Magic Data Science

28 July 2025

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.

Breaking my long silence to report that +2 Mace now appears on Gatherer.

justingarydesign.substack.comMark Rosewater — Designing for Emotion, Embracing Complexity, and 28 Years of Iteration (#103)Think Like A Game Designer Podcast

9 May

I've been listening to @justingary.bsky.social's podcast for awhile now, and was pleasantly surprised to see that this week's episode featured a guest -- and a game -- that you may be familiar with*: justingarydesign.substack.com/p/mark-rosew... A must-listen if you're interested in game design.

9 May

* it's @maro254.bsky.social, and Mood Swings!

youtube.comHow to Play Mood SwingsYouTube video by Magic: The Gathering

30 Apr

This is so cool. I think maybe all (?) of the art is from sketches/drafts of illustrations that have appeared on Magic cards. Seems fun, easy to learn, charming. Way to stick with it, @maro254.bsky.social

maro254.bsky.social@maro254.bsky.social

Here’s the how-to-play video for Mood Swings. #WotCStaff youtube.com/watch?v=-uf-...

30 Apr

30 Apr

22 Feb

📯I am pleased to announce the winner of the Lorwyn Eclipsed #P0P1 Challenge: 🎇 @formersquid 🎆 Going deep on BG, and hitting on the second-best Eclipsed creature (Kithkin), proved to be the keys to victory!

In second place, with 457.91 points, was Inflamed (who came in 3rd for TLA!), closely followed by @veserius.bsky.social at 457.87 -- and then several others close behind! Here is how the rest of the competitors finished out:

22 Feb

I finished somewhere at the bottom of the top third of competitors. I'm not satisfied with that result, and rest assured, changes will be made!

@formersquid, with 458.6 points, can now proudly display this trophy text: 𝐖𝐢𝐧𝐧𝐞𝐫 𝐨𝐟 𝐭𝐡𝐞 𝐄𝐂𝐋 #𝐏𝟎𝐏𝟏 𝐂𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐞🏆 No matter what happens, you will always be the winner of the ECL P0P1 Challenge, and no one can take that away from you.

15 Feb

#calledshots for the future The Hobbit set: RW draft archetype will be Dwarf Typal, or at least, heavily Dwarf-centric, maybe with a connection to treasure (tho that's more likely RG) The return of common Legends, at least to fit in some of the many lesser Dwarves

10 Feb

~ ≠ ≈ I'm open to using the ~ if you're texting or in a hurry, but on an earnings report, I'd probably take the time to figure out how to write the ≈. Also, some wild keming choices here

Magic Data Science reposted

10 Feb

Here is where ECL lands on the #PrinceVPauper ternary plot: decidedly _not_ Pauper. The rares and uncommons are relatively stronger than the commons, to a greater extent than pretty much any recent set. At common, look for good removal, friendly creature types, and... good luck!

Magic Data Science@mtgds.bsky.social

Here's a new visualization I'm trying out -- representing the #PrinceVPauper distribution of recent Limited formats. The axes are unitless, but essentially represent the relative win rates of cards, by rarity. The left-to-right dimension maps onto the common ↔ rare...

10 Feb

Magic Data Science@mtgds.bsky.social

Magic Data Science reposted

Some time ago, I came up with a method of measuring the "openness" of a card in Limited, which can summarize up to describing a particular format. One of the more interesting things I found was how relatively "on-rails" Bloomburrow was, with its 10 typal archetypes. You can...

@bsweitz123.bsky.social

5) The least Open cards -- the ones that are most narrowly drafted/played -- are mostly finely-targeted Typal synergy cards, many of which explicitly say the name of a creature type. This is very Bloomburresque.

Read the whole article if you want to know more about what I'm doing here, and if you want to see Openness scores for *each common* in the 17Lands era. mtgds.wordpress.com/2024/08/19/r... And let me know what kinds of patterns you observe.

3) Contrast those with the Typal hybrid "signposts", which are drafted and played much more like gold cards than hybrid. In the graph ^^^^, you can see the distinction between these two groups of hybrid commons -- the gluey Changelings are way at the top of the distribution.

4) The most "Open" mono-color cards are, as usual, the strong removal cards, which are (a) desirable and (b) not usually synergy-reliant

...as the graph shows, it also has the wide/low distribution characteristic of a fairly on-rails format. Here are some observations. 1) The most Open cards are, as usual, the colorless lands and artifacts, although even within these cards there are some archetype leanings.

2) The gluey hybrid Changelings are, as you might expect, pretty high-Openness. The "theoretical maximum" for these cards is a 7, which would be the 7 color-touching pairs playing the card equally frequently, and several of these are close, at 6+

... see this depicted in the graph as the mass of dots in the 1.5-3 range. You can read about my methodology, and how I interpret these Openness values, here: mtgds.wordpress.com/2024/08/19/r... Naturally, I was curious as to how ECL would look, given that it is also heavily typal. Well...

careers.hasbro.comGame Designer | HasbroKeep up to date on card files, build decks, and test interactions between cards. Give feedback to Design leadership on playtests, card designs, interactions, and impressions. Solve specific design pro...

4 Feb

Someday they'll post a job that requires 2+ years of graph-making and Limited-format-predicting experience.

(S)Cary 🎃@acry.bsky.social

There's a Magic Game Designer position up. Requires 2+ years competitive experience, not remote, exactly what it says on the tin: careers.hasbro.com/careers/job/...

2 Feb

@cubecobra.com Excited about the 1.5 update! Especially Scryfall drag'n'drop. Does this update mean that the job that calculates Cube Elo will run regularly? Also, at the moment, it seems like Playtest Data is missing?

30 Jan

Magic Data Science@mtgds.bsky.social

29 Jan

Clearly Alara, and specifically, Bant

maro254.bsky.social@maro254.bsky.social

Someone on my blog asked what plane has the best pizza. One of the planes I chose was Ravnica, so another person asked which guild made the best pizza. It was “Another Day of R&D Productivity Lost” (for fans of my old “Tales from the Pit” comic). #WotCStaff

Magic Data Science reposted

Here is a plot of GIHWR by rank within color, for commons and uncommons in ECL, per @17lands.bsky.social Among commons, it's G and then everyone else, although R and B have a higher top-end before falling off, while U and W are a little flatter. Among uncommons, ...

...the lines are pretty well tangled up, which is often a sign of good color balance, though there is a slight edge to green. Just going off of this graph, I'd say -- look for Elves, but none of the 5 primary typal mechanics are guaranteed to fail, if built reasonably.

I'm one of these, @maro254.bsky.social Watched 20 years, 20 lessons Listened to the DtW podcast episodes expanding on that Listened to every DtW at least once Had a million ideas for models to build Spent a year working on those intermittently Started MTGDS.

Good! Magic needs more math.

Matt Tabak@wotcmatt.bsky.social

I made this card, so if you think exponents are scaring me off, you’re not even close. (Zero, however, is even.) #WotCstaff

Good! Magic needs more math.

Chris Mooney@chriskmooney.bsky.social

I am fueled by all reactions to Mathemagics.

27 Jan

They made a card just for me! 🥰 🫶 WotC

27 Jan

Magic Data Science reposted

24 Jan

With all of the Treefolk and -1/-1 counters running around ECL, folks have been talking about how high-toughness the creatures in this set are. But, are they? 🤔 I went to make this graph, expecting ECL to be an obvious outlier, and it's... just not! As you can see... 👇

This holds up if you look at all cards, not just commons & uncommons. I just thought this was interesting, and a little counterintuitive. Especially after you add all of the -1/-1 counters, ECL is surprisingly low-toughness! (It's also interesting to see just how big...

24 Jan

... the creatures were in FIN. FIN > MOM > EMN > IKO are the four sets with the highest mean P+T.)

...all but 7 of these sets has higher average toughness than power, but the biggest toughness outliers are all old sets: Alpha, Alliances, & Antiquities. But even in the last year, Final Fantasy and The Last Airbender are substantially more tough-leaning than Lorwyn Eclipsed.

23 Jan

What does it say about a format if this is the highest GIHWR blue common?

23 Jan

I guess you could say that I did not expect this.

Magic Data Science reposted

22 Jan

Several folks on @maro254.bsky.social's Blogatog reported that players at prerelease were confused by the official 5-pair archetype guide, hesitating to explore the other five pairs, despite some support for them. So I made an unofficial guide to those secondary archetypes!

Magic Data Science@mtgds.bsky.social

ECL is another 5-archetype reference, but this one has an unusual twist: It includes three "allied" and two "enemy" pairs! This structure is mirrored in the #2CUDS as well as a common hybrid cycle, half of which are explicitly typal, half of which are Changelings.

22 Jan

Magic Data Science@mtgds.bsky.social

22 Jan

I know this would have been more useful about a week ago, but 🤷‍♂️ Maybe this will help someone, and if not -- I love these Archetype Guides, and wanted to dabble in the genre anyway 😄