The Second Round

The Making of "No 3735928559"
=============================

Life can be so unfair! At first fiveop was my opponent. We 
both aren't that long into corewar and that is why I thought 
my chances of winning to be close to 50:50. And now? Because 
of that mysterious bug Roy is my opponent. And fiveop got John 
Metcalf. Now we both are doomed, but at least we still have a 
little chance of winning (i.e. if our opponents don't submit 
their warriors ;-)

Normal papers don't do well under LP-conditions and LP-papers 
aren't that good under "normal" conditions either. That is why 
I think, that Roy will submit a stone, imp, clear or scanner. 
I hope it to be a scanner. If not I'm doomed.

At first I was thinking about a simple stone with imps, but I 
couldn't make it work properly under both settings. So I've 
looked for something better and found clears. SD-/SSD-clears 
don't work well under LP, so a simple d-clear had to be good 
enough.

That's it.

And no, that isn't a stupid name. It has taken several hours to 
choose it. ... quite a "beefy" name ;-)

To win you need a good name, I learned that from Jens
  so this is the time to fight fire with fire.

The name: "la flufis. mrobi'o" means:
  "The Fluffy is dead" in Lojban (the language Jens loves)

For the warrior, I could not decide on a strategy so
  I used just about everything

This warrior has:
      Scan    (qscan)
      Clear   (SD)
      Stone
      Paper   (LP type)
      Imps    (3point Afield impring)
      ...
      And a boot to make it complete

And if it works..? No idea, lets find out..

Against Mizcu I haven't much chance.  Maybe I will be lucky.

For my fans: ( Fluffy <_< >_>) I tried to write a stone/imp/partial clear
Depending on the LP settings.......I might get owned
With a super optimized cool qscan, though
So that makes me semi happy
But I got imps right
And qscans
That's good enough to lose

Few modern warriors would score good with such rules
but many older ones feel just fine: Iron Gate, Blue Funk,
B-Panama X worked with only a few changes. I've tried 
different strategies, but this worked best amongst my warriors.
It only looses to papers and spl/jmp scanners with good gates 
(Romulan Warbird for example). Let's hope that datagram's
entry isn't one of those. And it has some problems if
maxprocesses is 24. It'd be ridiculous if it is.

Fiveop is quite capable of writing and submitting any strategy.  The two different 
process counts make paper awkward, so Fiveop will probably avoid paper.

What will Fiveop assume I will submit?  The most obvious choice would be a scanner.  
Fiveop most likely knows I won't choose a scanner for this very reason.

Thinking about what to send doesn't seem to get me anywhere, so I'll flip a coin.  
Heads a scanner, tails a bomber...

A bomber it is!  Spl/jmp stun bombs work well under both process counts so I'll 
use these.  Just in case, I'll add a qscan, airbag and imps. :-)

I've decided to write a stone/imp, because I thought good papers won't 
be possiblle with only 20-50 processes. So only stones and scanners left and 
stones beat scanners ... This is the most obviouse choise and I hope Metcalf 
doesn't think I'm so dumb to choose it. ;)

Hi,

puuuhhh at qualification i get the right strategy. Three 
scissor are enough fodder for my Paper/Stone. But now, Chip is waiting...

After his nice work on corewin and his awesome 4 in a 
row success at Redcoders Frenzy Tournament he has my respect. I see two 
points of matter.

1) What strategy can score good under 8000 processes and under 
20-50 processes without use of MAXPROCESSES or any other detection 
of MAXPROCESSES. Try to detect if 50 Procs or more needs to spl 64 
proceeses and look if that works, and if not use LP-strategy...

2) What would Chip use ?

For 1) 

O.K, with 40 Processes you can design a nice and fast coreclear-paper 
with a lower need of procs per instance. But here, we don't know if 
it's 20 or 30 or 49. So common paper are not a good choice.

Scanner, usually working with 1 process, so there is no difference 
for them. After a direct hit at the spl a clear running with 50 procs 
is much faster down as one runnign with 8000, but when working both doing 
their job.

Stones - While bombing stones with 20-50 procs are as much as powerfull 
then the 8000er one and as much as weak like the clear...

So, scanner or stone seems the strategy of my choice.

For 2)

It seems that Chip preferences Recon2 as a scanner. In 3 of the 4 rounds 
he won he did it with a Recon2-clone.

And i hope for that and my entry will be a Anti-Scanner-Optimized Stone/Imp 
with Qscan as nice decoy.

Hope for the best - and may be the core with me...

1. Rule #1: Don't worry about what Sascha may or may not submit. In single-elimination 
battles like this, luck is about 75% of the result, and if I try to out-guess him, I'll 
end up going in circles.

2. So, grab a bunch of the best warriors I can find, and see which ones work best at 
both 8000 and 20-50 processes. The ICWS hills are '88 only, which probably won't produce 
the strongest warriors, so I'll raid 94nop Koenigstuhl. Grab 10 of the best, do a round-robin 
at both 8000 and 30, throw out the worst, repeat, and I'm left with 6 benchmark warriors at 
8000 coresize, mostly S/Ps.

3. Now I need to see what strategy works best against them at both process sizes. I'll adjust 
to 8192 coresize later. The strongest of them seem to be the S/Ps, and I'm about to choose 
one of them, when I happen to run Fizmo's Elven King (the basis for my qualification round 
entry) against them, and discover that it beats the tar out of all of them, at all process 
limits!

4. But I already used Elven King for the last round. Will Sascha anticipate this? Who cares; 
see Rule #1. If he guesses right, I'll tip my hat to him.

5. So I adjust Elven King to 8192 coresize, and find a step size that works best. It suicides 
at a very few of the smaller process limits, but it's the best I can do on such short notice. 
I'll submit it, and make a note to add Fizmo to my Christmas list.

 There are not many warriors that work well under both normal and lp
settings but scanners and oneshots. It might be hard to make a paper for this round.
 I don't know much about Simon's warriors, but i know he has some
nice oneshots, so i decided to create an agressive warrior against oneshots
and scanners, having some anti-imp cappability too, so i can
get some points if he submits a paper.

The strat is basically a stone/clear

The stone bombs with empty core and also with an anti-clear bomb
I don't know if i should call it bomber or stone.
 It is quite fast and disruptive, and self hits to become
a slow core clear. It seems to be very effective.

 The clear only makes one SPL pass and not even complete,
for not to hit the bomber with SPL's and slow-down the warrior.
It kills the bomber with dats and remains as an anti-imp tool. 
The clear just has to stun imp spirals, stones and other clears,
and does its duty well.

Obviously this warrior has nearly no chance against papers
 and certain weird lp variations like Roy's DaubFusion.
 I finaly attemped to add a Qscan, but had no time to 
fit one and make it work on a 8192 core.
 I hope not to see something paper-ish from Simon.

As usual I didn't have the time to make a proper entry, hence the 
name ;) My initial plan was a combination of KISS (2 ssd clears), 
CIA-style bomber (which would suicide after one confirmed enemy hit 
- a safeguard against scanners, KISS should work against everything 
else in LP), and a low overhead detection for high process round 
(which would simply kill one of KISS clears if a SPL in late game 
failed). Instead, this stone/imp will have to be enough :) I'm expecting 
a scanner or maybe a light stone/imp - I think these are only regular 
strategies that will work well in both environments. My warrior should 
have advantage against these both types - in theory. I see too many 
flaws in this warrior (especially against scanners) to have any 
confidence, and that's assuming I predicted opponent's strategy!

After some thinking, I concluded this round is too sharp. 
The key is guessing the opponent strategy correctly.
I assume that at low process, a scanner is excellent.
But that scanner is beaten by many other scanners.
Now, I am back to square one.
Then I decided to go with Spiritual Blank Dimension (SBD)
I changed the variables but still couldn't find the optimum.
Last minute call will be the 100% unchange code for SBD
But I feel bad for using 100%, 'no effort' deliverable
Besides, Fizmo will not be happy to see his warrior beign smashed by Michal
So, I decided to come back to my RF24 warrior of choice 'TIME'
The overall strategy so far:
   a) Never forget to encode the word John in this warrior
   b) Select a good name
   c) Change the constant so that this warrior is 'my creation'.
On item b) I was tipped during the last IRC session... choose a good name.
Or rather, choose an innocent name such as Harmless Fluffy Bunny
It may give some advantage for being so 'dearly'
Shown as per the title:
"Harmless" (hmmm, cool) + "Fluffy" (wow) + "Bunny" (so cuteeee)
How would you expect to 'kill' this lovely 'creature'?
That was what I was been told by the author
So first I am using 66% of the original charming name and add another 34% of neutral name
On item c) I am still changing the variable to meet the 8192 cores
status now - work in progress but must submit a variant
My hands and fingers sore... not sure how many more should I try.
Weak against f-scanner which is weak to wimp which is weak to SDM
if my opponent using f-scanner, I'm finished !

I had two options: 
  a) adapt Deathstar to tourney conditions - it should do well in both lp and normal
  b) expect a scanner and make a quickhack s/i - I did that. 
It was a little troublesome to make it work in LP and it's still not very good. I'll 
send a better version to 94t later I didn't want to use any of the stones from my other 
s/i, so I took a qsilver stone approach. in short: if my opponent sends a paper or a 
clear/imp, I'm finished. if he chooses a scanner... well, then he is history. a gamble, 
as in the round 1. however, I doubt that I'll do any better now.