Metagaming For Modern Tournaments

I’m a big believer in using metagame data to inform tournament decisions. Whether you use the information to guide your deck choice, change around some maindeck cards, or determine your sideboard bullets, players who are aware of metagame trends are much better prepared than those who are not. Unfortunately, it’s not always clear how a big pile of numbers should inform these important decisions. Nor is it always clear how we should translate a metagame breakdown, like my 7/1-8/1 update from last week, to actionable deck and card choices. Numbers can be daunting for some and arbitrary for others, and it’s hard to know what it really means when we say a “metagame is 8%-9% Jund”. Obviously, this translation of data to practice has big implications for your tournaments.

Arena Art

This article gives you some guidelines on how to tackle these issues and metagame for Modern events. I’ll discuss how you should “read” metagame numbers and translate those numbers to actionable metagame decisions. I’ll also explain how you can put the big-picture metagame numbers in a local context, tailoring these decisions to your area metagame. Whether you are attending a PPTQ later in the summer, heading over to an SCG IQ or Premier IQ, or getting ready for Grand Prix Oklahoma City in September, I’ll give you some tips to tune your decks to both the current metagame and any future metagames you dive into.

Translating Metagame Statistics: Quantitative

Whenever I post metagame articles, or read other ones, I see tons of questions about interpreting the numbers. The most common ones take the form of: “So if Deck A is N% of the overall metagame, does that mean it will be N% of my upcoming tournament?” Other commenters can be a bit more aggressive: “I went to a tournament this weekend and the metagame was nothing like the article predicted!” If you go to your event expecting a 9% / 8% / 8% split between Jund, Affinity, and Burn and then you see 20% of the field is on Twin and there are only two people each on those three decks, you are justifiably going to question any metagame breakdown you read the night before. These are reasonable questions when reading metagame data, and we should expect (even encourage!) critical consumers of Modern information to ask them.

As with most statistical questions, there are quantitative and qualitative ways to unpack this data and make sense of it. You’ll need both to succeed. In the spirit of the primarily numbers-based breakdown articles, I want to start with two quantitative approaches to metagame data. Then we’ll turn to the qualitative ones.

Dark Confidant MM2015The first quantitative concept is the idea of a margin of error. You’ve probably heard this term before in reference to surveys and polls (and we’ll hear much more of it as the American presidential campaigns kick into higher gear), and it’s a great tool for understanding metagame data. Margins of error are useful when you have a sample of results from a population and not the population results themselves. If we knew what every single Magic player brought to every single event, we wouldn’t be taking a sample of the data: that would be the population itself. Because we are taking data from reported events and Top 8s/16s, however, we are necessarily dealing with a sample of the overall population. Margin of error gives us an idea about the variation between our observed results in the sample (e.g. Jund being 9% of the observed 7/1-8/1 metagame) and the “true” results in the population (e.g. the actual number of Jund players between 7/1 and 8/1). There are lots of ways to estimate margin of error, based on the size of your sample, the distribution of results, how representative you think the sample is, etc. In all those cases, you are using margin of error to say that the “true” prevalence of a deck isn’t just the 9% reported in a breakdown, but the spread around that percentage.

We track metagame margins of error on our Top Decks spreadsheet. You can see these margins of error on the top of each metagame tab; as of this article’s writing, the “Paper” tab indicates a margin of error of +/- 3.46%. This means the “true” prevalence of a deck like Jund is not just the observed 10.85% we see in the table. Instead, it suggests the “true” prevalence is somewhere between 7.4% and 14.3%. It could even be lower or higher than that depending on event attendance! A smaller event is necessarily going to have higher variance, so a 16-player tournament could see Jund prevalence as low as 5% or as high as 20%. The trick here is not to fixate too heavily on the individual metagame number. Instead, it’s to think of those numbers alongside the margin of error. In fact, while writing this article I’ve decided to start incorporating this margin of error measure into future metagame breakdowns, so you can expect to see more of it in the future.

TwinOur second quantitative concept is that of relative magnitudes. That’s more or less a fancy way of saying “seeing if one deck’s share is bigger/smaller than another’s”. Metagame numbers do not exist in a vacuum. When you read a breakdown, you should not fixate on Jund being 8.9% of a metagame or UR Twin being 5.3%. Instead, you should look at the relative magnitudes of decks in the metagame: UR Twin sees a little more than half as much play as Jund. Or, to take the Affinity (8.4%) and Burn (8.1%) example, we might conclude these two aggro decks are about equally likely to appear at an event. In these cases and all the others we might construct, we aren’t focusing on the specific numbers but rather on the relationship between those numbers. This is hugely important in a diverse format like Modern. It’s going to be hard to prepare for every possible deck, so you need to make maindeck and sideboard choices to prepare for some decks more than others. The idea of relative magnitudes helps you do that, pointing you to prepare more heavily against one deck (e.g. Burn) over another (e.g. Merfolk).

Relative magnitudes are even more important with tier 2 decks than with those in tier 1. As we define them on the Nexus, tiers are prevalence-based measures more than performance-based ones. Although performance is certainly correlated with prevalence, there are other factors which can drive high deck shares beyond just a deck being “good”. This includes budget, playstyle preference, hype/Heritage Druidpopularity, and a host of other factors. Prevalence metrics might be theoretically more useful but, in practice, can be very arbitrary. There isn’t enough good data to track this and even the best data sources (MTGO match win percentages) can be complicated by all of the other factors described above. Because we focus on prevalence-based tiering, a deck’s tier needs to give it additional weight when comparing it to other decks. You could probably treat a tier 2 deck’s prevalence as half of what it actually is when comparing it to a tier 1 deck.

As an example of this, Elves has a 2.2% prevalence right now, which is about 50% of Merfolk’s prevalence. Because Elves is tier 2, however, you should probably treat Elves as having only a 1% prevalence for the purpose of comparing it with Merfolk; the deck isn’t played half as much as Merfolk at the average event. It’s probably played far less than that. There’s probably an exact value for the tier 1 and 2 weighting, but a 50% multiplier for tier 2 seems like a good starting point.

Using margin of error and relative magnitudes, you will be much better prepared for leveraging metagame data in a tournament setting. These tools give you a more realistic sense about what metagame statistics mean beyond just a single percentage share. You can also use these tools in any metagame, not just a Modern one. Interested in Standard or Legacy? These analytic methods will be useful in orienting you to those formats as well.

Translating Metagame Statistics: Qualitative

Numbers alone are never going to be enough to understand a metagame. As much as we might love a formula that could predict the metagame for any given tournament, it would probably be impossible to quantify all of the subtle qualitative variables at play in Modern. I hinted at these in the previous section: budget, playstyle, hype/popularity, local variations, and other factors all play a role in influencing the overall metagame data like we see on our site. I want to focus on two of these factors because they often go underappreciated in tournament preparation: budget and local variation.

Metagame breakdowns are often budget agnostic. They assume all players have the same budget and can all play whatever deck they want to, and while that might be truer of a Grand Prix level event, it’s Puresteel Paladincertainly not the case at a random PPTQ or SCG IQ. True, metagame numbers account for budget in some capacity (indeed, that’s one reason we see so much Burn and Affinity), but they don’t account for different effects budgets can have on different events. Going to a Modern FNM in an area known for Standard events? Expect more players trying to get into the format with budget decks. Going to an established Premier IQ in a major metropolitan area? You’ll see a lot more decks like Jund and Twin. Generally speaking, the higher the stakes, the less budget becomes a serious consideration. Of course, this assumes players go to high-stakes events with high-stakes expectations. Maybe players take their Puresteel Paladin/Retract combo deck (go go Cheeri0s!) to a GP without any serious intention of winning: they just want to try their chances in a competitive setting. The more you expect budget to be a factor at events, the more you should expect to see the cheaper decks (particularly the cheap tier 1 decks).

Budget considerations relate to another qualitative factor, that of local and regional variation. Everyone knows that one guy who always shows up with his pet Slivers or Allies deck (yep, it’s often tribal). Even collected companydecks like Death and Taxes, 8Rack, Storm, and other slightly more established decks will show up in this category. These decks aren’t “bad”, per se, but they aren’t tiered in the same sense as Twin and Jund, or Amulet Bloom and Abzan Company. If you know there’s a guy who tries to stick Collected Company into every offbeat creature and tribal variant from week to week, don’t think this guy is going to change his M.O. just because a metagame update says Company Werewolves and Soldiers are tier 5 or lower. These qualitative datapoints are critical in readying yourself for local events, especially at the PPTQ and IQ level. This doesn’t mean you won’t also see the tiered decks show up; even small tournaments have players who are up to speed on the most recent metagame developments and are packing the tier 1 front-runners from the most recent Grand Prix. The key is to (roughly) know what percentage of your local event is on the homebrews and pet decks and what percentage is going to go in the tiered direction.

Using Modern Metagame Data

If you combine these qualitative and quantitative approaches, you will be in a great position to understand the metagame and apply this data to your deck and card decisions. Whenever you read a metagame breakdown, whether on this site or others, you should use those percentages and numbers only as a starting point.

How would you use metagame data to inform your decisions for an upcoming event? What are some other factors you would consider in your own thought process? If there’s more interest on this “metagame interpretation” topic, I’ll definitely write some more articles on the subject. Until then, keep metagaming and keep on Moderning!

Sheridan is the former Editor in Chief of Modern Nexus and a current Staff Author. He comes from a background in social science data analysis, database administration, and academia. He has been playing Magic since 1998 and Modern since 2011.

15 thoughts on “Metagaming For Modern Tournaments

    1. Burn is definitely one of those decks you see a lot of. The deck isn’t as cheap as it used to be, but people certainly THINK it’s cheap (Merfolk is actually cheaper). It’s also probably the “easiest” tier 1 deck to play. At least, people believe it’s the easiest deck and can certainly succeed without knowing all its ins and outs.

      1. I think is why burn catches so much crap as being an “easy deck”.

        The fact is, you can play burn poorly and still do relatively well – a lot better than with a control deck, for example.

        That being said, the difference between a skilled burn pilot and an amateur is remarkable. People don’t give enough credit to how complex the deck actually is, they just see poor players score free wins with it and dismiss it as a one-trick pony, which it most certainly isn’t.

        I don’t play burn myself, but I can respect how difficult it can be to play correctly.

        1. Agree. We actually have a Burn primer coming out tomorrow that goes over the deck nuances. Hopefully it helps people get to the next level with the deck, especially those budget-minded players who pick it up for the first time.

  1. “If you go to your event expecting a 9% / 8% / 8% split between Jund, Affinity, and Burn and then you see 20% of the field is on Twin and there are only two people each on those three decks, you are justifiably going to question any metagame breakdown you read the night before.”

    No, no you’re not.

    That Jund is 9% of the meta globally has almost zero bearing on its share locally. Some meta’s have more, others have less. Take GP Singapore as an example – that GP clearly had an Affinity bias – that was a local flavour. You can’t transfer that to every area.

    What the meta %’s tell you is what percentage of the decks across the world are doing well. Jund, Affinity, Burn – are all decks that are around and you should be prepared for them in some sense as they represent strong decks with strong representation.

    Even if there’s only one Jund player at a tournament, you’d still do well to test vs the deck because there’s a good chance you’ll face the deck in the later stages/top 8 of the tournament. The actual percentage doesn’t matter – you need to determine what you’re likely to face at the top end of the tables if you want to win, and meta% shows you that.

    For example I know my meta, there’s very little Jund, almost no Grixis Control, a lot of Affinity and Twin and creature based aggro decks (Naya, CoCo, Elves, Hatebears). And in the early rounds those are decks I’m likely to face, but at the end of the tournament the tier decks will come through and it will be the one Jund player, a few Affinity, some creature deck that goes wide and possibly a Burn player to go with all the Twin.

    That’s how it goes – despite the actual percentages at the event not equalling the meta share, the top tables will over time reflect something close to the numbers represented.

    1. I agree with you. I was only saying that some players are going to think that and they aren’t ENTIRELY unjustified in their confusion. People just assume a lot from metagame numbers. But you are right that even at local events, where global metagame numbers are less applicable, the top tables will tend towards those global numbers.

  2. And then you have things like my run at the premiere IQ yesterday–4c kiki-chord without ghostway or any of the other strange tech, he had a tribal walls thing going, 4c gifts-loam-reanimator, infect, RG tron, agro elves with dwynen’s elite, naya collected company/hatebears, junk, scapeshift. And yes, I got blasted by 1-of magus of the moon all three games out of naya coco, triple coco/triple ezuri out of the elves deck g1, g2 stuck on three lands until he killed me on turn 8, two board wipes in hand, and back-to-back games against kiki chord where I hit >16 lands, and back to back t2 kills by infect. Played against zero of the twin, affinity, burn, and jund that I was prepared for/ready to beat, and all of the random-ass creature brews. By the time I played against real decks (junk, scapeshift) I was already playing against people who didn’t even know how their own cards worked.

    Days like this make me hate myself for playing a deck metagamed against tier 1 in a large event instead of just jamming burn or affinity.

    1. “Days like this make me hate myself for playing a deck metagamed against tier 1 in a large event instead of just jamming burn or affinity.”

      Dear Anonymous – maybe you should just play burn or affinity. Those tier 1 decks also have problems – they have decks played against them designed specifically to beat them, metagamed if you will. It is exceptionally frustrating to play/lose to those metagamed decks when those decks are not powerful enough to handle the remainder of the modern field. That is a frustration of the highest order.

      Also, you seem to be implying that playing affinity or burn is less skillful than your metagamed deck. You are in all likelihood mistaken. Everyone comes to the IQ ready to beat up on affinity and burn, having practiced the matchup multiple times and with express sideboard plans and counter sideboard play. Against your metagame deck, you get a few free passes because no one knows your list.

      Maybe you ran bad, or maybe your deck just cant handle a wide open meta in which there are many tier 2 options capable of taking down the title

    2. I’m not a big fan of over-metagaming for anything less than a GP. Even Premier IQs have a lot of random stuff from tier 2 and lower. It’s best to just play a deck you are good at. Bonus points if the deck has some linear and/or proactive ways to win. Jund and Grixis Control are good here because you can play a reactive gameplan in some matchups, and then just smash face with Goyf/Angler/Tas in others. One risk with decks like Kiki Chord (as much as I love the deck) is you are relying heavily on your minimal interaction in some matchups and bad situations. When you’re only running 4 Path or something similar, threats like Magus become very scary.

      Also, I agree with Josh here. Kiki Chord is an awesome deck, but a very risky decision for an open field. It’s also untested in the broader metagame (not even tier 2 yet), and its pilots might find it lacks firepower in key matchups. I think you would be okay going with at least a tier 2 deck, but the tier 3 or lower stuff can be extremely high risk, high reward (or worse: high risk, low reward).

  3. Thanks for your articles, it is very informative and analytical. to add on to your articles, these are some of my thoughts. it is sometime not possible for a player to simply switch deck and expect their result to remain the same. very often the meta changes because of certain type of cards that are played more. the emergence of these cards can push the meta towards certain colour or even meta that can affect the percentage of certain deck. so to use this as a basis, we can use the prevalence of that card in various archetype and how it affects the interaction with the various tier 1 and 2 deck and also how the various deck respond to it with their sideboard or mainboard. I am an affinity player. in the past I run ensoul artefact, however with the prevalence of kolaghan’s command, I change it to spellskite. this decision have an impact on my match-up against twin and infect in game 1 and such. if other deck type followed suit with a similar response, twin and infect metagame percentage will go down because it isnot favourable to them and such. therefore sometime it is not the archetype that matters but rather the card that is used that matter.

    1. Agreed that deck switching isn’t always feasible. Metagame changes (for whatever reason) can invalidate certain deck choices no matter how experienced you are with them. I’m personally okay with this: cyclical metagames are healthy ones, and if everyone had access to a deck that was always the best, that would be a solved metagame. But as you also say, it’s important for players to be flexible WITHIN their deck. This means changing around cards to fit the metagame. I think a lot of players don’t do this often, which is why they get blindsided by random Infect and Grishoalbrand opponents.

  4. When you talk about margin of error, are we talking 95% confidence interval?

    Also, I wonder how useful it would be to plot daily metagame and weekly tournament data and use the slope of each deck to predict inclines or declines in prevalence. I would be interested to see how these graphs looked like for decks and how we may use known weaknesses a la Rock Paper Scissors to predict declines in certain decks that are weak against decks showing a positive slope.

    1. Re: margin of error
      This particular margin of error was constructed using the 95% confidence interval.

      Re: Weekly tournament plotting
      I’ve considered something similar before, but encountered two issues. The first is sample size. I normally don’t care too much about a small N, but when your N can be < 5 in some weeks, that's a little too small for me. It's hard to extrapolate much from that, even if we bootstrap the sample (which doesn't even make too much sense in the context of Modern tournaments). The second issue is the disconnect between events. Just because some stuff does well in a bunch of Italian and North American events, it doesn't mean it affects Japanese events in the next week. Even within a continent (or even country!) it's unclear how those relate. This leads back to a broader question about regional and local metagames, but unfortunately it doesn't solve the problem: it just brings it to the forefront.

Leave a Reply