statsmash

Predictions and Projections for Albion 4


Tweet

Something fun Elo Rankings (like the one the site hosts hint hint) lets you do that no other traditional ranking system is estimate win probalities. Indeed, it’s baked right into the algorithm with how you update Elo values after a result.

So, here are projected Albion 4 results, the last PGR tournament of the season. Methodology will be below

Player Top 8 Top 4 Top 3 Top 2 Top 1
Dabuz 98.06% 91.4% 85.18% 74.48% 53.86%
MVD 95.72% 79.70% 68.3% 48.54% 20.18 %
Glutonny 92.22% 75.16% 62.38% 41.12% 16.34%
Raito 81.86% 54.88% 36.54% 19.94% 6.54%
Mr.R 70.94% 38.42% 22.72% 9.34% 2.3%
Leffen 24.6% 7.46% 2.84% 1.04% 0.18%
Meru 38.02% 8.74% 4.3% 1.13% 0.16%
Space 55.12% 11.5% 5.08% 1.44% 0.14%
Light 32.5% 7.3% 3.2% 0.78% 0.12%
quiK 49.68% 8.5% 3.2% 0.8% .08%

Dabuz is clear and away the favorite for the tournament, with an insane 91% chance to make top 4 and a 53% chance to get the chip. However, remember that 53% isn’t that likely. It’s a little better than a coinflip, and you certainly wouldn’t call a coinflip certain victory!

So the next time you call it and just believe that MkLeo or Tweek will take a tournament; remember, 50% is insanely high, but it’s still very likely one of them doesn’t win, even if they are clearly the best players in the world and have the plurality of win probability.

We can see where Elo rankings and seeding disagree with Leffen, Meru, and Space. Notice that Leffen actually has a lower chance to make it top 8, top 4, top 3, top 2 than Meru, but a higher chance to win the whole thing. By Elo, Leffen is considered the better player, but the seeding makes it much easier for Meru to make it higher.

But, when you get to the finals, you’re gonna face Dabuz or MVD, and then no amount of seeding will make it easier.

Leffen, really?

Leffen has a higher elo primarily because he has traveled more. Tournaments like Genesis are given much more weight than a European national in the algorithm, so his wins there elevate him above much of his European peers.

Is it deserved? Perhaps not. His Elo score doesn’t reflect his hiatus, and it is probably underrating the European players.

Who knows, maybe he won’t even enter.

Weird Worlds

Note that Leffen, Meru, and friends do not have a 0% to win the whole thing. Here’s a scenario where the Big Leff himself gets a W in Ultimate

Place:  1 ['Leffen'] 
Place: 2 ['Glutonny']
Place: 3 ['Dabuz']
Place: 4 ['Greil']
Place: 6 ['Space', 'Raito']
Place: 8 ['MVD', 'Yetey']
Place: 12 ['Mr.R', 'quiK', 'Meru', 'Jeda']
Place: 16 ['Light', 'RobinGG', 'iStudying', 'Loading...']
Place: 24 ['MVL', 'S1', 'Flow', 'PEW', 'Scr7', 'Tru4', 'Supahsemmie', 'Oryon']
Place: 32 ['Vreyvus', 'Young Eevey', 'Patrino', 'Whoophee', 'ShuC', 'cyve', 'Azrael', 'Elexiao']

Methodology

Too many players named Light

WHY ARE THERE SO MANY PLAYERS NAMED LIGHT

Brackets are very complex probabilistically; Dabuz may have a 98% chance to win against his top 32 opponent Oryon, but that 2% where he loses changes everything. There’s a sort of butterfly effect, and that makes calculating exact problities very hard.

So, the probabilities are calculated via Monte-Carlo sampling. Unlike 538, these are not “hot” runs - where the Elo values are updated as the simulation happens - because I really doubt it would change much.

The bracket was filled in manually by me starting with Top 32 and smash.gg’s projected results (for my own sanity). The only major impact that has is that Leffen starts in the lower bracket. Still the 6th most likely to win, despite that.

Contact me at stu2b@statsmash.io or @stu2b50 on Twitter

← Back