Testing Stall cards

[ratcharmer]

I've been testing this deck (http://elementscommunity.org/forum/index.php/topic,10503.0.html) and I've gotten a 177 | 23 ratio (88.50%). I know it was ttw testing and not your stall testing but I just wanted you to know I've already had experience with this stuff. Which deck would you prefer I use, if any?

Testing with the death deck would be great basically any deck from Jmdt's study would be a good one to pick, since the main things I'm looking for are 1) well known deck and 2) rush deck.

And THANK YOU for offering help

[Baily18]

Testing Grabbix

Code: [Select]

4vh 4vj 4vj 4vj 4vj 52q 590 590 590 590 590 5c1 5f9 5f9 5f9 5f9 5f9 5f9 5fc 5fc 5fc 5fc 5fc 5fc 5l9 5l9 5l9 5l9 5l9 5l9Unopposed TTW of 5.75
Turns Taken for Wins
4-1
5-7
6-8
7-4
Vs Druidic Staff:6.45/5.75: Stall Rating 1.122
Turns Taken for Wins
5-3
6-7
7-8
8-2
Vs Discord: 6.55/5.75: Stall Rating 1.139
Turns Taken for Wins
5-5
6-6
7-6
8-1
9-1
11-1
When you have no immos but phoenixes in hand, discord can slow you down alot. When you have immos and phoenixes for it, it can help you get that extra quantum you need to reborn that phoenix or grow that spirit, which made more 5 and 6 turn wins.
Vs Owl's Eye: 8.25/5.75: Stall Rating 1.435
Turns Taken for Wins
5-1
6-3
7-4
8-7
9-4
10-1
12-1
All of Grabbix's creatures are under three health(besides spirit and lycan buffed), and the reborn of phoenix prevents the summon of another phoenix.(without nova)Also, since the OE is likely to come out first turn your shriekers will not have a chance to burrow unless you evolve two at once.
Vs Vampire Dagger(Upgraded): 7.05/5.75: Stall Rating 1.226
Turns Taken for Wins
5-1
6-7
7-4
8-6
9-2
Vs Eternity: 11.05/5.75: Stall Rating 1.922
Turns Taken for Wins
7-2
8-2
9-5
10-4
11-2
14-1
15-1
17-2
21-1
Bad hand=Eternity slaughters you. Arsenic makes for a faster win and you need to play your whole hand in one turn to win faster.
Vs Vampire Dagger(Unupgraded): 6.4/5.75: Stall Rating 1.113
Turns Taken for Wins
5-3
6-9
7-6
8-1
9-1
Vs Shield(Unupgraded): 6.55/5.75: Stall Rating 1.139
Turns Taken for Wins
5-3
6-6
7-8
8-3
Vs Turtle Shield: 8.7/5.75: Stall Rating 1.513
Turns Taken for Wins
6-2
8-6
9-6
10-6
Got several bad draws.
Vs Diamond Shield: 8.75/5.75: Stall Rating 1.522
Turns Taken for Wins
7-2
8-7
9-6
10-4
11-1
Vs Improved Fog: 7.75/5.75: Stall Rating 1.348
Turns Taken For Wins
6-3
7-4
8-8
9-4
10-1
Vs Tower Shield: 7.5/5.75: Stall Rating 1.304
Turns Taken For Wins
5-1
7-11
8-5
9-2
10-1
Vs Devourer: 6.05/5.75: Stall Rating 1.05
Turns Taken for Wins
5-5
6-10
7-4
8-1
Devourers didn't really slow down the deck at all. Maybe stop a lycan from growing, or ash from reviving once in a while.
Vs Sanctuary: 7.2/5.75: Stall Rating 1.252
Turns Taken for Wins
5-1
6-4
7-9
8-4
11-1
9-1
Will edit as i gather stats.

[Pineapple]

6 SoGs is over 1.5!?
And it's only 6 cards that cost one raw non-elemental quanta each?

[ratcharmer]

6 SoGs is over 1.5!?
And it's only 6 cards that cost one raw non-elemental quanta each?

Well SoG costs more than 1 quanta to play, but yes it is performing extremely well so far.

Evaluating card balance was one of my goals in starting up the study.

That being said, many of you are probably noting that often times the ttw against these decks with only one real card are not very good compared to AI3. If this confuses anyone just think about how often AI3 drops a discord or phase shield on the first turn.

Thanks Baily18!

[silux]

I'm quite sceptical about empirical values since you can get as many experiments you want but the limit of probabilities approaches  (infinity).
So due we don't have endless time you can get sure datas from maths.However maths provides sure headache.

*every case of probability happens in 5(6x5 deck)x 5(6x5 ai)=25 cases
(it still doesn't mean that after 25 games you won't have a sure review of the deck yet, because often cases happen twice during testing.But if you run a test for every combination(outside elements) you will be sure about how the deck behaves)
If you want to test easier the deck give the ai deck with limited quanta and many copies like this

Code: [Select]

4vj 4vj 4vj 4vj 4vj 4vj 4vl 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vj 4vl 4vl 4vl
USEM by jmdt is quite straightforward so (2/30 10/30 6/30 6/30 6/30)
ai 3 currently has a rough medium of 40 cards, almost 8kind of cards and 13decks
(6+6+1+10+2_30)(8_40)(13)=520cases
error esteem (2/3+2/3)(8(1/40))(13)=52/15=<3.5
So 524 tests

I'm going to study another little maths...

[Baily18]

Im assuming i don't have to test Trident as Grabbix is pillarless. Man, Dimshield is going to take forever.

[Bhlewos]

Began testing unupped Mono-Death just now:

by Bhlewos

Hover over cards for details, click for permalink

Deck import code : [Select]

52g 52g 52g 52g 52g 52g 52g 52h 52h 52h 52h 52o 52o 52o 52o 52o 52o 52t 52t 52t 52t 52t 52t 542 542 542 542 542 542 542 8pk

Unopposed TTW of 7.45:
Turns taken for wins:
4 - 0
5 - 0
6 - 0
7 - 12
8 - 7
9 - 1
10 - 0

Notes: 9 turn win was due to a really bad draw (only 1 pillar first turn).

I'll start testing the weapons later today.

[ratcharmer]

You guys ROCK. I can barely keep up with adding all the new data to the OP.

@Bailey18:
Yes, of course feel free to skip any that you don't think you can get meaningful data out of. And, yeah, I've been putting off testing Dim shield myself. I'm seeing that one get a stall rating of at least 3.

@silux:
I'm not sure what you're trying to say in a lot of your post, but I'll respond as best I can. Sorry in advance if I misinterpret anything.
I am giving the AI decks with just 6x of the card being tested and the rest quanta, but I'm not sure what the advantage of using nova instead of towers could be.
I don't know why you're referring to AI3, since AI3 is not involved anywhere in this study.
Yes, there is a huge number of different orders you could draw the cards from a given deck, but almost all of them given very similar ttw values, to the deviation isn't nearly as scary as your math predicts.

[Bhlewos]

Just tested with upped discord.

by Bhlewos

Hover over cards for details, click for permalink

Deck import code : [Select]

6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6ts 6u5 6u5 6u5 6u5 6u5 6u5 8pj

Total games played:

ttw unopposed = 7.45

ttw opposed = 8.70
4 - 0
5 - 0
6 - 0
7 - 0
8 - 7
9 - 12
10 - 1

Stall rating: 8.70 / 7.45 = 1.17

Notes: Discord actually did quite a lot of damage by itself, usually I had 40 - 50 HP left at the end. Usually it was also played the first turn. Otherwise, not too much trouble.

And no prob ratcharmer, glad to help. =]

[silux]

I was going to say that 200 test aren't required.
(However 200test are little when testing an ai3 grinder(260minimum))

(2/30 10/30 6/30 6/30 6/30)USEM
(15/30 15/30)test deck

15x2=30cases
A medium of 15 tests is enough when testing weapons.

You should give some specifications about the cards
example
Poseidon on turn 1

Artic squid on turn 3

stall cards have different effectiveness also depending of turns.

[Bhlewos]

Results with Unupped Druidic Staff.

Code: [Select]

5c5 5c5 5c5 5c5 5c5 5c5 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7ac 7bu
Total games played: 20

ttw unopposed = 7.45

ttw opposed = 8.35
4 - 0
5 - 0
6 - 1
7 - 1
8 - 10
9 - 5
10 - 1
11 - 0
12 - 1

Stall rating: 1.12

Notes: Actually got one EM, in the 6 turn win. Throughout the game the AI did not play the Druidic staff AT ALL, lol.
The 12-turn win was unlucky, I did not get out creatures until the 4th turn.

[ratcharmer]

Hi, I'm back to updating this again.

Sorry I fell behind for a bit there. I got stuck trying to balance the quanta use for fate egg.