By Jason Waddell

The Emperor. The God of War. The Incredible Huk. The Game Genie Terran. The Gracken.

How good are these players, really? As a relative newcomer to the StarCraft 2 scene, it can be incredibly difficult for me to gauge how well the different pro players stack up against each other. For the most part I'm left with the overall sense of reverence that the community uses to discuss each player. Tasteless and Artosis wax poetic about MVP and Nestea. Day9 calls Select a baller, but other players are sick nerd-ballers. One day Wolf is claiming that MC's style has been figured out and the Kratoss Protoss isn't good enough for Code S anymore. The next day, MC is in the Code S Round of 16. Nobody can seem to agree on just how wide the skill gap is between the Koreans and the foreigners. Korean Top 6 in Anaheim; Korean Top 6 in Raleigh; Westerner wins Orlando.

It's enough to make your head spin.

Most importantly, how does this all relate to Providence? This Season's MLG finishes have been incredibly chaotic, with different Koreans flying over for each Event. There were groups of death, deceptively easy Pools, intimidating brackets, and more upsets than you can shake a stick at.

So, I did what any good statistician would do: gathered some data and fit a model. I recorded the result of every 2011 MLG Championship Bracket match between players who’d started in Pool Play, and added in the data from any competitor that had at least 5 matches against Pool Play players. Note that only matches against players in this category were recorded. For example, Slush's Orlando match against Inori was not included. The resulting data set was a list of 411 match results played between 52 players across the 2011 Season.

Next, I fit a statistical model—transformed Bradley Terry, for those so inclined (warning: link contains math)—to the data to give an estimated Ability Score for each player. Basically, youhand the model a pool of results. Its end goal is to produce an ability score for each player in that data set, such that for players A and B, the probability that A beats B is A / (A + B). So, for example, with Mvp, it knows he beat Moonan, Select, Machine and Boxer 2-0, and that he beat Kiwikaki, Ganzi, DongRaeGu and MMA 2-1. So, he should be a lot better than those first four players, and about twice as good as those last four. The model makes these evaluations for all players simultaneously, so all the players are correlated somehow. It knows how good Kiwikaki is, and MMA, etc., so all those things factor into MVP's estimate. Change one result and everyone else's estimate changes too, even if only slightly.

A couple of notes on the model: all it looks at are game results. It only processes whether you won or lost each game, and who your opponent was for that game. So, a win over MMA will affect a player's Skill Score much more dramatically than a win against Gretorp. The model looks at the number of games won and lost, rather than the pure match result (a 4-3 extended series win counts differently than a 4-0 win). Lastly, the order of the games has no effect, so a result in Columbus counts as much as a result in Orlando. Note that the model doesn't care about the race of the opponent or what stage of the tournament the games occurred in.

The cool thing about the model is that it gives an easily interpretable Skill Score. The multiplicative average of all the scores is 1, so a player with a Skill Score of 2 is roughly twice as good as the average player in the group. Also, If theoretical Player A has a score of 3, and Player B a score of 2, then the expected probability that A beats B in a game is 3 / (3 + 2), or 60% (again, races are not accounted for).

As with any small sample sizes, some of these estimates could be extremely far off the mark, especially in Tier 6 (see below). If a player only showed up to a single Event and underperformed for whatever reason, the estimate might not be representative of their true ability. The player estimates are simply a function of results produced on the MLG stage, and give a rough idea of where the players stand in comparison to the field. Inevitably some estimates will be off.

For example, MarineKing's estimate is based solely off of Orlando's performance. All the model knows is that he lost to 2-1 to Drewbie and MC, was swept by Huk, swept Boxer, Haypro and TLO, and beat Idra 2–1. The estimate based on these results has him ranked behind Rain and Naniwa. Is that an accurate representation of his skill? Was the loss to Drewbie just a fluke? That's all up for debate! The interesting part of the model is that it's unbiased by any outside opinions. Nothing but results factor into the estimate, regardless of the player's popularity or the community's opinion.

So how do the players stack up?

As noted above, the more games a player has in the sample, the more accurate their Skill Score estimate is. But even with some relatively small sample sizes, we see that the players' Skill Scores cluster into rough tiers.

Tier 1 – "The MVP Tier" – MVP (10.12)

MVP currently stands alone, in his own stratosphere, and his off-the-charts rating nearly forced the use of log-scale and scientific notation. His power ranking rating shows that he is an order of magnitude better than the average Pool Play opponent, and singlehandedly dispels the notion of a StarCraft 2 skill-cap What exactly does a Skill Score of 10 indicate? It means that MVP is expected to win over 90% of his games against the average player from the list. Even against such elites as Coca and Bomber, MVP is expected to win around 60% of the games, making him a heavy favorite in any Best of 3 setting.

In Anaheim, MVP looked like a god among men, amassing a 16-4 game record and leaving the likes of Ganzi, DongRaeGu, Boxer and MMA in his wake. He's back for more in Providence, and given his resume it's hard not to consider him the outright favorite, even from the Open Bracket. The seeded players are crossing their fingers in hopes that MVP doesn't find his way to their portion of the bracket, but with such a stacked Open Bracket it will be hard for them to dodge such bullets entirely. Did we mention that MVP's teammate Nestea will be making his MLG debut?

Tier 2 – "Code S" – Coca (7.49), Ganzi (7.29), MMA (6.83), Bomber (5.88), TheStC (4.58)

The second tier of the Pro Circuit Power Rankings is host to a fearsome fivesome of top Korean competitors. Their match records are littered with 2-0 sweeps over the MLG grinders, with rare losses to only the highest quality of opponents. These players, with one exception, are a selection of the GSL Code S mainstays. Coca, Ganzi and MMA were all group winners in the recent Code S Ro32. Bomber, by contrast, has been on a bit of a slide. After starting his MLG career on a 12-match winning streak that included a nearly flawless Raleigh Championship run, Bomber dropped three consecutive sets to close out Orlando. Upon his return to Korea, he dropped 0-2 out of the Code S group stage with losses to Gumiho and DongRaeGu. Slump or not, Bomber is still a force to be reckoned with, and has the advantage of already being seeded into Providence's Championship Bracket.

The last player in this tier, TheStC, has been notably absent from the highest Korean league. He currently resides in GSL Code A, and made an emphatic MLG debut in Orlando. There, he crushed through the Open Bracket, won his pool with a 10-1 game record and defeated the then-reigning Champion, Boxer, on his way to a 3rd Place finish. TheStC disappointingly fell a single game short of an automatic Code S slot when he lost to MC in Orlando, but given the prowess he's displayed, he might well be on his way to becoming one of the stars of the scene.

Tier 3 – "The Up and Downs" – MC (3.44), Puma (3.42), Noblesse (3.29), Polt (3.16), DongRaeGu (2.95), Losira (2.92)

The third tier stands as an example of how brutally competitive the current scene is. This group contains a handful of former GSL tournament winners and finalists, all of whom are struggling to maintain a seat in Code S. Losira, DongRaeGu and Polt are on their way to Code A after poor group showings, Noblesse was knocked from Code S last Season, former Champion MC was knocked temporarily to Code B, and Puma has never managed to make it through the GSL qualifiers. Despite facing struggles in Korea, these players mop the floor with all but the best foreigners. According to the statistical model, the Tier 3 players are all about three times better than the average MLG Pool Play competitor. Shrinking or not, the gulf in skill between Korea and the western world is truly staggering. We’re 12 players into the power ranking, and all 12 hail from South Korea.

Tier 4 – "Code A?" – Sase (2.50), Thorzain (2.28), Nada (2.18), Huk (2.10)

The labels become a little fuzzy around the fourth tier. Sase and Thorzain have each put in brief stints in Code A, and Nada is currently a Code A competitor. Huk, however, is a Code S mainstay. More than any other player, Huk's ranking suffers from the fact that games from the entire 2011 Season are included. He started the Season by placing 4th in his group in Dallas, transitioned to the unofficial title of "Best Foreigner" in Anaheim and Raleigh, then claimed the official title of Champion with a masterful display in Orlando. The fact that his early Season performances weigh him down to the fourth tier is a symbol of how far he's progressed over the year. In Dallas, Huk was swept by Socke and Kiwikaki. In Orlando, Huk swept July, Tyler, MarineKing and MC (in the Finals). It’s night and day.

Tier 5 – Idra (1.70), Boxer (1.63), Rain (1.62), Naniwa (1.62), HongUn (1.58), MarineKing (1.56), July (1.55), DeMuslim (1.51)

The fifth tier of the rankings is home to players who have shown flashes of brilliance, but whose styles are somewhat uneven. Idra has a terrifying ZvT matchup and ruthless Mutalisk harassment, but is less refined against Zerg and Protoss. HongUn pushes the boundaries with unconventional play, which is at times inspiring and at times simply suboptimal. July's raw aggression is peerless when executed properly, but can look shockingly tame at times (such as in the recent GSL November group stages). MarineKing, as suggested by the name, gives world-class displays of marine control and bio splits, but he’s known as the King of 2nd Place for a reason. Many of the most promising players fall into this skill range, and with further polish, their unique skill sets and approaches to the game could eventually define the competitive scene.

Tier 6 – The Rest

It is perhaps unfair to lump the remaining players into one bin, as a pretty large skill gap exists even in the lower half of the rankings. For example, LiquidHero's rating of 1.265 is more than 10 times higher than Gretorp's rating of 0.123. This last group is home to many players who have lost or are in the process of losing their favorable seedings. Within the community, players like InControl and Drewbie have been the poster children of players benefiting from a legacy seeding despite not producing any notable recent results. And it seems that the real under-performer of the season has been Tyler. Of his 25 matches included in the dataset, Tyler has won only 3. He logged 17 0-2 match losses, and an overall 11-46 game record. There's nowhere to go but up; luckily, with a dramatic increase in practice hours logged over the last few weeks, Tyler appears to have rediscovered the spark that fueled his success in the Brood War era.

On the whole, the South Koreans still hold a massive competitive edge over the rest of the world. Foreign stars like Huk, Idra, Sase and Thorzain, however, are becoming legitimate contenders, and might well be on their way to reaching the levels of top Koreans like MVP and Nestea. Outside of the “Best of the West” players, the foreigners still appear to be light years behind the Koreans, at least on paper. Players like Puma and TheStC struggle to break through the Korean ranks, but still find themselves heavily favored when facing foreign opposition.

It’s going to be incredibly interesting, to say the least, to see how all of this will play out in Providence. Whatever happens, we can be assured of two things. One, the best player will win. Two, no matter who wins, the road to that victory will be paved with some of the best StarCraft 2 the world has ever seen.