Issue 56 3 March, 1997
______________________________________________________________________________
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
Beppe Bezzi web page - http://www.aspide.it/freeweb/Bezzi
______________________________________________________________________________
Greetings.
Traffic on the hills is very little, and I'm rather busy, so I delayed the
issue. I hope as soon as traffic will come back to old values to begin
publishing CW weekly as usual.
This week we have an interesting hint by Robert Mcrae on hill balance of
power and how much a new warrior, He Scans Alone in the example, will move
the equilibrium.
--Beppe Bezzi
______________________________________________________________________________
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 38.6/ 23.7/ 37.7 Head or Tail Christian Schmidt 153.5 14
2 35.4/ 18.1/ 46.5 RetroQ P.Kline 152.7 26
3 38.0/ 26.3/ 35.6 Nine Seven Six M R Bremer 149.7 90
4 32.1/ 15.4/ 52.5 Return Of The Jedimp John K W 148.8 226
5 35.4/ 23.7/ 40.9 Gigolo Core Warrior staff 147.2 280
6 46.2/ 47.1/ 6.7 He Scans Alone P.Kline 145.2 41
7 39.5/ 35.6/ 24.8 Tides v0.1 Ian Oversby 143.5 10
8 32.0/ 21.7/ 46.4 unrequited love kafka 142.3 324
9 40.4/ 38.7/ 20.9 NCC-1701-A Philip Kendall 142.1 29
10 31.0/ 20.7/ 48.4 Trident^2 John K W 141.3 156
11 39.8/ 38.4/ 21.7 Solomon v0.8 Ian Oversby 141.2 72
12 39.5/ 39.4/ 21.1 QFalcon (vii) Ian Oversby 139.6 19
13 39.4/ 40.5/ 20.1 The Machine Anton Marsden 138.2 132
14 32.1/ 26.2/ 41.8 Terkonit 0.4 Christian Schmidt 138.0 32
15 34.0/ 30.3/ 35.7 Fast Fast Fast v2 Franz 137.6 40
16 37.7/ 38.7/ 23.6 Damage Incorporated Anton Marsden 136.6 267
17 35.5/ 36.0/ 28.5 Scanitator 4 Christian Schmidt 135.1 28
18 27.7/ 20.4/ 51.8 Impish v0.2 Ian Oversby 135.1 265
19 32.7/ 31.3/ 36.0 Instant Ogre Edgar 134.1 9
20 36.9/ 40.4/ 22.7 Instant Wolf Edgar 133.4 47
21 41.0/ 49.0/ 10.0 Memories Beppe 132.9 140
22 37.7/ 44.4/ 17.8 Oblivion Ian Sutton 131.1 91
23 26.2/ 21.9/ 51.9 blue spark 0.04 bjoern guenzel 130.5 6
24 31.1/ 33.5/ 35.4 Dust 7.0 Justin Kao 128.7 1
25 4.0/ 0.0/ 0.0 Test I Ian Oversby 12.0 2
Monthly age: 13 this month ( 9 last issue, 15 the issue before )
New warriors: 5 Turnover/age rate 38%
Average age: 93 ( 119 last issue, 115 the issue before )
Average score: 135 ( 134 last issue, 138 the issue before )
The top 25 warriors are represented by 15 independent authors: Oversby with
5; Schmidt with 3; Kline, JKW, Marsden and Edgar with 2. All others with one
lonely warrior each.
The fall of Blur2 and Probe, poor Anton :), made the hill much younger. Head
or Tail kept the top spot all time.
______________________________________________________________________________
94 - What's New
# %W / %L / %T Name Author Score Age
3 39.8/ 35.0/ 25.2 Tides v0.1 Ian Oversby 144.5 1
13 32.5/ 32.5/ 35.0 Instant Ogre Edgar 132.5 1
25 23.2/ 22.1/ 54.7 blue spark 0.04 bjoern guenzel 124.4 1
11 37.4/ 38.8/ 23.8 Test I Ian Oversby 136.0 1
25 29.5/ 35.0/ 35.5 Dust 7.0 Justin Kao 124.0 1
Tides is the only warrior to enter the hill in the top spots.
______________________________________________________________________________
94 - What's No More
# %W / %L / %T Name Author Score Age
26 1.9/ 0.9/ 1.2 Pulp v0.5 Ian Oversby 7.0 5
26 35.4/ 46.4/ 18.3 Probe Anton Marsden 124.4 396
26 24.5/ 26.0/ 49.4 Eva Peron David Randel 123.0 8
25 23.2/ 22.1/ 54.7 blue spark 0.04 bjoern guenzel 124.4 1
26 35.1/ 46.4/ 18.5 Blur 2 Anton Marsden 123.8 403
The two older warriors leave the hill.
______________________________________________________________________________
94 - What's Old
# %W / %L / %T Name Author Score Age
8 29.9/ 21.5/ 48.6 unrequited love kafka 138.3 324
5 34.2/ 27.2/ 38.6 Gigolo Core Warrior staff 141.2 280
16 35.4/ 41.5/ 23.1 Damage Incorporated Anton Marsden 129.4 267
19 23.8/ 19.8/ 56.5 Impish v0.2 Ian Oversby 127.8 265
6 28.8/ 16.9/ 54.4 Return Of The Jedimp John K W 140.7 226
No new entries, two losses
______________________________________________________________________________
OLD 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 Rosebud Beppe Bezzi 993 Stone/ imp
15 Iron Gate 1.5 Wayne Sheppard 926 CMP scanner
16 T.N.T. pro Maurizio Vittuari 925 Bomber
17 Agony II Stefan Strack 912 CMP scanner
18 Barrage Anton Marsden 876 Qscan -> replicator
19 Blue Funk Steven Morrell 869 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
______________________________________________________________________________
NEW HALL OF FAME
* means the warrior is still active.
Pos Name Author Age Strategy
1 Probe Anton Marsden 403 Q^2 -> Bomber
2 Blur 2 Anton Marsden 396 Scanner
3 unrequited love kafka 324 * Stone/ imp
4 Falcon v0.3 Ian Oversby 275 P-warrior
5 Gigolo Core Warrior staff 280 * Q^2 -> Stone/ imp
6 Damage Incorporated Anton Marsden 267 * Q^2 -> Bomber
7 Impish v0.2 Ian Oversby 265 * Stone/ imp
8 Return Of The Jedimp John K W 226 * Q^2 -> Stone/ imp
9 Rosebud Beppe 218 Stone/ imp
10 Q^2 Miro Anders Ivner 214 Q^2 -> Scanner/ bomber
11 Instant Wolf 3.4 Edgar 205 P-warrior
12 Goldfinch P.Kline 201 P-warrior
13 Simple v0.4b Ian Oversby 197 Stone/ imp
14 ompega Steven Morrell 189 Stone/ imp
15 Frogz Franz 172 Paper
16 Trident^2 John K W 156 * Stone/ imp
17 Memories Beppe 140 * Scanner
18 The Machine Anton Marsden 132 * Scanner
19 Tiberius 3.1 Franz 130 P-warrior
20 CC Paper 3.3 Franz 107 Paper
21 mrb-test m r bremer 106 ?
22 T.N.T. pro Maurizio Vittuari 105 Bomber
23 Jack in the box II Beppe Bezzi 100 P-warrior
Oblivio and NineSevenSix are near to enter the new HoF
______________________________________________________________________________
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
# %W / %L / %T Name Author Score Age
1 51.2/ 38.7/ 10.1 Pentagram 2.a J.A.Denny 163.8 23
2 48.1/ 35.3/ 16.7 Hexagram J.A.Denny 160.9 21
3 45.5/ 31.3/ 23.2 Flimsy v0.6 Ian Oversby 159.8 22
4 49.2/ 41.0/ 9.8 Goothmonger v1.2 Ian Sutton 157.3 46
5 41.1/ 28.8/ 30.0 Scanitator 3.0 Christian Schmidt 153.4 28
6 43.4/ 33.9/ 22.7 Versatility 1.7 Ross Morgan-Linial 152.9 49
7 40.7/ 28.7/ 30.6 Dust 7.0 Justin Kao 152.7 1
8 40.6/ 29.9/ 29.5 Dust 0.7.5 Justin Kao 151.4 48
9 46.3/ 43.0/ 10.7 Scankiller 0.1 Christian Schmidt 149.6 33
10 46.1/ 44.0/ 9.9 The Stainless Steel Rat Christian Schmidt 148.1 44
11 43.3/ 39.4/ 17.3 Microsoft Office '96 Justin Kao 147.3 98
12 32.6/ 17.9/ 49.5 DemonSpawn J.A.Denny 147.2 31
13 44.3/ 43.4/ 12.3 Short Sword 4 JS Pulido 145.2 3
14 41.7/ 38.6/ 19.7 Vivid Radiation 2.0 Matt Lewinski 144.9 70
15 44.0/ 43.1/ 12.9 Flamberge 13.4 Matt Lewinski 144.9 71
16 38.8/ 33.0/ 28.2 Zorm-B Anonymous 144.7 6
17 37.9/ 32.0/ 30.0 Apocalypse Matt Lewinski 143.8 79
18 35.9/ 30.3/ 33.8 Escargot 0.5 Justin Kao 141.5 68
19 32.9/ 27.1/ 40.0 Quantum Christian Schmidt 138.8 25
20 41.4/ 45.1/ 13.5 Mostly Harmless v1.1 Justin Kao 137.7 52
21 32.9/ 30.4/ 36.7 Heartworm Edgar 135.3 19
22 39.7/ 44.5/ 15.8 Bash the Rat V1.4 Ian Sutton 134.9 63
23 37.3/ 47.2/ 15.5 Dwa Michaly b Waldemar Bartolik 127.4 9
24 32.2/ 43.1/ 24.8 Time Lag 1c3 Ilmari Karonen 121.3 5
25 33.9/ 51.5/ 14.6 Dwa Michaly a Waldemar Bartolik 116.4 13
______________________________________________________________________________
The Hint
The Likely Impact of He Scans Alone on the Hill Population
For some time I have been playing with ideas about how populations of warriors
evolve. I think HSA will have an important impact on the hill, so it seemed
like a good time to formalise what I've been doing and *test* my views. What
difference, if any, will HSA make?
Background.
Imagine that we have a hill with very many warriors, all *sharing* a small
number of public designs. Players are continually trying to get warriors on
the hill by *selecting* from among those designs, but nothing new is written.
As new warriors challenge the hill, the population -- or more precisely, the
proportion of each design in the hill population -- will fluctuate around some
equilibrium. If you know the scores the warriors average against each other
you can calculate what it is.
To make it concrete, imagine that all the warriors where copies of Blur, Labo
or Torn, a scanner, paper and bomber respectively. This table shows the
average points scored by each against the others, so Blur scores 165 against
Labo. (For the moment lets ignore the uncertainty due to only having estimates
of the average scores).
Blur Labo Torn
Blur 142 165 99
Labo 117 101 265
Torn 189 22 149
>From this you can calculate the "equilibrium" population, which is
Blur Labo Torn
56.8% 25.0% 18.2%
At equilibrium, every warrior scores equally. In this case, all warriors score
139.93 points. For example for Blur, 56.8% x 142 + 25.0% x 165 + 18.2% x 99 =
139.93. (Nearly)
What happens if you deviate from equilibrium? Well, in this case, if you add
more papers the scanners move to the top of the table and more are added,
which push off the excess papers. This is a stable equilibrium and the
population will oscillate around it, never getting too far out of line before
new warriors arrive to take advantage of the disequilibrium.
Life gets more complicated as the number of designs is increased, because some
may not be present in the equilibrium at all. This means that you can't get an
explicit formula for the equilibrium, but you can still simulate easily. Just
model the hill, run it for several thousands of new submissions and
"hopefully" it will end up near equilibrium. If you prefer something more
solid than hope, replicate this process many times with slightly differing
tables of scores and you obtain a measure of the stability of the equilibrium.
[Theory Note -- I am ignoring potential problems here. In particular it is not
in general true that there is only 1 equilibrium -- consider a warrior with a
perfect handshake, but which loses to everything else. It may also be possible
for non-point equilibria and chaotic behaviour to exist, but I don't think
this matters too much in practice.]
Phew.
So with a small amount of code (Pascal source and Windows executable
available) we can take a bunch of warriors and ask which ones would make it
into the equilibrium.
The warriors in the equilibrium represent in a compact way the current "state
of the art". A new warrior pushes the state forward if it makes it into this
population. Warriors which don't make it can be tested against the population
to find out by how much they miss, which provides a global ranking in the
style of Mount Olympus.
So How about HSA?
As a practical example, I took a dozen warriors intended to cover a spread of
types; for simplicity I ignored P-space and Qscans. I make no great claims for
the construction of the list, I just picked some of the code I test against.
My lack of hill-success suggests this set is profoundly flawed :-/
He Scans Alone
Blur 2
Harmony
La Bomba Paper Copy
Die Hard Paper Copy
D-Clear Version (Everyone should have one:)
Rosebud
Impfinity v4g1
Seventyfive
Torch t18
Tornado 3.0 Copy
Tornado Dat Version (Minimalist DAT bomb, DAT clear)
Because I'm really interested in generic types like "bomber" and "paper", I've
usually just taken the central element, perhaps with a small decoy. "Copy"
means I've chopped the warrior out of something larger (Apologies to the
authors for any breakages:) and "Version" means I've adapted it in some way.
This is the resulting score matrix, based on 100 fights.
H B H L D D R I S T T T
HSA 150 171 152 177 195 102 141 91 102 171 127 112
Blur 120 142 199 165 97 148 145 212 167 117 132 99
Harmony 146 85 142 130 150 189 110 175 108 103 122 134
La Bomba 111 117 136 101 105 268 102 99 177 118 134 265
Die Hard 93 127 90 99 100 148 103 97 174 132 136 164
D-Clear 1 177 151 111 16 76 150 97 109 186 130 112 102
Rosebud 156 139 173 105 100 181 101 91 212 176 184 187
Impfin v4g1 205 74 112 105 109 175 118 102 195 144 166 223
Seventy5 198 128 189 102 69 114 71 81 150 113 114 201
Torch t18 120 165 181 136 93 154 89 123 167 138 116 135
Tornado 3.0 160 153 164 119 85 181 82 91 174 152 133 155
Tornado Dat 184 189 161 22 74 195 106 49 99 138 134 149
We can now crank the handle on our simulated hill, to obtain the following
equilibria. To get the following estimates I ran a 10000 rounds on the hill,
with HSA first excluded and then included. Weights converged to about 4
decimal places.
Pre-HSA HSA Score
HSA -- 18% 140.68
Blur 39% 29% 140.68
Harmony 0% 0% 122.77
La Bomba 14% 6% 140.68
Die Hard 0% 0% 120.85
D-Clear 1 0% 0% 127.81
Rosebud 23% 13% 140.69
Impfin v4g1 0% 12% 140.69
Seventy5 0% 3% 140.66
Torch t18 7% 0% 134.45
Tornado 3.0 0% 0% 136.56
Tornado Dat 17% 19% 140.71
The table shows the equilibrium weights Pre- and Post- HSA. The Score column
shows the score of each warrior post-HSA. Those with weights greater than 0%
should have identical scores, so the 140.68 to 140.71 range is due to
incomplete convergence. Those are not in the equilibrium can be ranked on
their score against the equilibrium warriors, from Tornado 3.0 which is just
out, to Die Hard and Harmony which are far out.
The above estimates are subject to error due to using only 100 rounds per
battle to estimate the strength of the warriors. As a test on this
uncertainty, I ran another 100 trials and in each trial perturbed every score
by +-10%. This resulted in interquartile spreads of roughly 6.5% for the
warriors which are in the equilibrium -- for example HSA had a weight of
between 15.6% and 21.7% half of the time. In practice this means that a
difference of weights of less than 10% _within_ a column could well be due to
chance, though we can be more confident about differences _between_ columns.
This is because a lucky result should benefit a warrior in _both_ tests
because the same scores were used.
We cannot be confident that Tornado 3.0 is really out of the equilibrium,
because it entered the equilibrium in 32 of these 100 perturbed trials, but
Harmony and Diehard never make it in.
Interpretation
Firstly, the warriors not in the equilibrium. It is quite possible that they
would get in if a different starting set were considered (in particular a set
containing a high-scoring warrior against which they excel), but against these
warriors they don't make it.
- Harmony does not waste time carpeting trails but is less effective against
Silks and loses to anything with SPL bombs.
- Die Hard does not split fast enough to over-write opponents, leading to too
many draws.
- D-Clear does well against the scanners, but little else.
Secondly, those in or borderline.
- Blur is highly effective and forms a large part of both populations. HSA can
lose a lot of time bombing Blur's carpet. It also handles Impfinity well.
- La Bomba is significantly hurt by HSA but is not, as I had feared,
eliminated. It spreads aggressively enough to score very well against anything
which can't stun it.
- Impstones overall are little affected; Rosebud is badly hit by HSA but
Impfinity takes up the slack. Impfinity seems to gain from its phenomenal
split rate, as hitting anything except the stones does not stun. Even with a
stone hit, hundreds of further bombs will be thrown by the other. HSA spends a
lot of time re-carpeting the DJN trails, so the imp-stones of the future want
messy cores... Hmmm... :)
- The fast bombers as a group just break even, even though with HSA pure
scanners now comprise a juicy 47% of the population. I suspect the rise of
Impfinity is what holds back the fast bombers -- the result would have been
different if I'd just used Rosebud as the generic stone-imp representative. My
DAT version of Tornado picks up enough against the scanners and rosebud to
compensate of worse performance against everything else. For the bombers,
dealing with imps will rise up the agenda and papers will recede somewhat.
- HSA itself takes a significant chunk of the population but it could use a
tweak against Impfinity. It doesn't replace Blur, partly because once it hits
a DJN trail it stops scanning and carpets the whole thing. Blur's more casual
approach to bombing is an advantage when the core is full of rubbish.
Assuming these results hold for hill warriors rather than just the examples
I've tested, then HSA *is* a substantial step forward. Silks have a major new
problem. Fortunately HSA's hyper-aggressive bombing ("positive" is the kind of
term the USAF would use:) makes it particularly vulnerable to decoys and
trails. Perhaps a stone/silk? Or a fast bomber with a good anti-imp clear?
Time to get coding...
Regards, Robert Macrae ff95@dial.pipex.com
_____________________________________________________________________________
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@etherspace.gen.nz>
|
