Crematorium - Computer Aided Speedbow Optimization Part 2

[Gorthos]

The way this works is, I start with a set of speedbow cards.  I program appropriate strategies for the computer to follow, and then let it simulate a thousand games with a single deck, seeing how many turns it takes to kill a 100 hit point tackling dummy that does not fight back.  Then I swap out one card for another, and see if the deck gets any faster.

This deck looks a lot like some other fast fire-based speedbows, but the difference is that this one has been computer optimized.  It's perfectly balanced.

Against the simulated tackling dummy, the TTW is about 4.75 turns.

Against a real opponent, of course it doesn't do that well because they're fighting back.  I haven't done manual TTW on this, though I can tell you 5 turn wins are typical, and it usually plays smooth as butter.  It takes some pretty heavy creature control to shut this thing down - it can sometimes lose two or three creatures and still pull off a 6 or 7 turn win.  What really stops it is shields.  Also, a life based healing rush can overpower it depending on the draw.

It's best for T50 grinding (though it doesn't do too bad in PvP 2, surprisingly).

I think I'll call it the Crematorium.

Hover over cards for details, click for permalink

Deck import code : [Select]

4vj 4vj 4vj 6u1 6u1 71a 7ae 7ae 7di 7dp 7dp 7dp 7dp 7dp 7dp 7dq 7dq 7dq 7dq 7dq 7ds 7ds 7ds 7ds 7mt 7mt 7mt 7mt 7mt 7mt 8po

[omegareaper7]

Why damsels and not gem finders?

[Pwnator]

Why damsels and not gem finders?

2 ATK instead of 1.

Quite surprised the computer preferred the Ruby drag over the Crimson drag.

[Calindu]

Why damsels and not gem finders?

2 ATK instead of 1.

Quite surprised the computer preferred the Ruby drag over the Crimson drag.

In my oppinion it's way better to use only gemfinders instead of damseflyes and BoL's because you can grow your destroyers.

[Gorthos]

There's enough earth quanta to grow them for a few turns just based on nova and immo - I have no graboids.

I just did an informal TTW against AI3.  51 games, 1 loss, TTW = 5.56.  Much slower than in the simulation because of nice things like chained Dimensional Shields, Sundials, Pandemonium and other CC making some games last 8 or 10 turns.  The TTWs for the 50 wins were:

4 4 6 9 6 4 7 5 5 8 10 5 5 7 4 5 4 4 4 4 5 6 7 6 5 4 6 4 5 5 5 7 5 10 6 5 5 5 6 4 5 7 6 8 7 5 8 5 6

The deck isn't really optimized for AI3 which does do some significant CC.  If I wanted to optimize against AI3 I would probably add some unupped Phoenixes, maybe a Graboid or two, and 1-2 deflags.

Against AI3 earth quanta was even less of a problem because half the time the lava destroyers only stayed on the field for a turn or two.

[SteppingStone81]

This kind of optimization does not work as well when you restrict substitution to only 1 card at a time.  Also, did you allow the mark to vary?

[Gorthos]

I manually experimented with adding multiple cards or changing the mark, then let the computer optimize from there.

[Calindu]

I have a question,what simulator you use?And can you use mediafire to upload it and give it to us?

[Gorthos]

It requires a lot of programming and fine-tuning to a particular deck to work (giving the computer the correct strategy of how to play the different cards involved in that type of deck).

Earlier today I tried optimizing CP Wyrms, which is much simpler than an immo rush (fewer cards and fewer rules about how to play the cards).  You can look at my source code for the CP Wyrms optimization if you want to see how I did it:
http://pastebin.com/8PRcEsNA
A different type of deck would require a different program.  I did at one point try combining everything into a single program, but that ended up being more complicated than I'm willing to deal with as a hobby project like this.

(Incidentally, my simulation says CP Wyrms is half a turn slower than Crematorium)