Street 134 5
14 33/ 37/ 30 Forked Lite Ning 4.074 Ansel Greenwood Serm 130 83
15 31/ 34/ 35 Pensive v0.3 Ian Oversby 128 2
16 22/ 18/ 59 RingWorm_v2.7 Calvin Loh 126 66
17 23/ 22/ 55 XXXOOO v2.0 Joshua Houk 125 41
18 29/ 33/ 37 Hyakutake Rising Scott Manley 125 78
19 30/ 39/ 31 Alien Kiss V1.0 bjoern guenzel 121 27
20 22/ 25/ 53 Stephanie's Pajamas v1.2 Robert J. Street 120 14
21 31/ 47/ 22 Optma Dwarf v4.0 Phoenix 114 24
22 22/ 30/ 48 Stephanie's Pajamas v1.3 Robert J. Street 114 1
23 29/ 51/ 20 GPT 09d David van Dam 108 13
24 10/ 12/ 79 ShadowImp Robert J. Street 107 15
25 23/ 42/ 36 Pommes-Ketchup V1.33 S. Schroeder 103 23
Ian Oversby has staked out quite a claim on the beginners' hill with six
warriors including the top two rankings. Ian has decided to share one of
his warriors with us.
;redcode-b
;name Auntie v1.0
;author Ian Oversby
;strategy v0.1 Searching for a good scanner.
;strategy v1.0 Shorter.
;assert 1
< boot deleted >
loop ADD.F cons, scan
scan CMP.I -1, 11
SLT.A #30, scan
fin DJN.B loop, <-100
MOV.I b1, }scan
count DJN.B -1, #13
SUB.A #13, scan
MOV.AB #13, count
JMZ.F loop, loop-5
cptr SPL.B #b1, #b2+1
b1 SPL.B #4, #b2+2
MOV.I *cptr, >cptr
JMP.B -1, <-10
b2 DAT.F #4, #5
cons DAT.F #3359, #3359
A quick look indicates a carpet bombing scanner strategy followed by a multi-
pass coreclear. While this strategy is successful on the b-hill, it will not
even come close to making the '94 draft hill. Timescape style replicators
used by most authors leave lots of decoys in core. A carpet bomber spends
too much time attacking them. A better strategy would be to attack with
spl/spl/jmp bombs like Memories. Also consider using a .8c scan engine
instead of the standard .66c engine.
top add.f step, comp
comp sne.i X, X+SPREAD
add.f step, comp
seq.i @comp, *comp
slt
djn.f top, <STREAM
A spiral clear may be needed to kill DieHard style imp/papers. Check out
Memories, Harmony, and SETI for examples. Then agressively pspace the
scanner with a decent bomber (or replicator) and you should gain a foothold
on the '94 hill. Good luck.
______________________________________________________________________________
The Hint by KURT FRANKE
How to Handshake
Whether or not it is right, when you challenge the hill, part of your
score comes from how well your warrior does against itself. If we are
paper and tie every time, that is about 4 points. If we are something
very aggressive and (nearly) split wins, that is about 6 points. And if
we use a handshake (or we are very, very lucky) we get this:
Program "Stepping Stone" (length 95) by "Kurt Franke"
(contact address "kurt@protector.lovett.rice.edu"):
has challenged the ICWS '94 Draft hill.
...
Program "Stepping Stone" (length 95) by "Kurt Franke"
(contact address "kurt@protector.lovett.rice.edu"):
;strategy Vampire
Stepping Stone wins: 200
Ties: 0
That is 12 points, 6 more than if I had let Stepping Stone split
wins. Sound good? Want to know how? Read on.
When challenging the hill, we fight the other programs two hundred times
each. Someone makes the first move in each round. The way this works is
the challenger -- that is us -- goes first on the first round, and then it
alternates (the defending warrior goes first on the second round, etc).
When we fight ourself, the version that goes first on the first round is
the version we get points for. So we want our warrior to realize when it
is fighting itself and have the right version win every time.
To recognize when we are fighting ourself, the obvious thing to do is
give our warrior a PIN and use pspace. For an introduction to pspace, see
Core Warrior #5. PIN's were not covered in that issue, but they are
simple. The acronym stands for P-space Identification Number. If we put
this line in our warrior,
PIN 123
Then we will share the same pspace as any other warrior that also has a
PIN of 123. Note that this is just a pseudo-opcode; it doesn't count as
one of our 100 lines and you cannot find it in core. It is just a message
to the simulator.
Now here is the handshake. place1 and place2 are just numbers. One
version of our code needs to execute:
stp #0, #place2 ;; clear the expected flag
stp #1, #place1 ;; set the flag for the other version
ldp #place2, #0 ;; check if the other version set our flag
While the other version simultaneously executes this:
stp #0, #place1 ;; (notice place1 and place2 are exchanged)
stp #1, #place2
ldp #place1, #0
If the load operations retrieve a 1, we know we are fighting ourself.
The only trick is that, when fighting ourself, the two versions are
(almost) identical so how can they be executing different code? Well,
there are two things different between the two versions. The first thing
is pspace location 0. p[0] is different from the other pspace locations.
It returns the result of the last round, and it is *not* shared when you
use a PIN. The second difference is that one version is one half-step
ahead of the other version.
It isn't obvious what can be done with the second difference, but using
the first difference,
last ldp #0, #0 ;; fetch the result of the last round
jmz lost, $last
stp #0, #place1
stp #1, #place2
ldp #place1, #0
...
lost stp #0, #place2
stp #1, #place1
ldp #palce2, #0
...
Now if we lost the last round, we execute the second part, and if we won
or tied the last round, we execute the first part. That is the basics of
a handshake. The rest is just code to figure out who should win and who
should lose the next round once we know we are fighting ourself.
Robert Macrae, who first explained the concept of a handshake to me,
solved the problem in a very complicated (but interesting!) way. Look at
his warrior Mason for details. Mason keeps track of how many rounds have
passed. If it is the first round or the last round was a tie, then there
is no difference in p[0] between the two versions and he just runs his
normal warrior. Once he gets a win or loss, he tries the handshake and
flags whether it worked or not. If the handshake fails (the ldp
instruction fetches 0), then Mason flags that he is fighting an enemy.
If it worked, then both versions commit suicide.
Commit suicide? Yes. He does that to find out which one moved first
on that round. (The one that moved first on that round will die first.)
Knowing the result of the suicide round and the round number that the
suicide occurred on, the warrior can calculate which version should win
the rest of the rounds.
Mason's handshake is lengthy (44 lines). It also has a nasty bug, which
was Mason's downfall. The bug is that it only does the handshake once and
then relies on values set in pspace after that. So, by brainwash or by
some paper making him execute the wrong lines in his code, he can be
fooled into thinking he is fighting himself and lose all the rounds:
Program "Hector 2" (length 96) by "Kurt Franke"
(contact address "kfranke@rice.edu"):
;strategy qscan -> replicator
Mason 2.0 wins: 0
Hector 2 wins: 187
Ties: 13
(Hector enjoyed that score immensely.)
Now let me describe my current handshake (the one in Stepping Stone).
My goals were to make sure I can't lose every time, and to reduce the
size.
The first goal is easily accomplished. Stepping Stone will run the
handshake code (that communicates through shared pspace) every round
unless he knows he's fighting an enemy. There is no way the handshake can
be fooled, unless my enemy has my PIN and the numbers place1 and place2
used for the handshake. The second goal is accomplished through giving up
the first round against everyone but myself. Let me explain;
automatically losing the first round is not as crazy as it sounds.
(You hear maniacal laughter in the distance.)
Suppose that on the very first round we have the challenger version (the
one we get points for) win. Then on subsequent rounds, all that must be
done is this: Run the handshake; if it works, then we are fighting
ourself and should have the same version win as won the last round. The
trick is to have the right version win on the first round. And my
solution is to use the second difference that exists between versions.
The version we want to lose is a step behind the one we want to win. Here
is the special code for the first round, which should be at the end of the
warrior:
search jmz.f search, {count ;; find the other copy
count sne.i #-100, {count ;; skip the next line if we are #2
jmp $0, {count
dat 0, 0 ;; just empty core after the warrior
All this is is a backwards scan for the other copy. On every cycle,
count is decremented. When we find the other copy, we compare the two
counts. If they are different, then that version must be one step behind
the other version, and so it kills itself. If the two counts are the
same, then we win by waiting for the other version to find us and kill
itself. And if we are fighting some other warrior, we surely die.
And that's all there is to it. Here is my handshake code (21 lines)
;redcode-94
;assert 1
;author Kurt Franke
;name Socrates
;strategy Know Thyself
;strategy Handshake -> wait
wait jmp wait ;; real tough guy :-)
;; - - - - - - - - - handshake constants - - - - - - - - - - - - - - -
PIN 1
ENEMYP equ 1 ;; flag for enemy warrior
WINP equ 2
LOSEP equ 3
HWARRIOR equ wait ;; our regular warrior
;; - - - - - - - - - handshake initialize - - - - - - - - - - - - - -
enemy ldp #ENEMYP, #0
jmn HWARRIOR, $enemy ;; if flag is set, fight
last ldp #0, #0
sne #-1, $last ;; check for first round
jmp search ;; special routine for first round
;; - - - - - - - - - - - - try handshake - - - - - - - - - - - - - -
shake jmz lost, $last ;; the actual handshake
won stp #0, #LOSEP ;; reset the expected flag
stp #1, #WINP
loser ldp #LOSEP, #0 ;; and then check it
jmz foe, $loser
jmp 0 ;; wait
lost stp #0, #WINP ;; reset the expected flag
stp #1, #LOSEP
winner ldp #WINP, #0 ;; and then check it
jmz foe, $winner
dat $0, $0 ;; lose again
;; - - - - - - - - - - - flag enemy warrior - - - - - - - - - - - - - - -
foe stp #1, #ENEMYP
jmp HWARRIOR
;; - - - - - - - - - special first round routine - - - - - - - - - - - - -
;; Determine who moved first without pspace
;; 1st version waits till timeout and second kills itself
search jmz.f search, {count ;; find the other copy
count sne.i #-100, {count ;; skip the next line if we are #2
jmp $0, {count
;; - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
end enemy
Try running it:
protector:~/redcode$ pmars -r 100 socrates.red socrates.red -b
Socrates by Kurt Franke scores 300
Socrates by Kurt Franke scores 0
Results: 100 0 0
Pretty nifty, eh? I'll conclude with two remarks:
First, it is pretty easy to make all of your warriors handshake with
each other to split wins and get more points. However I think only a real
weasel would do this. You are on your honor. ;-)
Second, I can't see any reason (except lack of space) why most of the
warriors on the hill wouldn't add a handshake. It adds at least 5 points
to the score and only costs two cycles once the enemy flag is set. But I
think it would be very boring if everyone started using handshakes. I
like variety. So perhaps we should just legislate against PIN numbers on
the koth servers. (?)
I hope you found all of this interesting,
- Kurt Franke
kurt@protector.lovett.rice.edu
______________________________________________________________________________
Questions? Concerns? Comments? Complaints? Mail them to people who care:
authors: Beppe Bezzi <bezzi@iol.it> or Myer R Bremer <bremermr@ecn.purdue.edu>
contributors this issue: Kurt Franke <kurt@protector.lovett.rice.edu>
|
