Ok, some things I noted about this round.
You have to be very fast, and be adaptive is crutial.

The warriors always start at the same position so everything 
is the same each round. If you lose one round, and don't change 
anything you'll lose all the others too.

Thats why I made two things:
 - Lightning fast clear
 - Self optimizing paper

In the first component speed is crutial, it boots very quickly 
and places a simple dat-clear close to the enemy. The distance 
is startes its clear from is loaded from p-space. If the opponent 
is slow/few processes there is a big change we'll find the right 
offset and 100% kill it!

When its almost done booting the clear we start with the second 
component. This is a simple Moore-style-paper (which I love!) The 
special thing about this version is its boot/setting. The paper 
uses p-space, it loads its steps from p-space and when it doesn't 
win it adjusts these steps itself! Because we can't really optimise 
before the match.... I decided to do it DURING the match :-)

No idea how I'm going to do, it all depend on what Chip is going 
to send.

Again, lots of interesting twists here. First, we have the known 
position of the opponent's warrior. But the write limit makes it 
impossible to strike at the enemy until a fang jumps twice. Plus, 
with 400 instructions to cover, I'm unlikely to cause any damage 
before he boots away. So don't bother with that strategy.

What components should I use in my p-spacer?
- Papers should do well. They can cover all of core, and 80 processes 
are enough to provide good effectiveness, plus less vulnerability to 
stun. I'm not good at papers, but I manage to put together something 
decent from scratch.

- Stones are hampered by the write limit. I would need at least three 
for full coverage, and if one gets knocked out, I would be left with 
a gap. Still, it would be good to have one. I boot four copies of 
Elven King's stone, with a seven-point imp for good measure.

- Scanners are even more restricted by the write limit. They can scan 
all of core, but they must send a fang to kill anything they find. 
Thinking about how to do it gets very complicated, so I'll forget about 
scanners.

So I construct some decent papers and stone/imps - and they all lose 
horribly on the Fortress hill! Are they really that bad? Perhaps others 
are building to the larger write limit of the hill. More likely, I'm 
going to get totally destroyed.

Oh well, all I can do is try. Everything goes into a P^3 switcher, with 
a name that (according to popular belief) can't lose, and off it goes. 
Good luck to all, and thanks for a fun tournament!

Something simple written right at the last minute. I never 
did get Corewin installed.
If Zul has put a lot of effort into this round, I hope he thrashes 
my warrior and get the win he deserves!

Fortress Hill featuring "Quick Strike", designed for CSE2005 Final Round.
It has weaknesses, however, I believe it forces the opponent to choose 
limited strategies. Will prevail until a "Fortress Protector" is deployed 
in the future. On the other hand, fortress shield will be too slow against 
scanners. What it does is quickly damages the other fortress, creates 
panic...no time to think.

This is my warrior from f hill. At the time of writing this comment, 
I _t0t4lly_ pwn the f hill:
HAS_b13 is N1 with 70(!!!) points gap to closest warrior made not by 
me. However, I'll not submit it because it slightly losses to Nenad's 
NTf5. I don't want to reveal Phluph 3 that fast, but I have to. It's 
an LP paper with BioMech's (my LP paper) surplus length trick (it's 
evident but it works very good; I haven't seen it in any other paper 
yet): length in copying code is larger than actual paper length. It 
allows to have some extra dat bombing without any special code for 
this and it also allows not to use ADD instruction and have a decent 
spreading of papers. Second thing is very important in this round:
my paper has only 5 instructions and complete boot takes only 14 
cycles, so practically noone can kill it before it's booted. Actually, 
right now noone can kill it at all ;) Just look at the hill: 0.6% of 
losses (0.1% if we exclude Fizmo's Viper to which it losses (oh no ;) 
5.3%) and it beats *everything*.  But it's partly due to PSpace. I'm 
switching boot distance with simple switch-on-loss-or-tie switcher.
Every time it's not winning, it increments a counter and next time this 
counter is multiplied with nice MOD-1 constant to get boot distance, so 
it coveres approx. every fourth cell in random pattern. Nenad is a very 
strong PRO player, but this time he is doomed (unless some kind of miracle 
happens). Why I don't want to reveal Phluph now? Look at quarterfinal 
results: 30% of participants sent HullaClones. Revealing good things too 
fast slows down the progress a lot. Release of Quake 1 engine slowed down 
devlopment of 3d engines for many years (up to present moment): most 
developers just bought it and used without changes. Ok, enough chatting, 
I'm risking to miss the deadline.