Recommend
 
 Thumb up
 Hide
11 Posts

KeyForge: Call of the Archons» Forums » General

Subject: Have any near duplicate decks shown up? rss

Your Tags: Add tags
Popular Tags: [View All]
Beau Bocephus Blasterfire
United States
Berrien Springs
Michigan
flag msg tools
badge
Avatar
mbmbmbmbmb
I understand the game generates random decks that are different and that there are a ton of different combinations, but being that it is random and all, I would think that over time some near duplicate decks should start appearing where there is only a card or two different between the decks. Just curious. I would consider them identical twin decks for the most part as even identical twins aren't perfectly identical.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ian Toltz
United States
Boston
Massachusetts
flag msg tools
designer
Your cat likes me more
badge
In brightest day / In blackest night / No evil shall escape my sight / Let those who worship evil's might / Beware my power--Green Lantern's light!
Avatar
mbmbmbmbmb
It's very likely that there are some similar decks.

But it's also incredibly difficult to actually find them, short of writing a program to scrape the API for all the decks and comb through it looking for similar ones (which is a non-trivial problem).
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Diane Mountford
United States
Minneapolis
Minnesota
flag msg tools
mbmbmbmbmb
It's also incredibly unlikely that very similar decks would show up in the same play group.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dobby
msg tools
mbmbmbmbmb
Asmor wrote:
It's very likely that there are some similar decks.

But it's also incredibly difficult to actually find them, short of writing a program to scrape the API for all the decks and comb through it looking for similar ones (which is a non-trivial problem).


I have a program that I've written that compares two decks and determines how similar they are. I also have all the decks downloaded. Just need to figure out how to do comparisons over large datasets lol
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Beau Bocephus Blasterfire
United States
Berrien Springs
Michigan
flag msg tools
badge
Avatar
mbmbmbmbmb
sdobby wrote:
Asmor wrote:
It's very likely that there are some similar decks.

But it's also incredibly difficult to actually find them, short of writing a program to scrape the API for all the decks and comb through it looking for similar ones (which is a non-trivial problem).


I have a program that I've written that compares two decks and determines how similar they are. I also have all the decks downloaded. Just need to figure out how to do comparisons over large datasets lol


Awesome. Please share your findings.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Ian Toltz
United States
Boston
Massachusetts
flag msg tools
designer
Your cat likes me more
badge
In brightest day / In blackest night / No evil shall escape my sight / Let those who worship evil's might / Beware my power--Green Lantern's light!
Avatar
mbmbmbmbmb
sdobby wrote:
Asmor wrote:
It's very likely that there are some similar decks.

But it's also incredibly difficult to actually find them, short of writing a program to scrape the API for all the decks and comb through it looking for similar ones (which is a non-trivial problem).


I have a program that I've written that compares two decks and determines how similar they are. I also have all the decks downloaded. Just need to figure out how to do comparisons over large datasets lol


Like I said, non-trivial. It's simple enough to go card-by-card comparing two decks, but that doesn't scale very well. I suspect there's something clever you could do to generate some sort of multi-dimensional value where you could then calculate the "distance" between two decks to come up with a similarity score.
3 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Beau Bocephus Blasterfire
United States
Berrien Springs
Michigan
flag msg tools
badge
Avatar
mbmbmbmbmb
Asmor wrote:
sdobby wrote:
Asmor wrote:
It's very likely that there are some similar decks.

But it's also incredibly difficult to actually find them, short of writing a program to scrape the API for all the decks and comb through it looking for similar ones (which is a non-trivial problem).


I have a program that I've written that compares two decks and determines how similar they are. I also have all the decks downloaded. Just need to figure out how to do comparisons over large datasets lol


Like I said, non-trivial. It's simple enough to go card-by-card comparing two decks, but that doesn't scale very well. I suspect there's something clever you could do to generate some sort of multi-dimensional value where you could then calculate the "distance" between two decks to come up with a similarity score.


I imagine it could be an interesting computer science project for some class.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Nushura
United States
Boston
flag msg tools
mbmbmbmbmb
bbblasterfire wrote:
I understand the game generates random decks that are different and that there are a ton of different combinations, but being that it is random and all, I would think that over time some near duplicate decks should start appearing where there is only a card or two different between the decks. Just curious. I would consider them identical twin decks for the most part as even identical twins aren't perfectly identical.


We do not know the exact algorithm used to generate the decks ....but they have said (see FFG's designer journal) that they made sure that decks were different enough. Whether "different enough" means 1, card, 3 cards or at least 10 cards I have no idea.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Kevin B. Smith
United States
Mercer Island
Washington
flag msg tools
designer
Avatar
mbmbmbmbmb
Asmor wrote:
Like I said, non-trivial. It's simple enough to go card-by-card comparing two decks, but that doesn't scale very well. I suspect there's something clever you could do to generate some sort of multi-dimensional value where you could then calculate the "distance" between two decks to come up with a similarity score.

As a start, I would go house by house. That way you're comparing 12 cards at a time instead of 36, and you would only have to compare decks with the same 3 houses. As a first pass, you could disregard any deck that has a maverick.

If I were writing it from scratch I might represent each possible card as a base64 character. Then compute the Hamming distance between the 2 house strings. https://en.wikipedia.org/wiki/Hamming_distance
4 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Dobby
msg tools
mbmbmbmbmb
peakhope wrote:
Asmor wrote:
Like I said, non-trivial. It's simple enough to go card-by-card comparing two decks, but that doesn't scale very well. I suspect there's something clever you could do to generate some sort of multi-dimensional value where you could then calculate the "distance" between two decks to come up with a similarity score.

As a start, I would go house by house. That way you're comparing 12 cards at a time instead of 36, and you would only have to compare decks with the same 3 houses. As a first pass, you could disregard any deck that has a maverick.

If I were writing it from scratch I might represent each possible card as a base64 character. Then compute the Hamming distance between the 2 house strings. https://en.wikipedia.org/wiki/Hamming_distance


I thought about doing something like that. I also looked into the Levenshtein Distance as well. I'm not sure it would give accurate results though, although I must admit I haven't looked into the math of it a ton.

Let's say I have a deck with 4 cards:
ABCD

And another with 4 cards:

ACDE

In my mind these two decks would be 75% in common, but would those distance calculations come up with that? Or would they say the distance is 3, and thus only 25% the same (the A being the same)?
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Kevin B. Smith
United States
Mercer Island
Washington
flag msg tools
designer
Avatar
mbmbmbmbmb
sdobby wrote:
Let's say I have a deck with 4 cards:
ABCD

And another with 4 cards:

ACDE

In my mind these two decks would be 75% in common, but would those distance calculations come up with that? Or would they say the distance is 3, and thus only 25% the same (the A being the same)?

It looks like that representation would be a list of cards in the deck. I think the comparison algorithms will have a lot easier time if the representation is of which possible cards are in the deck. In other words:

ACDEFFH
vs
10111201
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.