Do you know what my favorite thing about Deckstats is? The stats.
So right now I'm working on trying to mess with EDHREC's algorithms. I'm making some weird decks, to mess with the stats of those commanders. Don't worry, they're not popular ones (
Tobias Andrion,
Gosta Dirk, and
Livonya Silone, in case you're wondering). I'm probably going to have to wait until EDHREC refreshes it's data (likely at midnight tonight).
In the mean time, I'll show you what I have so far.
In the previous post, I calculated the average lands of decks not using
sol ring, and found that it was 5.
So something is definitely wrong with EDHREC's data collection. But I replicated this for other cards to see what's up. I've also attached how I did these calculations to my post. The first equation is how to calculate the average number of lands across EDHREC, and the second is that same formula rearranged, to give us what the average number of lands is needed to skew the average to 29 lands.
So here's what I got for the top cards in each color.
If you don't like data, just read the parts in bold.
139 768 decks can use
Swords to Plowshares73 795 decks use
Swords to PlowsharesTherefore, 65 973 decks do not use
Swords to PlowsharesOn average, decks that use
Swords to Plowshares have 36 lands
In order for the EDHREC average land count to be 29, the decks without
Swords to Plowshares have, on average,
21.17 lands.
153 979 decks can use
Cyclonic Rift80 970 decks use
Cyclonic RiftTherefore, 73 009 decks do not use
Cyclonic RiftOn average, decks that use
Cyclonic Rift have 36 lands
In order for the EDHREC average land count to be 29, the decks without
Cyclonic Rift have, on average,
21.23 lands.
142 909 decks can use
Cultivate63 343 decks use
Cultivate Therefore, 79 566 decks do not use
CultivateOn average, decks that use
Cultivate have 37 lands
In order for the EDHREC average land count to be 29, the decks without
Cultivate have, on average,
23.43 lands.
154 357 decks can use
Demonic Tutor50 621 decks use
Demonic TutorTherefore, 103 736 decks do not use
Demonic TutorOn average, decks that use
Demonic Tutor have 37 lands
In order for the EDHREC average land count to be 29, the decks without
Demonic Tutor have, on average,
25.01 lands.
137 306 decks can use
Blasphemous Act33 611 decks use
Blasphemous ActTherefore, 103 695 decks do not use
Blasphemous ActOn average, decks that use
Blasphemous Act have 38 lands
In order for the EDHREC average land count to be 29, the decks without
Blasphemous Act have, on average,
26.08 lands.
Interesting. The more popular cards have a greater skew to the data. I went with the top card in each color, but what happens when I go to the bottom.60 896 decks can use
Assemble the Legion9 028 decks use
Assemble the LegionTherefore, 51 868 decks do not use
Assemble the LegionOn average, decks that use
Assemble the Legion have 36 lands
In order for the EDHREC average land count to be 29, the decks without
Assemble the Legion have, on average,
27.78 lands.
Boring data stuff is over. It seems pretty consistent that the average lands for a certain card is around 37, and the decks without this card are 25 or so.
But this shouldn't be happening. We should expect the average to be the same for decks using
Assemble the Legion as those that don't. Same with
Cyclonic Rift, or
Swords to Plowshares, or any card on EDHREC.
The lower half is skewing the upper half, and I think I know why.
If EDHREC takes incomplete decks into account, then of course the average is skewed. I've often seen decks that are just spells, no lands, because the deck is still a work in progress.
I won't know for sure until I figure out how EDHREC collects deck lists.