| 
  • If you are citizen of an European Union member nation, you may not use this service unless you are at least 16 years old.

  • You already know Dokkio is an AI-powered assistant to organize & manage your digital files & messages. Very soon, Dokkio will support Outlook as well as One Drive. Check it out today!

View
 

Epic Duels Analytics

Page history last edited by Ken Kutcel 6 years, 3 months ago

Here I explain how I use some basic statistics from decks to predict how good they will be before playing them. I will describe some of the things I have learned from the analysis, explain how it works, and provide you with an excel file so you can predict the strengths of your decks before ever playing them too.

 

Obviously, the predictor is not perfect (maybe sometime in the future I will try to improve it more). Its about 75% accurate in predicting the tier of decks, so not amazing, but certainly better than me. An index score of .5 is about average. Anything above .7 is strong, and a score above .9 predicts that your deck is overpowered (Obiwan's score is .90). In converse, any deck scoring less than .1 has very little chance to win against a top tier deck. The worst deck out of the 54 I have tested scored a -.08, and it was indeed the worst deck I have ever seen.

 

The more unorthodox the deck, the less accurate the predictor will be, because the deck won't match the trends which the math predicts. For example, a deck with 11 "Heal to Full Health" cards would clearly be overpowered, but would not score well, because the strength of the healing in each card is way more than what is typically see in the game (a.k.a. Luke's in trouble - recover 3 health).

 

Outline

1) 2 Graphs showing Some Decks Strength as According to the Correlation Index Formula

2) Excel File to Calculate the Value for Your Own Custom Decks with Explanation

          a) excel file with snapshots

          b) what numbers do I put in?

          c) example cards of special card type categories

          d) example run through with a deck

3) Data on Strength of Decks based on Deck Color, Types of Talent Cards

4) Explanation of How I Made This Formula

-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

1) 2 Graphs showing Some Decks Strength as According to the Correlation Index Formula

 

List of Decks By Strength

 

This below graph shows the predicted strength of each of the original decks based on my resulting formula, which uses the values of each talent card type in a deck. The higher the number the stronger the deck.  Just for fun, in this graph I include the originals plus some of the decks my group uses often from this site. I hope to be able to expand this list soon.

 

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

2) Excel File to Calculate the Value for Your Own Custom Decks with Explanation

 

a) excel file with snapshots

 

The attached Excel file allows you to calculate the Correlation Score for your own custom decks.

deck strength predictor.xlsx

The excel file looks like the following screen shot:

 

To use the file, insert a numerical value for category, as like in the following screen shot (example used is Geektopia's General Grevious). You can also access the excel copy of this file here: grevious predictor.xlsx. The Correlation Index Vale will calculate automatically. (*You may notice that this Grevious score below varies from the one I have above in the graph slightly; this is a result of rounding error from switching the formula from google sheets to excel.)

 

b) what numbers do I put in?

 

So how do I determine the numerical value to be put in each category for my deck? 

 

For 'ranged major', 'red', 'blue', 'green', 'yellow', 'yellow+', and 'other', put a "1" for yes, and a "0" for no. For example, Obiwan would put a "0" for 'ranged major', a "0" for 'red', and a "1" for 'blue'.

 

I think "Major Health', 'Health per Minor', 'Number of Minors', and 'Total Health' are fairly straightforward. For example, Obiwan would have values of "18", "4", "2", and "26". As you may have noticed, I count the the red square as a point of hp (so I add one to number right before the red dot).

 

The slightly more difficult, partially subjective part of this is entering the values for the number of talent cards in each special card type category: attack, defense, combat power up, movement, draw, not an action, action control, discard, heal, direct damage, major or minor, total count. Though I do not always record it for each deck, I also track: minor only, dual attack/defense, and miscellaneous. I will try to mention several times that most cards fit into more than one category. For example, Jango's 'Rocket Retreat' is both an attack and a movement card. 

 

Even if some of the effects may not be used or won't be used (example 7, example 8), the card is counted under each possible category. Any card which has a negative effect which only impacts allied characters counts as a negative (Anakin's Anger, example 10), but cards which do something negative to an ally as well as an enemy do not count against that character (Obiwan's Force Balance).

 

This is what I mean by each special card type category:

attack: any special card which is an attack

defense: any special card which is a defense

combat power up: a card which changes the attack or defense value of other cards.  I do not count cards

       which must determine their own value, such as Mace Windu's 'Battle Mind', in this category. I do not think

       this effect is present in any of the original decks, but it is a common feature I see in custom decks.

movement: any card which moves a character

draw: any card that allows an allied player to draw

not an action: any card that does not count as an action or allows future actions to not count as

     actions

action control: any card which limits what an opposing character can do

discard: any card which makes an enemy player discard cards

heal: any card which heals an allied character

direct damage: any card that yields direct damage

major or minor: a special card which can be played by either the major or minor character

total count: since most cards count in more than one category, this is the sum number of special abilities seen

     in the deck

minor only: a special which can only be played by the minor character

dual attack/defense: a special which can be used as either an attack or a defense

miscellaneous: a special ability which does not fit into any of the above categories

 

c) example cards of special card type categories

The most difficult, debatable examples are at the end.

 

Example Card 1: 'Tap Reserves' from P.D. Mangus's Nute Gunray Deck

This card would count as an attack, a heal, and a minor only. There are two copies of the card in the deck, so I would add 2 to the attack column, and 2 to the heal column, based on this card.

Example Card 2: 'Tidal Wave' from Dorkistan's Aang Deck

This card would count as an attack, defense, dual attack/defense, and a movement.

Example Card 3: 'War Tactician' from my 'Imrahil' Deck

This card would count as an attack, and a combat power up.

Example Card 4: 'Jedi Mind Trick' from Obiwan John's Rey

This card would count as a draw (a very good one!) and a 'not an action'.

Example Card 5: Mark of the Exalt from a Lucina Deck

This card would count as an action control, and a defense.

Example Card 6: Wrist Cable from Hasbro's Boba Deck

This card would count as direct damage and an action control.

Example Card 7: 'Utility Belt' from Darth Jake's Batman

This card would count as direct damage, a heal, and an action control.

Example Card 8: 'Poke Your Knees' from Paul's Old Luke Deck

This card would count as an attack, an action control, and a discard.

Example 9: 'Limb from Limb' from Geektopia's Tarful and Chewbacca Deck

This card would count as a major/minor (because either can play it), and direct damage.

Example 10: 'Anger' from Hasbro's Anakin 

This card would count as an attack, but would count as a negative 1 towards drawing (similarly, I would count damage towards one's own character as a negative 1 towards healing).

 

d) example run through with a deck

 

Here, I will use Geektopia's General Grevious as an example (deck creator, I hope this is ok -- If you would prefer me to change it to something else, please let me know). FYI- I use the version of this deck where Magna Guards can't defend after 'Cowardly Retreat'.

Ranged Major: 0

Red: 0

Blue: 1

Green: 0

Yellow: 0

Yellow +: 0

Other: 0

 

Major Health: 16

Health per Minor: 5

Number of Minors: 2

Total Health: 26

 

Attack: 5 - Jedi Hunter x 3, You Lose Jedi x 1, Unorthodox Attack x 1

Defense: 2 - Cowardly Defense x 2

Combat Power Up: 1 - Unorthodox Attack x 1. This card changes the value of the defense card. Jedi Hunter

     does not count as a combat power up since the card determines/changes its own value, not that of other

     cards.

Movement: 8 - General's Orders x 3, Cowardly Defense x 2, Twirling Saber Charge x 2, You Lose Jedi x 1

Draw: 3 - General's Orders x 3

Not an Action: 0

Action Control: 0

Discard: 0

Heal: 0

Direct Damage: 2 - Twirling Saber Charge x 2

Major or Minor: 0

Total Count: 21 - 5+2+1+8+3+2 = 21

 

Index Score: .76

 

 

---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

3) Data on Strength of Decks based on Deck Color, Types of Talent Cards

Deck Type

 

Here, I show the strength of correlation between each deck type (by color) and win percentage (the bigger the number, the better). I have also recorded ranged characters similarly. 

 

 

 

Special Cards Type

 

Here, I list the strength of each type of effect a special card can have. This assumes that the special's strength is relatively typical to what is seen in other decks (e.g. direct damage might be four hit points to a single character. In contrast, a card which dealt 10 hit points would atypical, and thus not well represented by these statistics.)

 

I include some of the data I used in the below attached excel file. If you are interested in how I came up with these numbers, please see my explanation below. 

Epic Duels Data.xlsx

 

-------------------------------------------------------------------------------------------------------------------------

4) Explanation of How I Made This Formula

 

So, I have taken a plethora of statistics based on the 50ish decks often used within my playing circle.

 

Here, I have attached an excel sheet which rates each of the original decks based on the types of specials in the decks. Categories included in the analysis are: ranged/nonranged, color of deck, number of attack specials, defense specials, combat power ups, movement cards, draw cards, cards which don't count as actions, action control, discard, healing, direct damage, miscellaneous, and finally, the sum total of all types of effects seen in the deck.

 

For each above statistic, I take the correlation (r^2) between win percentage and that category. For example, the strength of correlation between win percentage and number of movement specials is .31. The formula takes the strength of correlation between win percentage and each category, then multiplies that by a fraction: the number of a cards that deck has in that category, over the maximum number of cards in that category which has been seen in any deck. For example, concerning Jango's movement category:

 

Movement R^2 * (# Movement Specials in Jango Deck / Max # of Movement Specials in Any Deck)

.31 * (6 / 8*) = .2325  (*The value of '8' comes from Geektopia's General Grevious Deck)

 

Then the resulting values across all categories are added together to yield one score. The higher the score the better. 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Comments (3)

Ken Kutcel said

at 12:39 pm on Dec 28, 2017

I have absolutely no idea if anyone outside of my playgroup will care about this. If you use it or have any questions please let me know. Most likely, I did not write all of this clearly!

Roman F said

at 10:19 pm on Dec 28, 2017

Hi Ken, thanks for your efforts on this.

A long time ago, in a forum far, far away (Yahoo!), I had a similar concept of an excel workbook that scored every Epic Duels deck, including each special card in it. I did a lot of the same things you're suggesting, such as classifying and scoring each talent card. An A4 was worth 4 points, an A7 worth 7, drawing a card was worth 2 points, etc. The workbook was semi-useful for a while, but I found it not worth the time to keep it up, for the following reasons:
- There are a lot of curves. Attack value helps you to a point, but going beyond that point has a diminishing return, but it's hard to know where and by how much. I used to give a +2.5 bonus for defense cards because they're so valuable, but I found that they, too, have a point of diminishing return. It's definitely true of movement, as well. You want a deck that can do some of all those things, without doing too much of one at the expense of the others.
- There's such a plethora of special cards and their ideas, that it was just too much to keep up with, and it sort of should be that way. I love when people design new special cards that challenge what you can do in Epic Duels. I could no longer score and classify all of them.

Where I landed was this: It comes down to a deck's killing power and staying power, and everything else is to help one and/or the other. If you could somehow quantify those 2 things, and figure out the curves for both, you might be able to get somewhere. You'd have to factor in how things like card draws increase the base killing and staying power, and how forced discards are a form of increased killing power, and that sort of thing. So, you'd still have to classify and score each special, but I think it has to come down to those 2 concepts. I'm not sure how to deal with movement, but probably in some form it's a way to help you either kill or avoid being killed.

Ken Kutcel said

at 8:58 pm on Jan 4, 2018

Roman,

I think your critique of magnitude (e.g. a 'A7' counting differently than a 'A4') is very valid and is certainly a hindrance of my design as it stands. Also, it is true, as you state, that having 5 defense cards instead of 4 may not be as valuable as having 1 defense card instead of 0. My system doesn't currently account for this either.

However, as it stands, I do think that this system is better than arbitrarily giving values to certain categories. I do not create the value for each special card category. It is mathematically determined through the use of correlational statistics (e.g. R2 - I won't go into more detail on that here, but certainly a lot of good online information explaining it if on the very small chance you were interested in learning about it and did not already know it).
https://www.khanacademy.org/math/ap-statistics/bivariate-data-ap/assessing-fit-least-squares-regression/v/r-squared-or-coefficient-of-determination

The fix for your critique would be to change my formula so that is not operating linearly, but used curved fitted mathematical graphing functions. I know how I would do this, but honestly it would take a lot of time, and I do not really foresee myself doing it. Thanks though above all else, for taking the time to look at the article and commenting. Sincere thanks in all you have done for this forum.

You don't have permission to comment on this page.