AI5 turns to win, lets talk about it.

[jmdt]

So, there has been lots of chatter about an AI5 version of a ttw study.

The ultimate questions are 1.) What are the best decks to look at and 2.) exactly how should testing go.

It seems peolple use everything from pure rushes to fg decks versus AI5.  There is no solid consensus what is best.  Also, seeing how there are 144 different AI5's, how many games gives a good representation of AI5 performance?

Discuss.

[QuantumT]

If I had to guess, I'd expect that the best ttw will be achieved by a rush with a minimal amount of control added. I'll submit something in a bit.

As for necessary number of games, 100 should still be plenty to give you the feel.

I submit something along these lines:

Code: [Select]

6rk 778 778 778 778 778 778 778 778 778 778 778 778 778 77g 77g 77g 77g 77g 77g 77h 77h 77h 77h 77h 77h 77l 77l 77l 77l

[willng3]

Yeah, I skimmed through Chat History a bit after I saw this thread.  100 games at least gives us an estimate of how fast the deck performs in comparison to another, thus giving a rough estimate of which is the fastest/most efficient.  For actual statistical analysis, 500+ games is an absolute must, period.  Of course I don't know of many people besides myself who have the time and patience to sit down and do this many games against such a tedious opponent.  FG grinder analysis is done because under normal circumstances it pays off through upped cards and such.  That same advantage exists here, just much less frequently, which leads me to believe that the biggest problem here is going to be community participation.  Under normal circumstances I am opposed to having stats for one deck come from multiple participants, but honestly here we might not have any alternative but to look past this rule of thumb.  AI5 definitely needs to be inspected more in depth, so yes, this is a good idea.

[RavingRabbid]

I might say that, for AI5, time to win is more important than turns to win.

[willng3]

I might say that, for AI5, time to win is more important than turns to win.

Time to win is directly related to turns to win.  Since people tend to...doze off, eat, get distracted by the TV, or think too much during a duel, Turns to Win is a much better way of giving estimates of speed for a deck.  Under normal circumstances you can safely say that the faster the deck can win, the less time it will take you to win a duel.

[jmdt]

I might say that, for AI5, time to win is more important than turns to win.

Time to win is directly related to turns to win.  Since people tend to...doze off, eat, get distracted by the TV, or think too much during a duel, Turns to Win is a much better way of giving estimates of speed for a deck.  Under normal circumstances you can safely say that the faster the deck can win, the less time it will take you to win a duel.

Thats both true and not.  When two decks have the same ttw, the win time becomes very important.  Win time is the stat that a grinder will look at most closely. i.e. what deck gets me to upgraded cards the quickest.

Speaking of upped cards, keeping a tally of upped cards won is an absolute must to the success of this study.  With any luck we can determine the average games to upped card, which will be huge in the viability of grinding AI5.

[QuantumT]

I take back the shrieker deck. The win percentage is too low. It just doesn't have enough control.

As for cards won, can't that be calculated?

[thatnewguy]

So for this study we must include spins also? Won't that throw off the electrum gained per game/ total? I guess we need to start making a tally for the +5 in spins for AI5 when it comes 'round.

[teffy]

Rainbow decks can have Hourglasses, time to win can be as important as ttw.
Expect >5 minutes per game.

Edit: Wow, Shrieker Rush better than I expected.

[zse]

I might say that, for AI5, time to win is more important than turns to win.

Time to win is directly related to turns to win.  Since people tend to...doze off, eat, get distracted by the TV, or think too much during a duel, Turns to Win is a much better way of giving estimates of speed for a deck.  Under normal circumstances you can safely say that the faster the deck can win, the less time it will take you to win a duel.

That might be true for "easy decks", but compare playing 1 turn with 'mutation with 23 creatures doing all sorts of tricks and stuff' vs. merely using Eternity lock to stall AI.

[jmdt]

So for this study we must include spins also? Won't that throw off the electrum gained per game/ total? I guess we need to start making a tally for the +5 in spins for AI5 when it comes 'round.

No, just note somewhere how many upped cards you winalong the course of 100 games.  The fact that you can win upped cards from AI5 while potentially using a deck good for AI3/T50 can completely change the way people play if we find it lucrative and possible.  People will not be bound to play the fg and can use the deck they make for AI3 to play AI5 as well.

All this assumes we find a rush that is actually good versus AI5 which may or may not happen.

[RootRanger]

I prefer turns to win over time to win. Turns is a solid, universal value. With one draw and one hand, every good player does the same thing in one turn. However, one person can play through a deck faster than others due to click and computer speed. I don't want personal clicking speed to interfere. What I'm saying is that two people will win in the same amount of turns in a certain situation, but two people will not win in the same amount of time in a certain situation.