How big a factor does luck play in Magic? How much of is Magic a game of skill? Does the luck have greater impact in Limited or Constructed? I’ve been thinking about this a lot lately, and I have tried to look for, and at, actual data. I think luck has an impact – but maybe not in the areas one would expect.
The $10,000 Topdeck
The first thing everyone thinks of is the amazing topdeck that wins the match. It happens all the time – right? If you’ve been around a while, you remember when Kai ripped the Morphling to win Pro Tour New Orleans. If not, you probably heard about Craig Jones pulling Lightning Helix off the top to steal the semi-final match from Oliver Ruel. It happens all the time, right?
Actually, it doesn’t.
There were a couple dozen Top 8 games played before the final participants in Pro Tour Hawaii went back to the beach. There have been hundreds and hundreds of games played since Kai’s Morphling saved him in the Big Easy. Other than those two, very few had amazing topdeck finishes. It’s just that people remember the dramatic finishes rather than the typical ones. Dramatic things stand out because they are unusual.
I live on a rural highway – and a couple hundred cars a day drive past my house. I have seen dozens, maybe hundreds, in the past week. I remember exactly two. I remember the first because I was shoveling snow on a wet, sloppy morning and I watched it spin out, crash backwards through the fence, and bounce into the field. I remember the other because, later in the morning, it slid about 150 yards with wheels locked and just missed colliding with a car stopped on the shoulder.
People don’t remember the typical events; they remember the unusual and atypical. Lawyers have a saying for this – “good cases make bad laws.” Good cases are the strange and exceptional – the things that happen once in a blue moon. When people make laws based on these cases, the laws are generally bad. They may deal with the exceptional case, but those laws generally have an adverse effect on more typical situations.
An example. Once in a great while, someone falls off a ten-story building, lands in a tree and walks away unhurt. It makes the news. There’s the atypical result – and when legislators fail to realize how unlikely that is, then they pass a law abolishing stairways and elevators and requiring that we jump instead.
Of course it’s an extreme example, but I have years of experience with legislation and legislators – and hundreds and hundreds of examples of bad laws passed to deal with unlikely and exceptional circumstances. But I digress.
The point is that it is the dramatic stuff that people remember, and the good stories that get retold. Memory, plus repetition, makes people think uncommon things are common. Here’s a quick example: which do you think is a more common occurrence, someone being murdered by a bomb, or being electrocuted by their air conditioner?
The US government publishes statistics on this sort of thing. According to the most recent FBI crime statistics and the US Consumer Products Safety Commission, ten times as many people died by being electrocuted by air conditioners than were murdered via explosives.
It’s just that no one would watch an episode of the Sopranos where someone gets zapped by their AC.
The point is that, while we all remember the lucky topdeck that won an otherwise unwinnable game, having a game decided by gradual shifts in tempo and board position is infinitely more likely. I watched a lot of replays of online games – mine, and the games of others – to check exactly that point, and I’m certain of it.
Moreover, even when Kai or Craig rips that card it was hardly luck that put them into a position to win. It was generally careful play and resource management that allowed them to hang on long enough to get to that point. Craig Jones carefully managed his life total and mana, including casting Char at Antoine’s head the turn before, so that he was in position for a Lightning Helix to win the game. Drawing the Helix may have been “luck,” but that sort of “luck” only favors the players good enough to get into that situation. [Agreed on all counts. – Craig, often berated for winning with the “lucky” Skizzik off the top.]
The Metagame
In Constructed, it isn’t in drawing cards where luck seems to have a big impact. It is in drawing opponents. In any tournament, in the Swiss rounds, the system tries to pair people with similar records. After that, however, pairings are random – and that is where luck sets in.
Every deck has good matchups and bad matchups. To some extent, it is pure luck that decides whether you will be randomly paired against a deck you can crush, or one that crushes you. Likewise, in an open event in the early rounds, it’s pure luck whether you are paired with a good player or a total scrub. Random eight-man events on MTGO are a great example of this – you could get paired against Nick Eisel, or me, or the rare drafter with a 1430 rating in any draft, and be playing either Mike Flores or a little kid with Craw Wurm.dec in constructed.
There is no way to control your pairing – and your pairings may have more to do with your chances of winning than anything else. In a multi-round tournament, bad players will tend to be eliminated from contention, but nothing will prevent you from facing bad matchups if the metagame is at all healthy.
I’ve known that for years – but I still remember the day it really hit home. It was years ago, at an Extended Pro Tour Qualifier. I was running GB Survival, and if I won the match, I could draw into the Top 8. Lose, and I would have to win out. Pairings went up, and I drew an opponent I knew was running Sligh. I could probably win that matchup if I mulliganed to four and my opponent got a god draw. Then they repaired, and my new opponent was running Pandeburst. Pandeburst was my worst matchup, and I faced it again I the final round.
That repairing took what was a matchup I won 99% of the time and replaced it with a matchup I would be lucky to win once in five.
Luck of the draw.
That happens at every tournament. The question is whether that luck really drives anything, or whether we just remember it. Do lucky matchups predict success, or do good players play through bad matchups and still make Top 8? Let’s look at some data.
At Pro Tour Hawaii, the only deck that made T8 in multiples (aside from the two very different Zoo builds) was Owling Mine. By that criterion, Owling Mine was a very successful deck. However, the deck has a huge weakness. Here are some excerpts from the coverage.
“Zoo [is] basically an impossible matchup for the Owl deck.”
“Before the match, Antoine explained that this matchup was so terrible that he just had to give up on it, and included zero cards in his sideboard to try to help out against the aggro matchup. This strategy worked for him in the Swiss, letting him stack his sideboard with cards that chewed through the slower decks in the field, but now his owls were being stalked by craven predators bearing flamethrowers. Ruel could have not shown up for this match and probably stood a better chance of winning it.”
The Owling Mine deck runs symmetrical card drawing engines, like Howling Mine, and tries to use Boomerang effects to bounce lands and large creatures to clog up an opponents hands. Zoo decks drop threats on turns through game, and don’t need many lands to function. Looking at The Ferrett wonderful StarCityGames matchup analysis, it is really clear how badly Owling Mine does against fast beatdown:
Against Boros: 16.67% wins
Against Gruul Beats: 33.33% wins
Against Mono-Red Beats: 0% wins
Against Ninja Stompy: 20% wins
Against Zoo: 12.5 % wins
On the flip side, Owling Mine had some great matchups. It won almost 90% of the dozen odd matches against Orzhov Aggro. It is pretty clear that the deck has some really good matchups, and some really bad ones.
The question is, did the Owling Mine decks that ended up in the Top 8 have better luck dodging bad matchups than those that did not make Top 8? In the event, fifteen players arrived with Howling Mine decks. Let’s look at how some of them did. (Unfortunately the decklists for some of the players are not included in the lists on either the Wizards site or in the StarCityGames database at the time of writing, so some of the matchups are blanks.)
Raphael Levy:
Round 1: beat Greater Good
Round 2: lost to Zoo
Round 3: lost to Red Deck Wins
Round 4 & 5: lost, but opponents’ decklist not available
(dropped)
Jeremy Berthoux:
Round 1: beat Jushi Control
Round 2: lost, but opponent’s decklist is not available
Round 3:
Round 4: lost to Boros
Round 5: lost to Heartbeat
(dropped)
Itaru Ishida:
Round 1: lost to RG Beats
Round 2: lost to Zoo
Round 3: lost, but opponent’s decklist is not available
Round 4: lost to UWB control
Round 5: won the mirror match
Round 6: beat Orzhov Descendent
Round 7: lost to BWG control
Round 8: lost to Eminent Domain
Tomahiro Kaji:
Round 1: beat UWR Firemane Angel
Round 2: lost the mirror
Round 3 & 4: won, but opponent’s decklist is not available
Round 5: beat Orzhov Descendent
Round 6: lost to Zoo
Round 7: beat Debtor’s Knell
Round 8: lost to Debtor’s Knell
Round 9: lost to Zoo
Round 10: beat Orzhov Descendent
Round 11: beat UR Control
Round 12: lost to Greater Gifts
Round 13: beat Enduring Ideal
Round 14: won, but opponent’s decklist is not available
Round 15: beat Izzetron
Round 16: won, decklist not available
JP Brichta:
Round 1: beat Zoo
Round 2: beat GBW control
Round 3& 4: lost to Izzetron
Round 5: lost to GBW beats
Round 6: lost to Zoo
(dropped)
Rashad Miller:
Round 1: won, decklist not available
Round 2: beat Greater Good
Round 3: beat BWG control
Round 4: lost to Greater Gifts
Round 5: beat GBW “The Rock”
Round 6: beat Orzhov aggro
Round 7: lost to Greater Good
Round 8: drew (mirror match)
Round 9: lost to Roxodon Hierarch
Round 10: beat Orzhov aggro
Round 11: lost to Ghazi-Glare
Round 12: lost to Eminent Domain
Round 13: beat UBr Tron
Round 14: lost to BW Control
Round 15: lost to Zoo
(dropped)
Antoine Ruel (T8):
Round 1: beat Eminent Domain
Round 2: beat BWG control
Round 3: lost, decklists not available
Round 4: beat Orzhov aggro
Round 5: beat Zoo
Round 6: beat Izzetron
Round 7: beat BWG control
Round 8: beat Heartbeat
Round 9: beat GBW aggro
Round 10: beat Izzetron
Round 11: beat Orzhov agro
Round 12: lost to Zoo
Round 13: lost to Orzhov aggro
Round 14: lost to Tallowisp (Ghost Dad)
Round 15: beat Izzetron
Round 16: beat Greater Gifts
Taigo Chan (T8):
Round 1: beat Gifts Control
Round 2: lost to GBW control
Round 3: lost to Ghazi-Glare
Round 4: beat Orzhov agro
Round 5: beat Greater Good
Round 6: beat GBW aggro
Round 7: lost to GW Wrath
Round 8: drew with Rashad Miller (mirror match)
Round 9: beat Orzhov agro
Round 10: drew with Greater Good
Round 11: beat BWG control
Round 12: beat BW aggro
Round 13: beat Debtor’s Knell
Round 14: won, decklists not available
Round 15: beat Greater Good
Round 16: beat Ghost Husk
So what does the evidence show? Some players, like Raphael Levy and JP Brichta, clearly had bad matchups. Levy beat the controlling Greater Good, but lost to Zoo and Red beats. Brichta played Zoo twice and GWB aggro once, but he also lost twice to Izzetron, which was, overall a 70-30 matchup in favor of Owling Mine. Both those players failed to finish Day 1.
By comparison, Antoine Ruel also faced Zoo in the early rounds, and a couple of other aggro decks. The difference is that he won those matches. He also won all his matches against Izzetron. Taigo Chan faced a few control matchups, dodged Zoo, and split with other decks. So far, it is inconclusive. Time to quantify.
I am going to try to divide the matchups into categories based on the won-loss percentages for each category. For example, Owling Mine beat Orzhov Aggro 88.9% of the time, so that matchup falls into category one (decks against which Owling Mine has a win percentages of 80% or better.) This is still all based on The Ferrett analysis, but I am going to exclude any matchup with insufficient data (e.g. Owling Mine met UWB Firemane Angel exactly once, and won, but I am not going to say that makes it a category one matchup.) Even with exclusions, the numbers involved (in terms of matches, etc.) are small enough that this is probably not statistically valid, but it should still give us some indications of what happened.
| Player | 100-80% | 80-60% | 60-40% | 40-20% | 20-0% |
| Antoine Ruel | 3 | 6 | 1 | 0 | 1 |
| Taigo Chan | 3 | 4 | 4 | 1 | 0 |
| Rashad Miller | 1 | 2 | 5 | 0 | 0 |
| Raphael Levy | 0 | 0 | 1 | 0 | 2 |
| Jeremy Berthoux | 0 | 1 | 0 | 1 | 1 |
| Itaru Ishida | 1 | 2 | 1 | 1 | 1 |
| Tomahiro Kaji | 1 | 2 | 1 | 0 | 1 |
| JP Brichta | 0 | 3 | 0 | 0 | 2 |
Let’s take this one step further. We can average the matchup percentages for each player. For Antoine Ruel, (no, Raphael Levy – the numbers are easier) that means averaging one match with a 50% win percentage and two with a 10% win percentage – giving an average of expected win percentage for opposing matchups of about 23%. Doing that, here are the results:
Antoine Ruel: 54 % Day 1, 58% overall
Taigo Chan: 61% Day 1, 69% overall
Rashad Miller: 73% Day 1, 58% overall
Tomahiro Kaji: 78% Day 1, 58% overall
Itaru Ishida: 58% Day 1, no Day 2
JP Brichta: 46% Day 1, no Day 2
Raphael Levy: 23% Day 1, no Day 2
Jeremy Berthoux: 36% Day 1, no Day 2
I have to repeat the caveat again – the sample base is small, and the data is somewhat circular. Even so, it seems pretty clear that drawing or not drawing favorable matchups can have a big impact on performance. Three players had an expected win percentage well below 50% – meaning that they faced mainly unfavorable matchups, and they dropped. All of those making Day 2 had, on average, favorable matchups. Only one player, Ishida, had a favorable average but did not make Day 2 – and Ishida’s matchups were all over the board.
On Day 2, the two players having more favorable matchups – reflected by their overall percentages being higher than their Day 1 percentages – made Top 8. Those players who showed marked declines in their overall percentages did not make Top 8.
I think the data shows that the random pairings in Swiss rounds are one area where luck really does play a roll, at least in Constructed tournaments.
PRJ