Graphical analysis of the post deathrite meta

[janchu88]

hey,

i wanted to get a little bit in touch with some graph visualization and decided id work on some mtg related stuff, in short i did the following:

- got 2750 decklists of legacy winning decks in the period between 15Jul2018 and 07Feb2019
- calculated a "synergy" value between all of those cards
- calculated a gradient of how each card saw its meta share increase or decline since the first results after the ban up to today. So if on day 1 (after ban) some new cards or decks were tried but it didnt work out longterm the gradient is negative, if some card increased its share or made it newly into the format the gradient is positive. If the card wasnt really affected at all longterm then the gradient is close to 0
- caluclated the individual meta share of each card
- and finally created a graph out of it. Imagine it like a simulation where each card is represented by a bubble and those bubbles have springs between them according to their relationship to each other. The size of the bubble stands for the meta share and the color indicates if there is an down, or uptrend for this specific card.

Why all this? As David Kriesel (some Data analyst) said: the eyes are the broadband connection to the brain! :)

I have explored it to some degree and made found some interesting details. But im gonna leave the interpretation to you guys and hope this can lead to some discussion afterall

Download as SVG File (8mb): https://drive.google.com/open?id=19JT1T7Tr4XomJKtv1PRbaB737hVla_NT
Download as PNG file (25000px x 25000px 50mb): https://drive.google.com/open?id=1tD84MeKHLTp7PvV1lSs_CyGX-uBO4B9p

And a preview what it looks like:

https://abload.de/img/previewl2kgy.png (http://abload.de/image.php?img=previewl2kgy.png)

If there are any questions feel free to ask!

[FourDogsinaHorseSuit]

Now I clearly understand this simple and in no way hard to read graph, but for those other, less sophisticated, posters what is the take away here?

[tescrin]

* That basic island sees more play than Volcanic Island (and the majority of the other cards in the format)
* That basic Plains, swamp, and mountain each see more play than most cards in the format

Thus, basic lands are broken. I think it's time we acknowledge that and ban basics.

But seeing as wotc likes to IGNORE DATA I think I'm gonna buy up a bunch of basic lands before they start spiking!

[apple713]

* That basic island sees more play than Volcanic Island (and the majority of the other cards in the format)
* That basic Plains, swamp, and mountain each see more play than most cards in the format

Thus, basic lands are broken. I think it's time we acknowledge that and ban basics.

But seeing as wotc likes to IGNORE DATA I think I'm gonna buy up a bunch of basic lands before they start spiking!

guru island is already through the roof bro

[janchu88]

Now I clearly understand this simple and in no way hard to read graph, but for those other, less sophisticated, posters what is the take away here?

that is the thing, it depends. In the end there are a lot of thing to be seen here driven by data rather than gut feelings, just to metion a few:

- i think the overall meta hasnt finally settled yet, though its shape roughly will stay the same
- If you were a long term Bug player like me: Bad news, without DRS bug is officially dead
- you can clearly see blue destroying heavily distorting the color pie. We already knew that, but here it is displayed
- You can clearly see people tried goblins after the ban but longterm it still could not stick even without DRS
- DnT seems to suffer from the Ban overvall
- the data indicates that lands suffered from the ban overall, maybe that can be explained by the heavy uptick in combo based strategies? knight of the reliquary based strategies seem to have improved though and fight for the spot
- plus tons of other things you can see

but the interpretation of what you see is up to you. You have to ask yourself questions for each individual thing you can see.

Something i found to be very interesting btw. I made my first tests of the proccess based on some older data when DRS was still legal in 2017.... It indicates people where right, DRS pushed UBx into being the core of the format :laugh:

https://abload.de/img/2017agjyj.png (http://abload.de/image.php?img=2017agjyj.png)

[sco0ter]

Chalice of the Void being directly next to Aether Vial is weird. It indicates they are played in the same deck, no?
Same for Mountain + Lord of Atlantis...

Would be great if we'd see the cards being arranged better.

[janchu88]

It does not indicate they are in the same deck, but that they share a common playstyle and therefore push into the same spot in the meta

i think i would explain it like this: Merfolk ist trying to be the bridge between parts of the blue core like FoW and stompy elements like i.e. Chalice. But on the other hand it is not really connected to i.e. brainstorm and ponder which has an insane influence and pushes it out of that spot. And there it happens to be pushed towards the mountain and red stompy corner. The coloring indicates that it is drifting further away from the core and, whereas the red stompy decks gain influence overall.

I would sum this up as:

- Even though being a mono blue deck it doesnt really do the blueish stuff. Basically this indicates it shares more elements with like an aggro (red) stompy deck than the classical cantrip mafia, which is true if you think about it
- It is one of the worse stompy decks and therefore loses ground to the better stompy decks pushing into this spot

edit:

after taking a look at the bigger picture you could even define the succesful archetypes that define the current metagame

https://s16.directupload.net/images/190208/temp/2ws9db3w.png (https://www.directupload.net/file/d/5357/2ws9db3w_png.htm)

and all other strategies dont really seem to find their spot in this meta. The random surroundings are decks we are seeing and occasionally are doing well, thats why the gradients are very high towards the outer side but they cant stick and those decks dont really define the format unless you wat to categorize them as random garbage as a whole :D.

PS: Loam / Dark depths decks should be the supertitle.

[Bithlord]

Thus, basic lands are broken. I think it's time we acknowledge that and ban basics.


Poor forest. Sitting in the corner eating his paste and wondering what he ever did...

[Humphrey]

this looks more like a modern chart or some kind of legacy and modern mix. tons of cards on it that never see play in legacy

[janchu88]

this looks more like a modern chart or some kind of legacy and modern mix. tons of cards on it that never see play in legacy

Partially this can be explained by the scaling / resolution issue. The bubbles for the major cards would need to be even way bigger in order to make the subtle differences in meta share more visible for the not so heavily represented cards. The reson behind this is that all meta share values are normalized to a scale between 1 and 0 and the bubble representation scales from 1 to 10. It is kind off a tradeoff, cause otherwise the map had to be even bigger and the resolution should have been even higher in order to make the small areas readable. But nevertheless people tried those cards in legacy with limited success. What you see there is the constant attempts of porting modern into legacy. This "random" cloud around the core is the garbage category i meant earlier, which cannot get real traction in legacy but sees some random top8´s from time to time. The chart is 100% based on data from legacy events.

But its a good feedback, this can be solved to some degree by making the bubble size not perfectly linear in order to emphasize the differences in this lower areas better

[Kap'n Cook]

Partially this can be explained by the scaling / resolution issue. The bubbles for the major cards would need to be even way bigger in order to make the subtle differences in meta share more visible for the not so heavily represented cards. The reson behind this is that all meta share values are normalized to a scale between 1 and 0 and the bubble representation scales from 1 to 10. It is kind off a tradeoff, cause otherwise the map had to be even bigger and the resolution should have been even higher in order to make the small areas readable. But nevertheless people tried those cards in legacy with limited success. What you see there is the constant attempts of porting modern into legacy. This "random" cloud around the core is the garbage category i meant earlier, which cannot get real traction in legacy but sees some random top8´s from time to time. The chart is 100% based on data from legacy events.

But its a good feedback, this can be solved to some degree by making the bubble size not perfectly linear in order to emphasize the differences in this lower areas better



I kinda think you shouldn't scale down at all for the picture's sake. Let us see the Jupiter-sized Brainstorm/Ponder bubbles. That would truly show the legacy representation.

[bruizar]

Blue violates the color pie due to Polluted Delta, Flooded Strand, Misty Rainforest and Scalding Tarn. Time to axe.

[SpatulaOfTheAges]

I like it but tend to agree that maybe a map that had the top 500 cards but was scaled to show how comparatively huge BS and company are would be more accurate.

[Jain_Mor]

I like it but tend to agree that maybe a map that had the top 500 cards but was scaled to show how comparatively huge BS and company are would be more accurate.

This please, or less.

[somethingdotdotdot]

Really interesting graph. How are the distances between the bubbles calculated? Is it based on likelihood certain cards are present in the same deck?

Seems like miracles, dnt, infect, lands, snt, and eldrazi post are all on a downswing. Chalice decks like moon, steel, and eldrazi stompy are gaining ground. Spell-based combo and gy-based combo are doing well. Depths, elves, and maverick seem to be doing well. Burn seems to be well positioned as well.

Overall, slightly less blue and dnt in favor of more prison/non-blue combo.

[Ronald Deuce]

Hey, guise! If we make a graph with a bunch of "data" and shit, then we doctor it to make Brainstorm look like it's the best card in the format, that'll mean everyone will think (as we know, of course!) that it's the best card in the format!

Seriously, isn't there already a thread for this nonsense?

[FourDogsinaHorseSuit]

Hey, guise! If we make a graph with a bunch of "data" and shit, then we doctor it to make Brainstorm look like it's the best card in the format, that'll mean everyone will think (as we know, of course!) that it's the best card in the format!

Seriously, isn't there already a thread for this nonsense?

That Already exists (https://www.mtggoldfish.com/metagame/legacy/full#paper)

[Bithlord]

Hey, guise! If we make a graph with a bunch of "data" and shit, then we doctor it to make Brainstorm look like it's the best card in the format, that'll mean everyone will think (as we know, of course!) that it's the best card in the format!

Seriously, isn't there already a thread for this nonsense?

So, are you suggesting that it's not an incredibly strong card in the format?

[Ronald Deuce]

So, are you suggesting that it's not an incredibly strong card in the format?

No, I'm saying there's already a thread for people to make freewheeling diatribes about Brainstorm; this thread appeared to be about making an accurate representation of data.

[FTW]

Hold on, are you saying that the blue core is good in Legacy?

Thank god we had data to show it. All this time I was playing Leatherback Baloth like a fool.

[bruizar]

Hold on, are you saying that the blue core is good in Legacy?

Thank god we had data to show it. All this time I was playing Leatherback Baloth like a fool.

if welder beats is a plan, leatherback baloth looks fine to me.

After spending some more time looking through the viz I have to say I love how this represents the legacy metagame. The gradients really help identify the trend, blue (and goblins) has declined and has given way to more dragonstompy, reanimator, arclight phoenix, high tide, turbo depths, elves, slivers :D, spirits, burn, UB ninja, dredge. That's a lot of extra diversity!

Also found some really cool tech in the periphery of the viz: Throne of the High City, Rise // Fall and Estrid's Invocation

Any chance we can have this in an interactive Tableau public viz with a scrollable timeline so we can see the evolution over a longer period of time?

[sco0ter]

hey,
- calculated a "synergy" value between all of those cards
- calculated a gradient of how each card saw its meta share increase or decline since the first results after the ban up to today. So if on day 1 (after ban) some new cards or decks were tried but it didnt work out longterm the gradient is negative, if some card increased its share or made it newly into the format the gradient is positive. If the card wasnt really affected at all longterm then the gradient is close to 0
- caluclated the individual meta share of each card


What are the algorithms behind this? What software did you use?

And related to your last answer (Merfolk cards being near the Chalice.dec cards): How does the algorithm knows, that both are aggro-control strategies and therefore put them near to each other?
I think it's still confusing have such short distances between unrelated cards.

Red are declining strategies and green rising ones?

[tescrin]

1. What are the algorithms behind this? What software did you use?

2. And related to your last answer (Merfolk cards being near the Chalice.dec cards): How does the algorithm knows, that both are aggro-control strategies and therefore put them near to each other?
I think it's still confusing have such short distances between unrelated cards.

3. Red are declining strategies and green rising ones?

1. couldn't tell you for sure but it's likely just which cards see play near eachother most often; as that's the only data he's using. As in, if the card sees inclusion next to other cards commonly, they are placed closer together

2. I think you've forgotten that Merfolk started using chalice around the time of TNN/Cavern of Souls. Not 100% why that ended up being the best idea but it's still a thing if you go to the merfolk thread.

3. Yes. So when he says Blue is seeing slightly less meta share; it's because Brainstorm, Force, and (other card I assume is ponder but didn't check) are all tinged slightly red. When someone says Merfolk is seeing less play, it's because all the merfolk cards are red.

back on 2: His algorithm doesn't *really* group by strategy; but cards that are only seen in certain deck archs will be grouped between those decks. For example, TurboDepths doesn't use PFire, but you'll find PFire somewhere near its cards I bet; because both aggro loam and lands use Depths and PFire. You'll probably see Maverick cards floating somewhere between the D&T section and the Elves section (GSZ, Mother of Runes)

Thusly, the cards most tightly grouped in the center are those that are found in the most deck archs. What he meant by "the meta hasn't solidified yet" (or w/e) was that you see several extraneous cards in the middle that show multiple deck archs competing at the center for meta-share.

(for those of you who work in companies that hold quarterly all-hands about how the company is doing; this is very similar to "market share"; the more "market share' a card has, the closer it is to the center; and the more total copies found in decks the larger the circle.)

Funny enough, this means that the people who want Brainstorm to be larger are somewhat missing the point; the graph *already shows it's at the middle*, as well as Force of Will.

[SpatulaOfTheAges]

No, I'm saying there's already a thread for people to make freewheeling diatribes about Brainstorm; this thread appeared to be about making an accurate representation of data.

That's literally what people were talking about and you complained.