Issue 18 February 26, 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.
______________________________________________________________________________
Greetings.
It's me again--unfortunate for you. Bezzi threatened to fire me for the poor
job I did on the last issue. I didn't think you could get fired from a job
you did for free. Apparently the rules are different in Europe.
My apologies to J K Lewis for mistaking him with J K Wilkinson. How could I
have done that? JKL = redcode-quiet. JKW = redcode-verbose. Won't happen
again. Remember the deadline for Lewis' tournament is March 5th. Mail
entries to jklewis@umich.edu.
______________________________________________________________________________
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
Last challenge: Mon Feb 26 05:33:26 PST 1996
# %W/ %L/ %T Name Author Score Age
1 43/ 30/ 26 Grilled Octopus v0.2 David Boeren 157 16
2 42/ 34/ 24 Chameleon M R Bremer 149 10
3 39/ 32/ 29 Thermite II Robert Macrae 146 539
4 38/ 30/ 32 Tornado 3.0 Beppe 145 340
5 37/ 31/ 31 T.N.T. Maurizio Vittuari 143 264
6 33/ 22/ 46 Barrage Anton Marsden 143 12
7 31/ 18/ 51 Clisson P.Kline 143 25
8 42/ 42/ 15 quiz Schitzo 142 938
9 41/ 40/ 19 Frontwards v2 Steven Morrell 142 791
10 43/ 46/ 11 Memories Beppe Bezzi 139 77
11 34/ 30/ 36 Trade Wars P.Kline 139 34
12 40/ 42/ 17 Mirage 1.5 Anton Marsden 138 452
13 43/ 48/ 9 Wild Card M R Bremer 137 92
14 28/ 18/ 54 Impfinity v4g1 Planar 137 738
15 26/ 15/ 59 Hazy Shade Of Winter John K W 137 167
16 29/ 22/ 49 Haziest Shade Of Winter John K W 137 4
17 39/ 41/ 20 Wraith Kurt Franke 137 18
18 37/ 38/ 25 C I A Anders Ivner 137 1
19 34/ 32/ 33 Torch t18 P.Kline 136 1317
20 29/ 24/ 47 Link Kurt Franke 134 41
21 32/ 29/ 39 Jack in the box Beppe Bezzi 134 1305
22 23/ 12/ 66 Evol Cap 6.6 John Wilkinson 134 553
23 26/ 19/ 55 Blizzard Anton Marsden 132 258
24 30/ 29/ 40 patroclus Kafka and Kurt 131 334
25 37/ 46/ 17 test me 129 2
Weekly age: 159
New warriors 10 Turnover/age rate 6%
Average age: 333 ( 308 last week, 249 the week before )
Average score: 139 ( 139 last week, 132 the week before )
The 25 warriors on the hill are represented by 14 authors.
( Lost Karl Lewin's Night Train )
Boeren has been nothing but dominate. Grilled Octopus v0.1 was on top of
the hill most this week, but was strongly contested by Memories and
Frontwards v2, each claiming the number one rank for more than a few
challenges. And then replicators Night Train and Hector 3 fell off the hill
hurting the scanners' scores and dropping them into the middle of the pack.
None the less, David started a new test series to consolidate his position.
Clisson took the top spot for awhile as testing continued, but quickly fell
to the successor--Grilled Octopus v0.2. Since it's introduction, only
Chameleon--another new warrior--has been able to steal the lead, and for
only one challenge. Can his dominance continue?
This week also marks the return of pspace. Somewhat neglected for a few
months, pspace is back with a vengence. Is it any coincidence that bombers
started falling off the hill shortly after? Blue Funk 5 was doing very well
at the beginning of the week, but with Boeren testing his new pspacer and
Trade Wars making it onto the hill, it was easy for Marsden's new paper
warrior to force Blue off. seventyfive was pushed off much the same way.
With the code bloat associated with p-warriors, qscanners should score some
easy points. Thermite has enjoyed a boost in score, and Barrage is doing
well. I suppose we'll see the real test of booting versus decoy making as
more qscans appear to do battle. The pspace cycle: Here we go again . . .
Vampires are attempting a comeback with Paulsson's classic myVamp seen on
the hill earlier. Wraith has managed to cling onto the lower half of the
heap. It's too early to tell if it has staying power.
M R Bremer (me) finally returns with Chameleon and Wild Card. Both programs
have been ranked number one albeit for only one challenge each. Ivner didn't
much care for the loss of seventyfive, so he's back with C I A. And what
exactly are intelligence bombs anyway?
Lots of significant losses this week with Night Train and Boombastic falling
the farthest. Night Train by Karl Lewin stops at 10th place in the Hall Of
Fame at an age of 755. Maurizio Vittuari's Boombastic will rest at 21st
place at an age of 521.
Torch t17 and Jack in the box are both still alive and kicking. Will a
pspace comeback help or hurt the dynamic duo? Find out next week.
______________________________________________________________________________
94 - What's New
# %W/ %L/ %T Name Author Score Age
1 43/ 30/ 26 Grilled Octopus v0.2 David Boeren 157 16
2 42/ 34/ 24 Chameleon M R Bremer 149 10
6 33/ 22/ 46 Barrage Anton Marsden 143 12
11 34/ 30/ 36 Trade Wars P.Kline 139 34
13 43/ 48/ 9 Wild Card M R Bremer 137 92
16 29/ 22/ 49 Haziest Shade Of Winter John K W 137 4
17 39/ 41/ 20 Wraith Kurt Franke 137 18
18 37/ 38/ 25 C I A Anders Ivner 137 1
20 29/ 24/ 47 Link Kurt Franke 134 41
25 37/ 46/ 17 test me 129 2
______________________________________________________________________________
94 - What's No More.
# %W/ %L/ %T Name Author Score Age
26 38/ 45/ 18 Test J E Long 131 27
26 36/ 48/ 16 test Beppe Bezzi 125 3
26 33/ 45/ 22 Boombastic Maurizio Vittuari 122 521
26 2/ 2/ 0 test P.Kline 6 23
26 36/ 50/ 14 seventyfive Anders Ivner 122 311
26 1/ 1/ 1 Blur Anton Marsden 6 27
26 17/ 12/ 72 Night Train Karl Lewin 121 755
26 25/ 30/ 45 Blue Funk 5 Steven Morrell 120 152
26 22/ 23/ 56 Hector 3 Kurt Franke 120 266
26 1/ 1/ 2 Grilled Octopus v0.1 David Boeren 5 192
______________________________________________________________________________
94 - What's Old
# %W/ %L/ %T Name Author Score Age
19 34/ 32/ 33 Torch t18 P.Kline 136 1317
21 32/ 29/ 39 Jack in the box Beppe Bezzi 134 1305
8 42/ 42/ 15 quiz Schitzo 142 938
9 41/ 40/ 19 Frontwards v2 Steven Morrell 142 791
14 28/ 18/ 54 Impfinity v4g1 Planar 137 738
22 23/ 12/ 66 Evol Cap 6.6 John Wilkinson 134 553
3 39/ 32/ 29 Thermite II Robert Macrae 146 539
12 40/ 42/ 17 Mirage 1.5 Anton Marsden 138 452
4 38/ 30/ 32 Tornado 3.0 Beppe 145 340
24 30/ 29/ 40 patroclus Kafka and Kurt 131 334
5 37/ 31/ 31 T.N.T. Maurizio Vittuari 143 264
23 26/ 19/ 55 Blizzard Anton Marsden 132 258
______________________________________________________________________________
HALL OF FAME
* means the warrior is still running.
Pos Name Author Age Strategy
1 Torch t18 P.Kline 1317 * Bomber
2 Jack in the box Beppe Bezzi 1305 * P-warrior
3 quiz Schitzo 938 * Scanner/ bomber
4 Iron Gate 1.5 Wayne Sheppard 926 CMP scanner
5 Agony II Stefan Strack 912 CMP scanner
6 Blue Funk Steven Morrell 869 Stone/ imp
7 Thermite 1.0 Robert Macrae 802 Qscan -> bomber
8 Frontwards v2 Steven Morrell 791 * One shot scanner
9 Blue Funk 3 Steven Morrell 766 Stone/ imp
10 Night Train Karl Lewin 755 Replicator
11 Impfinity v4g1 Planar 738 * Stone/ imp
12 HeremScimitar A.Ivner,P.Kline 666 Bomber
13 La Bomba Beppe Bezzi 650 Qscan -> replicator
14 myVamp v3.7 Paulsson 643 Vampire
15 Armory - A5 Wilkinson 609 P-warrior
16 juliet and paper Bremer & Bezzi 607 P-warrior
17 Phq Maurizio Vittuari 589 Qscan -> replicator
18 Hector 2 Kurt Franke 561 Qscan -> replicator
19 Evol Cap 6.6 John Wilkinson 553 * Replicator
20 Thermite II Robert Macrae 539 * Bomber
21 Boombastic Maurizio Vittuari 521 Bomber
22 B-Panama X Steven Morrell 518 Stone/ replicator
23 Mirage 1.5 Anton Marsden 452 * Scanner/ bomber
24 NC 94 Wayne Sheppard 387 Stone/ imp
25 Cannonade P.Kline 382 Stone/ imp
Anton Marsden joins the ranks of the Hall of Fame replacing Torch t17.
______________________________________________________________________________
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 Feb 26 06:05:01 PST 1996
# %W/ %L/ %T Name Author Score Age
1 52/ 14/ 34 Pretentious Ian Oversby 191 78
2 52/ 19/ 29 Gluttony JEL 184 88
3 58/ 33/ 9 Assassin XII-b Andy Nevermind 184 52
4 55/ 35/ 10 test J E Long 176 57
5 55/ 34/ 11 CoreWarrior M R Bremer 175 38
6 48/ 22/ 31 Shabba J E Long 174 70
7 47/ 20/ 33 Hyakutake C/1996 B2` Scott Manley 174 59
8 45/ 29/ 26 Uvavu II (revistited) Scott Manley 162 80
9 45/ 29/ 26 Uvavu II (test 25) Scott Manley 161 76
10 49/ 41/ 10 Paper Eater David Boeren 157 69
11 49/ 43/ 8 Assassin XI Andy Nevermind 155 63
12 40/ 39/ 20 Swan's Lament Chris Arguin 142 26
13 38/ 35/ 27 Jezebel's Revenge CXX Andy Nevermind 141 67
14 41/ 43/ 16 Lenochka 3 Julian 140 7
15 44/ 49/ 6 Fei-Oh v0.6 David Boeren 139 75
16 31/ 25/ 44 AB Scanner 2.9.3 Chris Hodson 138 27
17 42/ 46/ 12 XPEH 2 Julian 138 6
18 39/ 42/ 19 Swan's Lament Chris Arguin 136 10
19 39/ 46/ 14 test Julian 133 8
20 39/ 54/ 7 XPEH Julian 123 48
21 28/ 45/ 27 Evil Elf Andy Nevermind 112 5
22 30/ 50/ 20 Pommes-Ketchup V1.03 S. Schroeder 110 4
23 28/ 47/ 26 Inchworm J E Long 109 64
24 27/ 51/ 23 Blubbi V1i.0b bjoern guenzel 103 11
25 28/ 68/ 4 A-Cluster v2 Iain Hogg 88 1
Is your only goal in life getting a warrior on the 'pro' hill? ( If it is
you should seek professional help. ) Want to know the tricks those mensa
geniuses use to tweak their programs? Just mail your warrior to us at Core
Warrior. We'll attempt to point out strengths, weaknesses, and strategies
you can use to improve your core warring skills. And requests for hints are
always welcome.
______________________________________________________________________________
The Hint
Modern replicators breed like epileptic rabbits. Scanners need to take
extreme measures to eradicate Timescape and DieHard style papers. The
trick is to stun them as rapidly and thoroughly as possible. Simple spl/jmp
bombs aren't enough. Even spl/spl/jmp bombs may be inadequate without a
multipass coreclear. So what type of payload should a warrior deliver to
maximize stunning power? A paper product typically employs parallel
processes to achieve replication. I tested 6 stun bombs by executing 6
parallel processes on the first bomb instruction. The number to total
processes after 100, 200, 300, etc. cycles was recorded and entered into
the graph below.
#Processes
1000 +
- 56
900 + 56
-
800 + 5
- 6 2
700 + 5 2
- 6
600 + 5 2 1
- 6 2 1
500 + 5 1
- 6 2 1
400 + 56 2 1 4 4 34 34
- 3 3
300 + 56 2 1 3
- 2 13 3 4
200 + 56 13 4 4
- 123 4
100 + 12356 4
- 4
0 +------+------+------+------+------+------+------+------+------+------+
100 200 300 400 500 600 700 800 900 cycles
These are the payloads delivered:
1. spl 0 2. spl 0 3. spl 0 4. spl -1 5. spl #N+1 6. spl #2, 0
jmp -1 spl 0 spl 0 spl -1 . . . mov -1, }-1
jmp -2 spl 0 spl -1 mov -N, }N
(spl 0) (spl -1)
The raw data:
Cycle
---
100 70 85 89 (105) 49 ( 67) 101 85
200 132 161 142 (188) 96 (128) 202 180
300 195 238 189 (265) 129 (193) 303 277
400 257 314 228 (330) 192 (252) 402 375
500 319 391 263 (393) 193 (315) 501 474
600 382 466 294 (454) 240 (378) 598 574
700 444 543 325 (513) 341 (443) 695 673
800 507 619 356 (564) 384 (504) 786 773
900 569 696 383 (617) 385 (567) 877 873
1000 632 771 414 (676) 384 (628) 974 973
Initially, the bombs were limited to 3 instructions, but the split carpets
performed so miserably I executed them with an extra instruction just to see
how much improvement it would make. The 4 line versions are shown in
parentheses. The graph contains the 3 line versions.
The analysis:
The spl/jmp bomb is used for a reference point. It is not effective against
modern replicators.
The spl/spl/jmp bomb is more powerful and attacks three locations as opposed
to two. There is little difference if the jmp instruction is the first to
execute.
The spl 0 carept is only effective if four instructions long. However, its
performance degrades rapidly if the lead spl 0 is not the first instruction
to be executed.
The spl -1 carpet is ineffective even in the 4 instruction version. A 5
instruction carpet with the middle split executing first generates about the
same number of processes as the spl/spl/jmp bomb.
The spl/mov bombs generate the most amount of processes. The incendiary
bomb is the equivalent of a 6 instruction split carpet when the move is
executed. However, sometimes the other instruction is executed reducing the
effectiveness of the bomb. In addition, the powerful part ( the move )
attacks only one instruction. It is very easy to miss the active part of a
silk module with this bomb. The last spl/mov bomb is slightly less effective
in stunning capability, but is more effective overall. If the move is the
first instruction to be executed than the bomb is the equivalent of a 6
instruction spl 0 carpet ( same as the incendiary bomb). However, if the
spl #2 is first to execute, we get a continously expanding split carpet--
forty-three instructions long after 1000 cycles. This could even overwrite
other silk modules. There are some disadvantages however. The split carpets
can color core thus distracting the scanner. And the bomb only attacks two
instructions.
Real life test:
Wild Card is a spl/spl/jmp scanner. At this time it is scoring around 131
points on the hill. A similar scanner using spl/mov bombs only scored
111 points. Changing the bombs to spl/spl/mov increased the score to 115
points, BUT the stun power of this bomber is not as good as the spl/mov.
The double split spends more time increasing the size of the carpet and not
stunning. The increase in the score is most likely due to the increased
coverage of the bomb. I didn't spend any time trying to integrate the bomb
components into the executing code. The resulting warrior is more fragile
and larger resulting in an increase in losses. It also attacks the split
carpets. By eliminating these problems, perhaps someone could write a more
credible, quick stunning scanner.
______________________________________________________________________________
Questions? Concerns? Comments? Complaints? Mail them to people who care:
Beppe Bezzi <bezzi@iol.it> or Myer R Bremer <bremermr@ecn.purdue.edu>
|
