Metagame is the most generic characteristic of the collectible card game and its health indicator. Healthy Meta is when the player has some wide choice between well-performing decks and may pick something according to their preference. As for Marvel Snap, the game is very young, but it's already gathering interest, and of course players are willing to know what Metagame looks like. We decided to develop our own methodology to solve this problem.   

Important Marvel Snap Meta Prerequisites and Usual Errors

For the most collectible card games Metagame snapshot usually gets divided between Aggro, Midrange and Control decks, and archetypes are derivative of each of these Supertypes. Many games use card properties like color to make up archetypes. But with Marvel Snap it’s much simpler and much harder at the same time. Here’s the main specifics:

• Game is always limited to 6(7) moves and terms like “Aggro” and “Control” are a bit of a stretch.
• Cards don’t have usual “credentials” like rarity, color, clan etc. All CCGs tend to add as many ways to systematize cards as possible, but with Marvel Snap all those cards are pretty much mixed, and specs they actually have sometimes are obscure.
• The only real way to systematize the cards is based on what that card does. And not every mechanic has its own strong keyword emphasized, many cards just have a description text.
• The main force moving archetypes is synergy between cards.
• Random element of Locations is making archetypes more vague and less controlled by the player.
• Small number of cards available, which limits variability. That leads to a lesser number of possible archetypes in general.
• Decks are small. We are dealing basically with 12-card singleton decks, which is several times less than most of the CCGs have, since 30-40 cards is sort of an industrial standard. On the one hand, it simplifies the game, but on the other it makes each card very significant. So when you change even 1 card, the whole deck might feel differently, so the temptation to call the New Archetype any variation raises high.

Also, we are talking about Superheroes here, which adds its own fingerprint. Meaning, there’s always a fanbase for the superhero willing to make specific cards to be central in the archetype just because of the hero depicted on it.

The most common mistakes with Metagame approaches we observe in Marvel Snap community are:

• Attempts to put a usual label “Aggro” or “Control” where they totally don’t belong. To be clear, aggro is when you try to kill an opponent as fast as possible, with minimal number of turns. How exactly are we going to minimize the number of turns in the game where you always have 6 turns? Yes, the shape of the mana curve might look like a control or aggro deck in other games like Hearthstone or MTGA, but that doesn’t actually mean that low-mana deck will be more “aggressive”.
• Attempts to name everything a New Archetype. Player builds a deck, picks some card they like the most and calls it a new archetype. Well, if you have a typical Move deck and replace in it a few cards, you are not getting a new archetype, you are still dealing with Move deck.
• Too many archetypes. There are slightly more than 200 cards out there, so you can’t make up hundreds of archetypes using those. And if every deck will be considered as a new archetype, then how are we going to calculate the Meta?

Having all those prerequisites, let’s move to our methodology. 

Marvel Snap Pro Meta Methodology

When we approached this task, we had big experience of metagame analysis for other CCGs like MTG Arena, Legends of Runeterra and TES: Legends. We see what’s going on within community and decided to find the approach which will lead us to the systematic and complete meta:

• As small human involvement as possible. We just analyze the data and create the algo, and we don’t modify the output manually.
• As few archetypes as possible. Occam's razor never gets old, so we are trying to achieve as few instances of new archetypes as possible.
• Reasonable archetype formulation basis. We tried to find out what forms the Archetype, and we noticed that data spikes are based on card abilities, specific staple cards and card synergy. Those three sources are basic for archetype detection.
• Only the latest patch contributes to Meta. When we get too much data, the error rate is climbing as well. And cards are getting nerfed for a specific reason: to shift the meta. That is why we cut off metadata samples when a new patch launches.
• As wide coverage as possible. Having only 200 cards, we should understand that the community will find popular and powerful (meaning winning) decks, so the majority of new decks submitted should be variants of those, therefore majority of new submitted decks should have autodetected archetype.

Based on that, we Implemented following Marvel Snap Meta Snapshot methodology:

• A week since the new patch hits we gather and analyze the data, detecting spikes.
• Spikes are detected based on card mechanics, primarily on Destroy, Discard, Move and Ongoing effects. But also, we add “Increase Power”, “Add Card“ and some mechanics. We call those “Supertypes”.
• Two kinds of Archetypes are detected: basic Archetype based on mechanics (for example “Move”) and card-specific archetype (for example “Cloak Move”). Each card-specific archetype belongs to the general mechanic-based Supertype.
• Then we autodetect archetypes for all decks we have and count them. After that we get meta share percent. We abandoned the “Tiers” term on purpose, because at this point all Tiers are very subjective and strongly rely on human attitude. But the percentage shows Meta share, and each reader can decide for themselves whether 7% is Tier-A or Tier-B.

Currently almost 80% of decks are getting auto-assigned archetypes, which leads us to the conclusion that the system we have is pretty stable and represents the situation in Meta realistically.

We render Metagame Snapshots on a weekly basis, you can find the latest one here.