Guest posts by SickPillow

25

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 16, 2010, 12:21:24 pm »

So... I was too fast in making changes to how i judge the winner.  The new system does not find the most effective deck at getting rares, and does not find the best deck for getting wins in the least amount of time.

The new system looks at score gain per turn.  I thought this value would take into account many factors, such as wins, speed and EMs (and it does this).  But the problem is that Elemental Masteries do not actually help your deck get rares.  Its basically just somewhat of a "bonus".  If you want the fastest speedgrinding deck for the purpose of getting rare cards, then this system of determining the winner fails.

I think I will have to go back to my original idea that i stated in my first post; that decks will be judged by their win rate as well as TTW (ie speed).  Score gain and electrum gain are actually just superfluous information, meant only as another indicator of overall performance.

Winners will be determined by dividing Wins by TTW for both decks, and comparing the results.  The deck with the higher score is the winner.



The whole thing is my mistake.  Changing the winner back and forth is going to be quite confusing to people. 

I will be editing the "new rules" again to reflect these changes.
I will also be adding how I will deal with "abnormal" games (outliers in the data)

AAnnd i will have to throw out the current matchup that I am working on, since the king deck will again be the old shrieker rush.   

26

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 15, 2010, 08:43:39 pm »

Also, I am curious, do you take into consideration the time of day/week you test? I know the T50 is a lot different on a Sunday morning versus a Thursday afternoon.

I try not to play any matches for KOTH on the day the scores get reset.  There are just too many changes happening on the T50 list right after the reset.  It is hard to get a good "snapshot" of all the decks.. and make sure that the two decks i'm testing actually face somewhat of the same competition.

Also i usually try to play all of the matches in one batch, switching back and forth between the decks as i go.  But sometimes I finish the match in 2 days.  Playing 230 games in a row while keeping stats is sometimes a bit extreme.

Basically i'm trying to maximize the chance that the two decks are facing the same decks. 

27

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 15, 2010, 05:10:33 pm »

King Of the Hill

This competition will pit T50 decks against eachother in a one-on-one matchup, to slowly find the best deck against T50 opponents.  Now, most people say: "You can't compare decks in T50! T50 opponents change every day!".  But I say, This is not True. 

For the competition I will play both decks against T50.  I will switch between decks every 10 games.  This way if any changes occur in the T50 decks, it will affect the performance of both decks.  I will play at least 100 games (unless one deck makes me very angry) for both decks, which should assure somewhat comparable statistics.  The win/loss ratio and turns-to-win stats will be compared and one "king of the hill" deck will be crowned.

Post your deck if you wish to challenge the leading KOTH deck.  Otherwise I will make my own pick.

*disclaimer* Please note that I am not playing 10,000 games with each deck, so it is possible that the worse deck can win a match-up.  Also, if i say how your deck "sux SoO hard", plz forgive me 

I quote your 1st post here.  You say that you are looking at 2 criteria to determine the best deck against T50 opponents: win/loss ratio and turns-to-win stats.  I'm pretty sure this part of the post has not changed since this study was started, if so forgive me.  So from the outset, you collected data on win% and ttw, which validates this claim.

Recently you 'changed the rules' to look at all parameters instead of only win%, but I thought, and your 1st post claims, that you were doing this all along.  The title of the study backs up this notion 'KOTH (slowly finding the best T50 deck)'.  Having 2 KOTH would serve little purpose as this study is to find the best deck versus T50.  This study is a novel idea as no one before has studied how well decks perform versus T50.  I hope you test many more decks versus the T50 in the future.

Ultimately where issues lie is that we all have a difference oppinion how to interpret the results.  What exaclty makes the best T50 deck?  That question has a different answer for many people.  It all depends on what the person is looking to do.  Most T50 farmers are looking to get rares as fast as possible.  Some people even skip non farms.  These people would want the fastest deck possible to get their rares efficiently; one that wins fast and also loses fast.  The next large group of people is those who shun the fg's and play T50 for electrum.  If there is a good number of farms out, the electrum for t50 can exceed that from the fg's while actually netting you positive score.  These people also want a fast deck, but would be worried about win % as well.  Before AI3 EM decks were in vogue, T50 was a decent place to farm score, however at the moment, score is more of an afterthough.

So rares, speed, electrum (after selling rares), and win % are the most important factors to the average joe T50 farmer in probably that order.

Looking solely at win% tells only a small portion of the picture when examining a decks performance, especially versus T50.  If you only have 1 data parameter available, then by all means use that for analysis.  When you have say 5 parameters to look at use all the data you have to make your case.

I am not attacking you or your study; I am rather trying to help you develop a an acurate and maningful scientific study.  Look at it as a thesis defence of sorts.  Keep up the good work.

hey jmdt.  i understand where you are coming from..   finding the best deck to play against T50 was my original goal.  That meant finding the most effective deck.  But as i played all the decks and tried to find the winner, i had a hard time nailing down my actual criteria of judging. 
i do like the idea of finding the most effective deck.  But effective can mean two different things.  Most effective in gathering rares, or most effective at winning.

i agree that most people will probably want the deck that is most effective at getting rares in the least amount of time.  This would be your Speed-Grinding deck.

but i also think there are people who are interested in finding the best deck at winning.  The most winning deck possible. 
For example, ME.  I don't play T50 to get rares.  I don't play T50 for score or electrum.  I play T50, because for me it is the most Fun. 
Truthfully i never want to be Grinding.  I don't even like the word.  I want to play T50 because you get the most variety of opponents.  you get fun new decks constantly being created by other players.  To me, T50 IS the endgame.  I like it more than playing FGs, which i do sometimes, and way more than AI3, which i find boring.

Having 2 "king" decks would allow me to search for both the best SpeedGrinding deck, as well as the best "Winning" deck (or "SlowGrinding" lol)
So far I have worked tons on finding the best SpeedGrinder... i don't think making a second "king" deck would detract that much from that study.

The two categories would be very simply labeled, with big arrows pointing to the SpeedGrinder as the deck for fastest "Grinding" of rares.
While the category for "most winningest" deck would be for people like me, who like to play T50 to challenge their deck and have fun.

28

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 15, 2010, 02:02:56 am »

New Rules  (For Deciding the Winner of a Matchup)

due to umm.. popular demand (or criticism), i will now determine the winner of a matchup by two statistics, points gained and total turns passed.

to determine a winner i will divide the total score gain by the total turns taken.
whichever deck gains the most score per turn wins.

the win/loss record will no longer figure into which deck wins the matchup.


This system will make it so that there will always be a winner, even when statistically there is no significant difference between the two data sets and the winner is basically random.
Also the system will favor fast decks and decks that get a lot of elemental masteries.

NOoo  Why, oh why...
Is there any need for this change. We already know that if you want to gain stuff-loads of score, you should play AI3 with The Ultimate Speed EM Deck (http://elementscommunity.org/forum/index.php/topic,10563.0.html) by jmdt (or whoever made that first). That deck can gain you some 2500-3000 score on 100 games, depending on play-style & luck. There's no way anyone can beat that playing T50, unless one faces only farm decks. We also know that best way to get tons of electrun is to use Hope/RoL on AI6. Gaining coins and score is just a byproduct of playing T50 in my eyes... The only thing that was unique in this study was that it compared how different decks WIN on Top50, and my opinion is that you should keep it that way.

This is the way I see the difference in Match-up 13:
Wins 83 vs 81. Difference 2 wins!
TTW 6,70 vs 6,67. Difference 3 turns in 100 games = 0,45%, no meaningful difference.
Elemental Masteries 13 vs 14. Who cares really?
Score 1802 vs 1850. Difference 48 points = 0,48/game, no meaningful difference. (And BOTH lose to AI3 grinding.)
Coins 2337 vs 2470. Difference 133 coins = 1,33/game, no meaningful difference. (And BOTH lose to AI6 grinding.)

My analysis of match 13 was the exact same as yours.  and i am a really big fan of judging by win percentage, and trying to look at the "whole picture" by examining all the data.  but really, i will get complaints about being biased for every match where the data is close.  So basically i have to choose some sort of system that will easily let me justify the winner.

If i said that i want to solely base my decision on win percentage, then someone would submit their "deckout" stall deck, which maybe wins 92% of the time but takes countless of turns.  Then i would get endless complaints about having such a "slow" deck as the "king" deck.

I sort of wish there were 2 categories, with 2 different King decks.  One is for finding a speed grinding deck with a decent percentage win rate.  And category 2 would be to find the best deck for Win%.
 
Trying to find the best deck for Win% is much more exciting because many awesome decks are slow.  They are slow, but fun to play, and may win at a really high rate.  Finding this deck is way harder than finding the best "speed-grinding" deck.

So maybe i should have 2 categories?  and stop the arguments?  And people could choose which "king" deck to challenge?

29

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 08:40:15 pm »

Great Work, Sick Pillow, I've been monitoring this thread since the beginning and I realize how much time and effort you spent on making this. +Karma

Thanks.  i appreciate it   

30

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 08:38:51 pm »

New Rules  (For Deciding the Winner of a Matchup)

due to umm.. popular demand (or criticism), i will now determine the winner of a matchup by two statistics, points gained and total turns passed.

to determine a winner i will divide the total score gain by the total turns taken.
whichever deck gains the most score per turn wins.

the win/loss record will no longer figure into which deck wins the matchup.


This system will make it so that there will always be a winner, even when statistically there is no significant difference between the two data sets and the winner is basically random.
Also the system will favor fast decks and decks that get a lot of elemental masteries.

I just changed the winner of the last matchup, due to the new judging rules. 
I'm hoping this change will stave off any further arguments over who is the winner of a matchup.
It will definitely make the whole thing easier for Me!   

31

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 06:38:13 pm »

Not so New Rules  (For Deciding the Winner of a Matchup)

The winning deck is the deck that wins the most matches in least amount of time.

Therefore the only stats which matter are Wins and TTW (TTW being a measure of average speed.  it would be better if i had actual time measurements).

To determine the winner:
The wins will be divided by TTW (i'd much rather do Wins/time, but I have not yet done time studies).  This will be done for both decks, and the results will be compared.  Whichever deck has a greater value will be crowned the winner.
This is sort of a strange win statistic, something like: Wins per average turns required to win.  I'd like to come up with a better system, but this is the best one I have so far.


I will also be looking at the data to see if there are any wins that took an abnormally large amount of turns.  These wins will be turned into losses if this helps to increase the win/TTW stat.

There is a reason why I am doing this and I will give an example: (this has happened)
One deck in the T50 is a deck that consists of only antimatter and liquid shadow.  I am testing 2 rush decks.  If I play any of my creature cards, I will lose the match.  The enemy deck will antimatter and liquid shadow any creature I put out.  Therefore the only way that I can win is by passing turns until the other deck runs out of cards.  The winning deck now gets a win with 23 turns passed.  One game of 23 TTW can severely alter otherwise great stats.  Basically the winner of this matchup would be decided by how many times a deck runs into this opponent.

So instead of a win I will give the deck a Loss instead.  This is the only change I make.  I do not want to punish a deck for my playstyle.  If I were Perfect Grinder I would immediately quit out of this matchup. (but instead i play it out and win)

32

Forum Archive / Re: Everything you want to know about the T50

« on: September 14, 2010, 05:47:16 pm »

I have a question too. Is the deck that appears the deck equipped when you submit your high score, or whatever you are currently using?

its whatever you are currently using.

33

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 03:53:39 pm »

The percentage is right.  I didn't alter the xx/xx win/loss value.  that is just a mistype, which does not affect the statistics, same math is shown in the zse thread 3 posts down.

there is no inconsistency in the data.

this was the change made:
80/100=80% -> 70/90=77,8% and
79/100=79% -> 65/86=75,6%.

now i changed it to show it before and after it was altered.  this is a bit silly, but it seems you want it.

Please note the added emphasis. I understand how it's easy to overlook something when you are editing a post, but dismissing it while conducting a study is still a mistake that needs to be acknowledged. I'm not trying to undermine the study. I'm trying to help you make it as true as possible.

i'm sorry but look at your previous posts, you do try to undermine the study. 
you claimed my "numbers where all over the place" 
you state that my math is wrong and that my "data is flawed".
you stated that faster ttw always leads to more points and electrum, and therefore my data is wrong  (this is not true)
you claim that i refuse to admit a mistake (which makes no sense, since i immediately admitted to having a mistype)

instead you could have easily done this way differently, by making a couple inquiries and having me fix the one mistype.
or asking why i would still choose a certain winner when it only won in win%
instead you attacked the study.  i definitely felt under attack   

34

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 03:19:04 pm »

if i had chosen skillgannon as winner, i would have also had complaints of people.  They would have quoted me as saying "But to me, in this comparison, I am going mainly by win percentage, and how strong a deck is."

that is exactly what i have previously said.  that i wasn't going to dethrone shrieker rush, when it put up a better win percentage.  and i'm still not going to do it.  sorry.

Skillgannon's deck has a higher electrum and score gain while also posting a lower ttw.  In the end who cares about win % when everyone that plays T50 is looking for the most electrum or score in a period of time. 

Skillgannon's deck clearly does this better based on your results.

truthfully at this point i don't care.  if everyone thinks that i should go by score and electrum instead of win percentage... that's fine by me. 
i was trying to stay consistent with what i was doing... by favoring win percentage.
but if i will have to argue about it every time... then i might as well make it score and electrum... and save myself grief.

35

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 03:05:20 pm »

I am not claiming anything that I haven't observed and carefully reviewed already, SickPillow. Look at the results of your win rate of that match.

Results:

I played 100 games for each side, 200 total. 

Stats!:
 Mono-Earth Shrieker Rush:
Win/Loss= 70/20, 77.8% (no farm decks)
TTW of 6.375
21 Elemental Masteries - 10 from farm decks
Points gained: 2062
Electrum gained: 2788
 Da Fastest Speedbow Eva!:
Win/Loss= 79/21, 75.6% (no farm decks)
TTW of 6.190
22 Elemental Masteries - 14 from farm decks
Points gained: 2058
Electrum gained: 2728

You played 100 games for each side, correct? So after removing the farm matches, the Shrieker deck had a net total of 90 games played. However, for your claim to be true (farm wins removed), the net games played for jmdt's deck should be 86. Looking at the results shows a win/loss record of 79/21, which is a total of 100 games. It is obvious that you made a mistake somewhere. My complaint is that you either didn't notice it or have refused to admit it. The inconsistencies in this data, by definition, mean that the data is flawed, no matter what the outcome.

The percentage is right.  I didn't alter the xx/xx win/loss value.  that is just a mistype, which does not affect the statistics, same math is shown in the zse thread 3 posts down.

there is no inconsistency in the data.

this was the change made:
80/100=80% -> 70/90=77,8% and
79/100=79% -> 65/86=75,6%.


now i changed it to show it before and after it was altered.  this is a bit silly, but it seems you want it.

36

Forum Archive / Re: KOTH (slowly finding the best T50 deck)

« on: September 14, 2010, 02:59:52 pm »

You can't throw out outliers in resulting data to justify a winner. The entire KOTH study is based on 100 games played by two different decks, with the raw results being paired up and the winner takes the hill. All top tier decks will have similar results against the top 50, so eliminating the possibility of "winning by a nose" makes close matchups an automatic win for the defending champ. That isn't fair. It's your study, but if you want to have accurate results, you must not declare winners on speculation.

is that really winning by a nose?

one deck wins 2 more games, the other gets 1 more EM, score and electrum are also virtually the same.

if i had chosen skillgannon as winner, i would have also had complaints of people.  They would have quoted me as saying "But to me, in this comparison, I am going mainly by win percentage, and how strong a deck is."

that is exactly what i have previously said.  that i wasn't going to dethrone shrieker rush, when it put up a better win percentage.  and i'm still not going to do it.  sorry.