Silvestri Says – Stats

This week is a Magic Online extravaganza where we talk about all the wonderful data replays provide us with. In this case my friend and stats guy Rolle provided me with a ton of useful data for both Standard and Theros Draft. Today we’ll be breaking down that data and seeing where the online metagame differs


Total Games› 43385

Screen Shot 2013-10-29 at 2.48.57 PM

Screen Shot 2013-10-29 at 2.50.29 PM

Looking at only decks with at least 3% of the field or more we get the following breakdown:

Tier 1: (53-56% win rate)
Esper, Mono-U, GR

Tier 2: (51-53% win rate)
Azorius, Mono-R, Boros

Tier 3: (Sub 50%)
Everything else, notables being: Mono-G, MBC, BWR Mythics, WBG (Junk)

Screen Shot 2013-10-29 at 2.51.13 PM

How the Meta is Changing

Play percent by date1
Play percent by date2

Disclaimers: Because of how the data is collected, there’s a limit to how much actual deck separation can be done. As a result, I couldn’t tell you what the exact percentages are from GR Devotion vs. GR Aggro. It does tell us that at least one of the archetypes is doing significantly well and the other isn’t doing badly enough to notably lower the win percentage. If we cross-reference this with cashing Daily Events lists, we also see that GR devotion decks largely outnumber the aggro lists about 5 to 1.

Same goes for the massive field of mono-black and mono-blue. Yes, some pet decks are scooped into the pile, but the vast majority are all the archetypes you’d expect. Mono-black control and mono-blue devotion.

These stats are also not the end-all be-all for deck selection purposes. They give a lot of insight into what to play for the majority of Magic players, including those who are good but consistently stick with worse options. This not able to take into account a theoretical best deck that perhaps would be the best choice for a given metagame with an extremely competent pilot behind them. I doubt Reid Duke would want to take advice from just this data.

For most of us, it provides a useful pool of information collected from a number of games no team, let alone individual, could battle through in a reasonable amount of time. It also showcases the evolution of the online metagame through a very tumultuous time.


1) Mono-Black Control Isn’t That Good

Mono-black control just hasn’t been all that impressive online in the majority’s hands. Notably, a handful of the grinders who play on known accounts still have solid win percentages, like BBD and Todd Anderson, but most MBC players are coming up short. With that kind of metagame penetration, mirrors have to be taken into account, but even then it shouldn’t have a sub-50% win percentage. When mono-red consistently had over 20% meta output it remained in the 51-52% win range.

What could be the possible explanations for this massive popularity and low-win results?

A. The Magic Online metagame simply is less favorable than the real life one for MBC

You can see this in how much red is online and how quickly the metagame skews toward newer popular decks. Why does this matter? Because MBC suffers when pilots are uncomfortable with a given match. The deck is very much a 1-for-1 machine hoping to get ahead when it doesn’t have [card]Pack Rat[/card] to overwhelm the opponent. Not understanding what threats are correct to spend removal on is a major detriment.

B. The online metagame is quicker to adapt and adjust to a new “top deck” compared to real life

Part of this is likely attached to how fluid cards are online. Eating a 5-10% money loss to change decks or potentially even gaining money when swapping popular decks for less popular ones is a lot easier than real life. When decks online do well, people often either switch over to them for a bit to try them out or switch to a counterplay deck. In this case, mono-red, after the Waves deck pushed many red players out of the metagame.

Now the meta is flooded with a very good match for red decks and we see a quick resurgence.

C. The deck was cheap

Anything that costs less than 150 to build out of the gate will get people to immediately jump on board. Of course, building it now is 229, but that’s beside the point. It was cheap when the initial rush of people came in to build the thing.

2) Jace and Revelation Decks Aren’t Going Anywhere

Every other week people lament the death of control or talk about how this is the week slow control decks will get pushed out for good. It just isn’t happening and won’t for the foreseeable future. Maybe the next sets will make the metagame too hostile, but for now both Esper and UW remain solid choices. Esper may be seen as the better of the two choices, but keep in mind that if an aggro resurgence comes to bear, UW is the place to be.

Notably, UW gets very little chatter and is largely dismissed as a worse Esper deck, but if red comes back big then I’d expect the win percentages to be exchanged for the two decks. It’s important to keep in mind moving forward why and when Esper is the superior control deck.

3) Red Decks are Underrepresented in Real Life

Both small red and Boros flavors are succeeding online in a field full of other very publicized and respected decks. Yet, they get very little attention, let alone major play. The mono-red link to the cost of the deck and availability of Daily Events (a.k.a.: It’s a grinder deck) is well-known. However, I don’t think people should completely write these two decks off because of that.

I think both decks suffer from a lack of organized advances and attention from skilled players in the real-life metagame. How many players are well known for Islands, Jund, and a host of other masteries? How many of them pick up red decks on a consistent basis? Owen showed off that RDW is a deck that can succeed even when the metagame aims at the deck.

Of course, the answer can largely be traced back to [card]Master of Waves[/card] and the deck that blew up immediately following the Pro Tour. Once again, though, we have to ask ourselves how many people will be playing the deck at a given tournament versus all of the other good matches in that same tournament? Oh, and whether RDW is just dead to Waves without exploring any others avenues—like, say, adding another color to the deck.

4) The Best Decks Online are Either Devotion-Based, Good at Controlling Devotion, or Small Aggro

You’ll notice a distinct lack of midrange decks, and the devotion decks not going way over the top seem to be doing worse than the two that do. Decks focusing on 1-for-1s without clear end goals (midrange decks) falter. It wouldn’t surprise me to learn that MBC’s adoption of four Pack Rat, other than being impressive in the mirror, is a direct response to other decks simply forcing it to take a more active role in matches.

Small aggro doing well is interesting, some can be blamed on Esper not doing enough early outside of tripping over its own mana and praying for [card]Supreme Verdict[/card] on turn four every game. I think a lot of it has to do with how small aggro and Waves attacks the opponents, both put out a swarm with limited protection/reach. Both aspects of small aggro can potentially kill the opponent and opponents often need to successfully answer both to win. These small decks also have the option to go bigger post-board or simply tweak which mode they want to lean on, significantly changing how matches play out.

Now we move onto Limited.


Total Games 56950

Screen Shot 2013-10-29 at 2.54.42 PM

Screen Shot 2013-10-29 at 2.55.47 PM

Screen Shot 2013-10-29 at 2.56.04 PM

Card Win % Spreadsheet

Card Win % Only C/UC Spreadsheet

Improved Expected Win % Spreadsheet

You’ll get the most out of the card data by sorting by Win % or Likely Win % and by common/uncommons to compare peers against peers.

Note that the “Likely Win %” takes into account the percentage of time you’ll be able to cast your spell. It’s percentages are based on the average likelihood that you will reach the CMC in number of lands across all the replay games of Theros. Note this tends to have a much more accurate turnout when it comes to commons and reverses the bias for high CC cards. While I may not agree with all the potential conclusions drawn from this data in regards to certain high-end bombs, I think it does a great job of filtering out the common expression of high-end bombs being dominant and puts the commons into real view.

One of the major takeaways from the data, no matter how you process it, is that high-end bombs tend to win a lot of the time. However, they also end up played far less often than what people expect. Based purely on my own discussions with players, I find many people take for granted that they’ll hit six mana by turn seven. Cards like [card]Sea God’s Revenge[/card] and [card]Abhorrent Overlord[/card] are often lauded without any real consideration given for how often they sit dead in your hand. It’s not that all high-end cards are bad, they need a substantial game-changing impact or they likely aren’t worth putting in your deck.

Rolle added this alongside the data:

“THS is “average” speed. Going back to overvaluing expensive cards, I don’t think people fully understand what this means though.

You reach 7 mana in 41% of games. 6 mana in 66% of games. 5 mana in 85% of games.

How many 6 or 7 mana cards can you really afford to play that don’t just win you the game? It also means you are likely only hitting monstrosity in about half your games, unless you are ramping, and I believe this is reflected in the archetype win%.

The advantage of a huge data set is consistency. It takes away personal anecdote and the result is for Theros—the best strategy over time is just focus on reading the draft and trying to get in open colors, then drafting a proper curve, and this means more 2-drops than anything else if possible, then 3-drops, 4-drops etc. unlike M14 which barely needed a curve at all.

Synergy in the form devotion, ramping or even heroic should not overrule “solid” cards. It’s not that synergy can’t work, its just never going to be as consistent in a format that is average speed or faster.”

I feel like curve considerations are something that get put on the back burner by too many otherwise reasonable Limited players. Just because a format isn’t blazingly fast doesn’t give you a pass to ignore curves, especially in a format where the best draft decks are built around them. Looking at the best performing archetypes, the common theme is that being able to pressure early and interact early with aggression are two of the most important aspects of the format.

Green is a great example of a deep color with a strong mid- and end game and crucially few early interactive cards. This makes it an excellent companion for blue because blue has those early interactive cards with [card]Voyage’s End[/card] and cheap evasive creatures to race or block other evasion.

Red is a support color as talked about last week and is slightly under drafted online.

Black is likely over drafted because the mono-black devotion deck is likely one of the best decks in the format when it comes together. However odds are good it’s like trying to draft a successful Storm deck in Cube. Very satisfying when it happens, and very hard to put together consistently.

On card win percentages: First, replay data can never fully take the place of actually playing the game out. They aren’t the final say on how good or bad a card is, even if we tweak the data against potential bias. Second, the gap between the highest echelon of Limited Magic players and theory may be far larger than we imagine. I hope to explore this down the road, but what it may come down to is simply a fundamental difference in how certain players approach and play Limited compared to the majority.

What’s interesting is that on the whole this may not actually matter to you. Unless you consistently are playing against players on that level, odds are the data points help. Knowing what people consistently draft and do well with in 8-4s may serve as a guidepost.

Finally, having this data is interesting because it lets us make more useful distinctions than just calling a card good. It was really hard to talk about a pair of cards valued equally, because you quickly run out of arguments one way or the other. At that point, it came down to whatever the preference was. Now you can look at the big board and see if the two cards are really that equal after all. It also gives us an interesting look into major discrepancies in card evaluations. For example, on Frank Karsten’s excellent breakdown last week, Tom Martell posted about his undying love for [card]Favored Hoplite[/card] and was roundly mocked.

[card]Favored Hoplite[/card] – 55.7% win rate, putting it as the 6th best common/uncommon in white, likely closer to 3rd/4th once we take data bias into account ([card]Hopeful Eidolon[/card] costing 1, but bestow for 4 amusingly dodges the Expected Win Rate solution). Even if we assume it is in 6th, it’s a cornerstone in heroic strategies and we know that it’s performed at a reasonable level among MTGO players.

Anyway, the data kind of speaks for itself in this case. I’ll let Rolle sum it up the format as a whole:

“It’s a pretty well-balanced format. WR heroic really is the strongest archetype. The reason it beats UW heroic looks like the fact that red is slightly under drafted, you are contending with UB and UG for the strong U cards. That said UG and WU are strong. WG, WB, and UB are all solid. GR and UR are a little behind.

GR I think is behind because the ramp plan plays a bit like combo in draft. You often don’t have the redundancy you need to have consistent ramp into fatty that you do in Constructed. Plus, the GR uncommon is pretty sucky by comparison. The UR uncommon is also very bad.

I think this counts for a lot, in formats like this a lot can ride on the strength of your uncommons.

The mono-color decks are behind, B and R do the best obviously since they reward devotion the most. But people easily forget the mono-color decks where they’ve had to play sub-par cards then lost R1 and remember vividly when they had those absolutely crushing draws that devotion enables (which may have still lost in R2 when they just get outclassed by someone with higher average quality).”

My own anecdotes have largely been in line with the data, with the exception that I thought WU was the best deck in the format outright. Perhaps that’s incorrect, but honestly the good WU decks feel like I’m playing Block Constructed, while the bad ones are just average WU Skies decks in Limited. It also amuses me how good WB is, even though the mana is just so painful and black devotion synergizes so badly.

Whew, that’s a lot to take in. Have fun looking at this rare insight into thousands of games of Magic and see if you pick up a thing or two.

Josh Silvestri
Email me at: [email protected]

3 thoughts on “Silvestri Says – Stats”

  1. Pingback: Limited picks orders and cheat sheets | Drafting with fourouttheforty

  2. Pingback: American Control | Magic the Gathering: Blogger

  3. Pingback: Exava is the New Shitstorm

Comments are closed.

Scroll to Top