The last metagame update before Modern is once again permanently changed by a Horizons set. While Modern Horizons 3 looks tamer than its forebearers, looks are often deceiving and players are famously bad at cold reading new cards. We'll see just how tame this actually is next month.

The Outliers

As has been normal, there are statistical outliers in the data. The top deck for both Magic Online (MTGO) and top two in paper (Rakdos Scam, Yawgmoth and Amulet Titan, respectively), are statistical outliers and, per policy, were removed from the data analysis though they're reported in their correct place on the tier lists. This completely defies my expectations prior to performing the tests.

In the case of MTGO, I thought there'd be more outliers than there actually were. The gap between the top decks and the rest is so massive I expected a repeat of April's outliers, or even more. However, the tests all pointed to just Scam. Everything else was close enough to the trend line, apparently. In paper, things played out as I expected. The gap between the top decks and the rest was pretty definitive, but I always confirm rather than assume.

May Population Metagame

To make the tier list, a given deck has to beat the overall average population for the month. The average is my estimate for how many results a given deck "should" produce in a given month. To be considered a tiered deck, it must perform better than "good enough". Every deck that posts at least the average number of results is "good enough" and makes the tier list.

Then we go one standard deviation (STdev) above average to set the limit of Tier 3 and the cutoff for Tier 2. This mathematically defines Tier 3 as those decks clustered near the average. Tier 2 goes from the cutoff to the next standard deviation. These are decks that perform well above average. Tier 1 consists of those decks at least two standard deviations above the mean result, encompassing the truly exceptional performing decks.

The MTGO data nearly exclusively comes from official Preliminary and Challenge results. Leagues are excluded, as they add analytically useless bulk data to both the population and power tiers. The paper data comes from any source I can find, with all reported events being counted.

While the MTGO events report predictable numbers, paper events can report anything from only the winner to all the results. In the latter case, if match results aren't included, I'll take as much of the Top 32 as possible. If match results are reported, I'll take winning record up to Top 32, and then any additional decks tied with 32nd place, as tiebreakers are a magic most foul and black.

A Note on the Data

Daybreak is now releasing the total results from every MTGO Preliminary, Challenge, and League 5-0. After some experimentation, I'm sticking to just using the Challenge Top 32 results and 3-1 or better from the Preliminaries. The first reason is that, ultimately, nothing changed. The population metagame list didn't change between my normal method and the experimental versions. Various treatments for the power metagame did change the order of the tier list, but the composition varied only marginally.

The second reason was that dealing with all that data is significantly more work, even with automation. I'm not a great programmer, and setting up and training the bots, then auditing the results, took significantly longer than my current system, plus I'd have to redo it monthly. Since it made little difference, I'm not going to make more work for myself. There are other sites that put together winrates with all the new data anyway, so I don't feel that anything's being lost. It also means that comparing the paper to MTGO results is easier.

The MTGO Population Data

May's adjusted average population for MTGO was 15.44. I always round down if the decimal is less than .20. Tier 3, therefore, begins with decks posting 16 results. The STdev was 31.57, so add 32 and that means Tier 3 runs to 48 results. Again, it's the starting point to the cutoff, then the next whole number for the next Tier. Therefore Tier 2 starts with 49 results and runs to 81 Subsequently, to make Tier 1, 82 decks are required.

May's population is down from April's yearly high. January had 1,400 decks, February was 1225, March hit 1042, April hit 1664, but may was down to 1424. This was despite there being a number of extra super qualifiers on the schedule, but the Preliminaries were much smaller than normal, and it looks like some didn't fire. I'm pretty sure that's what caused the drop, and that it's down to players waiting for MH3.

The fall in sample size is accompanied by a fall in diversity. The total number of decks in my data set fell from 103 to 79. This means that the unique deck ratio also fell from .062 to .055, the worst I've measured this year. The number of tiered decks fell as well. There are only 18 decks on the list, down from 26 decks in April.

Deck Name	Total #	Total %
Tier 1
Rakdos Scam	220	15.45
Yawgmoth	145	10.18
Living End	141	9.90
Amulet Titan	128	8.99
Counter Cat	119	8.36
Izzet Prowess	97	6.81
Tier 2
UW Control	62	4.35
Tier 3
Izzet Murktide	45	3.16
Jund Creativity	42	2.95
Burn	37	2.60
Hardened Scales	27	1.90
Rack Scam	26	1.83
Temur Prowess	24	1.68
Coffers	24	1.68
Jund Saga	24	1.68
Goryo Blink	19	1.33
Mono-Green Tron	16	1.12
Mono-White Emeria	16	1.12

I put a lot of the blame for the state of these statistics on MH3. Players are just waiting for the new set rather than putting work into the current format, and as such they've crawled back to their known decks. It's an understandably rational decision. However, when coupled with MTGO's small player base, it means that events quickly start to look recursive. I understand players being bored and/or frustrated with the online metagame, but this is what happens when everyone is waiting for a big change to happen in a few weeks. At least Prowess is back?

Classifying deck variants was considerably harder this month than previous months. Around the middle of May, some Rakdos Scam players cut their Ragavan, Nimble Pilferers for Caustic Bronco. Moving around numbers usually doesn't represent a change in gameplay, requiring a deck to be reclassified, but cutting cards entirely can be. The Rags made a comeback around the 20th, but I'll be watching this one.

Meanwhile, the indistinct blob of decks lazily lumped together as Domain Zoo grew some new offshoots. The primary configuration remains Counter Cat by a lot, with Domain Murktide, Domain No-Murktide, and Domain Scam as frequent alternatives, despite all being listed in the same database. May saw some more variants branch off, such as this one that cuts Wild Nacatl for Caustic Bronco and Orcish Bowmasters and this variant that utilizes Break Out and plays more like classic Zoo. These decks play very differently, and should not be classified together, but I seem to be the only one who cares.

The Paper Population Data

Paper's dataset continues its rollercoaster ride. January had 803 decks, February 890, March had 311, April was up to 559, and now May is down to 389. The real diversity is holding steady, which is positive news. April had 88 unique decks and a unique deck ratio of .160. May has 66 decks and a ratio of .169. The ratio's up a bit, but that doesn't really mean much in context.

The number of tiered decks fell insignificantly from 24 to 23, which is quite good given the low population. The adjusted average population was 4.64, so 5 results make the list. The adjusted STDev was 5.82, so the increment is 6. Therefore, Tier 3 runs from 5 to 11, Tier 2 is 12 to 18, and Tier 1 is 19 and over.

Deck Name	Total #	Total %
Tier 1
Yawgmoth	47	12.08
Amulet Titan	45	11.57
Rakdos Scam	27	6.94
Izzet Prowess	23	5.91
Counter Cat	20	5.14
Tier 2
UW Control	17	4.37
Living End	15	3.86
Izzet Murktide	13	3.34
Goryo Blink	13	3.34
Coffers	12	3.08
Jund Creativity	12	3.08
Tier 3
4-C Creativity	10	2.57
Hammer Time	9	2.31
Wrenn White Blue	8	2.06
Bant Rhinos	8	2.06
Domain Rhinos	7	1.80
Jund Saga	7	1.80
Merfolk	7	1.80
Burn	6	1.54
Mono-Green Tron	6	1.54
Wishshift	5	1.28
Rack Scam	5	1.28
Domain Murktide	5	1.28

While the composition of Tier 1 is pretty similar to MTGO, there's vastly different conclusions to be drawn. Looking at MTGO would lead one to believe that Scam is oppressing Modern, while the paper results make it look like a fairly middling deck. The average power statistics are revealing and point to the paper conclusion being more accurate, as will be seen below.

May Power Metagame

Tracking the metagame in terms of population is standard practice. But how do results actually factor in? Better decks should also have better results. In an effort to measure this, I use a power ranking system in addition to the prevalence list. By doing so, I measure the relative strengths of each deck within the metagame so that a deck that just squeaks into Top 32 isn't valued the same as one that Top 8's. This better reflects metagame potential.

For the MTGO data, points are awarded based on the population of the event. Preliminaries award points based on record (1 for 3 wins, 2 for 4 wins, 3 for 5), and Challenges are scored 3 points for the Top 8, 2 for Top 16, and 1 for Top 32. If I can find them, non-Wizards events will be awarded points the same as Challenges or Preliminaries depending on what the event in question reports/behaves like. Super Qualifiers and similar higher-level events get an extra point and so do other events if they’re over 200 players, with a fifth point for going over 400 players.

Due to paper reporting being inconsistent and frequently full of data gaps compared to MTGO, its points work differently. I award points based on the size of the tournament rather than placement. For events with no reported starting population or up to 32 players, one point is awarded to every deck. Events with 33 players up to 128 players get two points. From 129 players up to 512 players get three. Above 512 is four points, and five points will be reserved for Modern Pro Tours. When paper reports more than the Top 8, which is rare, I take all the decks with a winning record or tied for Top 32, whichever is pertinent.

The MTGO Power Tiers

As with the population numbers, total points are down, from 2770 to 2499. The adjusted average points were 27.11, therefore 27 points made Tier 3. The STDev was 56.83, so add 57 to the starting point, and Tier 3 runs to 84 points. Tier 2 starts with 85 points and runs to 142. Tier 1 requires at least 143 points. Bant Rhinos took Mono-Green Tron's place on the power tier.

Deck Name	Total Points	Total %
Tier 1
Rakdos Scam	384	15.37
Yawgmoth	273	10.92
Living End	249	9.96
Amulet Titan	235	9.40
Counter Cat	208	8.32
Izzet Prowess	169	6.76
Tier 2
UW Control	106	4.24
Tier 3
Izzet Murktide	71	2.84
Jund Creativity	65	2.60
Burn	59	2.36
Hardened Scales	55	2.20
Rack Scam	52	2.08
Jund Saga	50	2.00
Coffers	44	1.76
Temur Prowess	38	1.52
Goryo Blink	33	1.32
Mono-White Emeria	31	1.24
Bant Rhinos	27	1.08

There's a lot of movement inside the tiers but no movement between them. Given the huge gaps that's not surprising. I've noticed that big gaps are quite common on MTGO but not so much in paper. While the usual suspects of groupthink and small playerbase could be the entire problem, I do wonder how much to blame the rental services. Players get a deck for a month and play them constantly, far more than you usually see in paper. I wonder how much they're locked into their deck for the whole month and if that is a reason you see huge swings and gaps all the time in the MTGO data.

The Paper Power Tiers

As with the population, total paper points are down from 1163 to 645. The adjusted average points were 7.52, setting the cutoff at 8 points. The STDev was 9.98, thus add 10 to the starting point and Tier 3 runs to 18 points. Tier 2 starts with 19 points and runs to 29. Tier 1 requires at least 30 points. There's a lot of movement between the tiers, while Domain Murktide fell off the list.

Deck Name	Total Points	Total %
Tier 1
Yawgmoth	85	13.18
Amulet Titan	79	12.25
Rakdos Scam	41	6.36
Izzet Prowess	37	5.74
UW Control	37	5.74
Counter Cat	35	5.43
Tier 2
Living End	26	4.03
Jund Creativity	24	3.72
Izzet Murktide	21	3.26
Goryo Blink	21	3.26
Tier 3
Coffers	16	2.48
Wrenn White Blue	16	2.48
Bant Rhinos	16	2.48
Hammer Time	15	2.33
4-C Creativity	13	2.02
Domain Rhinos	11	1.70
Jund Saga	11	1.70
Merfolk	10	1.55
Mono-Green Tron	10	1.55
Burn	9	1.39
Rack Scam	9	1.39
Wishshift	8	1.24

Composite Metagame

That's a lot of data, but what does it all mean? When Modern Nexus was first started, we had a statistical method to combine the MTGO and paper data, but the math of that system doesn't work without big paper events. I tried. Instead, I'm using an averaging system to combine the data. I take the MTGO results and average the tier, then separately average the paper results, then average the paper and MTGO results together for final tier placement.

This generates a lot of partial Tiers. That's not a bug, but a feature. The nuance separates the solidly Tiered decks from the more flexible ones and shows the true relative power differences between the decks. Every deck in the paper and MTGO results is on the table, and when they don't appear in a given category, they're marked N/A. This is treated as a 4 for averaging purposes.

Deck Name	MTGO Pop Tier	MTGO Power Tier	MTGO Average Tier	Paper Pop Tier	Paper Power Tier	Paper Average Tier	Composite Tier
Rakdos Scam	1	1	1	1	1	1	1.00
Yawgmoth	1	1	1	1	1	1	1.00
Amulet Titan	1	1	1	1	1	1	1.00
Counter Cat	1	1	1	1	1	1	1.00
Izzet Prowess	1	1	1	1	1	1	1.00
Living End	1	1	1	2	2	2	1.50
UW Control	2	2	2	2	1	1.5	1.75
Izzet Murktide	3	3	3	2	2	2	2.50
Jund Creativity	3	3	3	2	2	2	2.50
Goryo Blink	3	3	3	2	2	2	2.50
Coffers	3	3	3	2	3	2.5	2.75
Burn	3	3	3	3	3	3	3.00
Rack Scam	3	3	3	3	3	3	3.00
Jund Saga	3	3	3	3	3	3	3.00
Mono-Green Tron	3	N/A	3.5	3	3	3	3.25
Bant Rhinos	N/A	3	3.5	3	3	3	3.25
Hardened Scales	3	3	3	N/A	N/A	N/A	3.50
Temur Prowess	3	3	3	N/A	N/A	N/A	3.50
Mono White Emeria	3	3	3	N/A	N/A	N/A	3.50
4-Color Creativity	N/A	N/A	N/A	3	3	3	3.50
Hammer Time	N/A	N/A	N/A	3	3	3	3.50
Wrenn White Blue	N/A	N/A	N/A	3	3	3	3.50
Domain Rhinos	N/A	N/A	N/A	3	3	3	3.50
Merfolk	N/A	N/A	N/A	3	3	3	3.50
Wishshift	N/A	N/A	N/A	3	3	3	3.50
Domain Murktide	N/A	N/A	N/A	3	N/A	3.5	3.75

Average Power Rankings

Finally, we come to the average power rankings. These are found by taking the total points earned and dividing them by total decks, to measure points per deck. I use this to measure strength vs. popularity. Measuring deck strength is hard. There is no Wins-Above-Replacement metric for Magic, and I'm not certain that one could be credibly devised. The game is too complex, and even then, power is very contextual.

Using the power rankings helps to show how justified a deck’s popularity is. However, more popular decks will still necessarily earn a lot of points. Therefore, the top tier doesn't move much between population and power and obscures whether its decks really earned their position. 

This is where the averaging comes in. Decks that earn a lot of points because they get a lot of results will do worse than decks that win more events, indicating which deck actually performs better.

A higher average indicates lots of high finishes, whereas low averages result from mediocre performances and a high population. Lower-tier decks typically do very well here, likely due to their pilots being enthusiasts. Bear this in mind and be careful about reading too much into these results. However, as a general rule, decks that place above the baseline average are over-performing, and vice versa.

How far above or below that average a deck sits justifies its position on the power tiers. Decks well above baseline are undervalued, while decks well below baseline are very popular, but aren't necessarily good.

The Real Story

When considering the average points, the key is looking at how far off a deck is from the Baseline stat (the overall average of points/population). The closer a deck’s performance to the Baseline, the more likely it is to be performing close to its "true" potential.

A deck that is exactly average would therefore perform exactly as well as expected. The greater the deviation from the average, the more a deck under or over-performs. On the low end, a deck’s placing was mainly due to population rather than power, which suggests it’s overrated. A high-scoring deck is the opposite of this.

I'll begin with the averages for MTGO: 

Deck Name	Average Points	Power Tier
Jund Saga	2.08	3
Bant Rhinos	2.08	3
Hardened Scales	2.04	3
Rack Scam	2.00	3
Mono-White Emeria	1.94	3
Yawgmoth	1.88	1
Amulet Titan	1.84	1
Coffers	1.83	3
Living End	1.77	1
Counter Cat	1.75	1
Rakdos Scam	1.74	1
Izzet Prowess	1.74	1
Goryo Blink	1.74	3
UW Control	1.71	2
Tier 1	1.68
Burn	1.59	3
Izzet Murktide	1.58	3
Temur Prowess	1.58	3
Jund Creativity	1.55	3

Yawgmoth held off Amulet Titan to be April's MTGO Deck of the Month. Scam was the worst performer of the Tier 1 decks.

Now the paper averages:

Deck Name	Average Points	Power Tier
UW Control	2.18	1
Jund Creativity	2.00	2
Wrenn White Blue	2.00	3
Bant Rhinos	2.00	3
Yawgmoth	1.81	1
Rack Scam	1.80	3
Amulet Titan	1.76	1
Counter Cat	1.75	1
Living End	1.73	2
Hammer Time	1.67	3
Mono-Green Tron	1.67	3
Izzet Prowess	1.61	1
Izzet Murktide	1.61	2
Goryo Blink	1.61	2
Wishshift	1.60	3
Domain Rhinos	1.57	3
Jund Saga	1.57	3
Baseline	1.54
Rakdos Scam	1.52	1
Burn	1.50	3
Merfolk	1.43	3
Coffers	1.33	3
4-C Creativity	1.30	3

In a shocker, UW Control wins Deck of the Month in paper and is the overall best performer. Never would have called that, but there it is. Meanwhile, Scam underperformed in paper.

Analysis

Actual change in this metagame is on hold until after MH3 releases. MTGO's players aren't putting much effort into tuning their decks and instead went back to old favorites to just wait for the new cards. I say this confidently because Living End did quite well all month, yet the amount of graveyard hate didn't measurably change online. It looked to me like players were more interested in maindeck tweaks than sideboard changes, which strongly points to everyone coasting to the finish of this particular metagame.

The New Horizon

With that in mind, the long-term impact of MH3 remains unknown. There is nothing obviously format-redefining like the evoke elementals from MH2 or obviously broken like Hogaak, Arisen Necropolis in MH1. There are certainly many powerful cards and others with a ton of potential, but nothing that will obviously heavily change Modern forever. Things are going to change, but not as much as before.

In the short term, Affinity, Elves, Colorless Eldrazi, and Energy decks will be everywhere once MH3 releases. Those decks all received obvious pieces and players will be eager to build around them. Probably too eager, and I'd expect that many of these decks will fail quickly as players discover that many of the cards are overhyped. My bet for greatest disappointment is Ugin's Labyrinth, as the price to make it work is filling a deck with 7-drops which will just clog the hand most of the time. Remember, the odds of opening a given 4-of is under 40%.

The other area to watch is graveyard decks. There will be many players looking for a combo-win with Buried Alive for Thassa's Oracle and Leveler into Victimize for the win. Leveler has already seen one speculative price spike. I have doubts that this can be sustained, but we'll wait and see.

Financial Considerations

As with any new set release, the prerelease prices reflect hype and excitement rather than any sustained demand. Those with MH3 inventory should look to move their stock quickly while those looking to buy should hold off until after the release day to see where prices stabilize. I'd freely speculate on selling the hyped reprinted cards, as those with high prices will not hold once the new supply becomes available.