Lightnews — Scholar-powered news

Magic Data Science

@mtgds.bsky.social

2.9K followers 90 following 1.8K posts

Doing math with Magic; doing Magic with math. (Chromatic Lantern; Julie Dillon)

Posts Media Videos Starter Packs

Pinned

Magic Data Science @mtgds.bsky.social · Jul 28

A thread that serves as a directory for articles I've written, links to things I've created, and recurring themes.

This is a good place to start if you're new to Magic Data Science.

2 1 12

Magic Data Science @mtgds.bsky.social · 1h

Hmm... Probably? If I was doing it again today, I'd probably do it by "minimum" rarity.

Magic Data Science @mtgds.bsky.social · 2h

That's just text matching. And honestly, that one is pretty low-priority, I'd say.

UB/UW is a big one; maybe percentage old-border-available? Maybe counting deviations from singleton? And/or highlighting extreme deviations? (100 Ornithopters).

Magic Data Science @mtgds.bsky.social · 3h

You didn't ask, but here is an old Twitter thread with lots of ideas for different things to count in comparing cubes:
x.com/mtg_ds/statu...

Magic Data Science on X: "Here are some details on the main event cubes themselves. They range from 160 to 540 cards in size, and there is huge variance in their composition. For example, this table shows the proportion of cards with each type. For example...👇 https://t.co/sSHPXZYxyI" / X

Here are some details on the main event cubes themselves. They range from 160 to 540 cards in size, and there is huge variance in their composition. For example, this table shows the proportion of cards with each type. For example...👇 https://t.co/sSHPXZYxyI

x.com

1 1

Magic Data Science @mtgds.bsky.social · 4h

This is cool! Your barplots might benefit from showing % in addition to, or instead of, count. Just because that makes it easier to directly compare two cubes of different sizes. Maybe on a second axis? Maybe just on hover? 🤷‍♂️

1 1

Magic Data Science @mtgds.bsky.social · 4h

Finally, in green, I think it's "supertrample"

What are your picks?

1 2

Magic Data Science @mtgds.bsky.social · 4h

In red, I think Ball Lightning might be the right answer, but for now I'm going with spellsling-to-deal-direct-damage. Red likes spells, direct damage, and pinging (repeatedly dealing damage). Guttersnipe and Thermo-Alchemist are variations on the theme.

1 3

Magic Data Science @mtgds.bsky.social · 4h

In black, I think it's the sacrifice-a-creature-to-lifedrain effect -- two very black things working together. The Blood Artist effect is probably my second pick -- it just doesn't itself directly require a cost.

1 2

Magic Data Science @mtgds.bsky.social · 4h

In white, I think Soul Warden's creaturefall/lifegain hits both white's preference for many small creatures, and its strength as the main lifegain color. Lifecreed Duo even adds flying!

In blue I'm leaning toward "curiosity" -- drawing cards when you deal damage to an opponent.

1 3

Magic Data Science @mtgds.bsky.social · 4h

As a follow-up to @andy.luckypaper.co's fun game from yesterday:

Which specific mechanic (not necessarily just keywords) is the "green-est"? And the same for white, blue, black, and red?

Doesn't have to be a primarily creature-based mechanic, but mine are. I'll put my thoughts 👇

2 3

Magic Data Science @mtgds.bsky.social · 4h

...though, that we don't know anything about the counterfactual -- what if they hadn't mulliganed the hand they drew? On average, mulliganed hands must be worse than typical, and so would have ~some~ lower win rate. But we can't really observe that.

1 3

Magic Data Science @mtgds.bsky.social · 4h

No, because it's not probabilistically derived from theory or anything, it's just based on @17lands.bsky.social data. Basically, if you have two otherwise identical players with two identical decks and opponents, the player who mulligans has a 14% lower expected win rate.

Worth noting....

Magic Data Science @mtgds.bsky.social · 5h

Probably the most surprising thing I've found, though, is that one mulligan, in most Limited formats, costs about 14% in win probability, holding everything else constant. That's huge!

1 1

Magic Data Science @mtgds.bsky.social · 5h

I think my favorite finding is that, when you model how much better cards (in Limited) get when put in increasingly synergistic decks, the gold signposts (#2CUDS) consistently stand out.

This is very pleasing to me.
mtgds.wordpress.com/2022/02/28/k...

1 1

Magic Data Science @mtgds.bsky.social · 5h

This is a very clever inversion of the trope

1 1

Magic Data Science @mtgds.bsky.social · 17h

Reach

Magic Data Science @mtgds.bsky.social · 20h

Yeah; came here to say that this idea is suggestive of pick draft, which... I think makes it more probable.

Magic Data Science @mtgds.bsky.social · 21h

Mostly baseless speculation:
I'm guessing this is more of a collation innovation than a printing innovation. But 🤷‍♂️

6 19

Magic Data Science @mtgds.bsky.social · 23h

This is the best kind of use of "social media" -- thought exercises!

I came up with a bunch of ideas; what are yours?

Andy Mangold @andy.luckypaper.co · 1d

Fun game @phantooom.bsky.social just invented: for each evergreen keyword ability, what is the card that uses it best? For example: Celestial Colonnade is the best use of vigilance, Ball Lightning is the best use of Trample, etc.