The International approaches, and with it comes floods of predictions, odds, and opportunities to win arcana or go broke trying. Despite Secret's unprecedented dominance, the savvy fan may have some serious reservations: the top-ranked team to open the International has never been the top-ranked team by the end.


Should Team Secret win The International next month, they will be the first team in Dota history to go into a multimillion dollar tournament as the statistical favorite and win. That's just one more in a long list of records they will take, including highest ranked team yet, longest consecutive days as the #1 team in the world, and likely most career days as #1 team in the world.

Secret has been dominating in an unprecedented manner, and against more difficult opponents than has ever been concurrently fielded.

But how often do favored teams win? Why do otherwise very successful projection models break down in the face of The International and similar events?

I'm defining the favored team as the team with the highest Elo. Elo is the ranking system used in Chess. Basically, teams earn points based on how often they win and against whom. In Dota, Elo correctly projects over 85% of games but only 21% of tournament winners. For this article we are looking only at the single team statistically most likely to win, not an array of teams each of which is likely to place well. Only one team can win and only one team can be a true "favorite".

The teams expected to win tournaments rarely do.

This makes sense if you think about it: if 16 teams are competing, the odds of any one team beating all 15 other teams are much lower than the odds of them beating any individual opposing team. If a tournament has more teams, the winning team is statistically more likely to be an underdog. Pinnacle eSports' betting odds project Secret with 42.8% likelihood to take first place, which is quite high even for a top-ranked team.

TI Odds according to pinnaclesports.com


Obviously enough, tournaments with closely-ranked teams also have lower winrates for favored teams. This is pretty easy to understand: the more closely-matched teams are, the less likely any specific team will dominate.

This is bad news for Secret fans:

this year's International has the most closely ranked teams in the history of the event.



TI5 will also have the highest-ranked teams in the history of Dota 2 multimillion dollar tournaments. Fortunately for Secret, average team skill at an event doesn't have a significant impact on whether or not the best-ranked team will win a tournament.

Tournaments also come in all shapes and sizes. Since 2011, there have been 172 ticketed professional Dota 2 tournaments, of which 98 have been since 2014. That's $27,517,230 up for grabs... with another $17,000,000 around the corner.

Don't be alarmed by the following chart: I'll walk you through it nice and slowly.

The Y Axis is the average Elo of all teams participating in the main event of a tournament while the X Axis is the standard deviation (the average distance from that average) of those ranks.


What we have

here is every tournament in Dota history graphed by the variation between participating teams' skills and the skill of participating teams on average

. This means the lower left has tournaments like The Canada Cup or Korean Dota League which are either region locked or have prize pools too low to attract tier 1 teams. Because of this, these tournaments don't have the most skilled teams, but all the teams are about the same skill level.

On the opposite end of the spectrum in the top right you have tournaments like EMS seasonal cups which have a qualifying component and an invitational component and end up pitting tier 2 or community teams in the same pool as big hitters such as Vici Gaming or Evil Geniuses.

Those tournaments have high skill level, on average, but they also have a very large variation between the teams as a whole.



The top-ranked team is equally likely to win at any range of skill diversity

but only if the skill of participants trends with it.

In other words, as the average skill goes up the average diversity must go up, too. Otherwise the top-ranked team's odds of winning change.

The middle square has every "average" tournament, with the squares on either side representing everything more than one standard deviation away from that average.

The grey circle is the home to the most premiere tournaments, including ESL Frankfurt, Dota 2 Asia Champtionships, and of course The International. It also happens to be in the only section of the grid where winning as the top team is less likely to a statistically significant degree.



Secret prepares to defend their position against the world's top competitors.


Between The International 2011 and The International 2012, the only match Invictus Gaming managed to take was against Mouz. Granted, ticketed games were much less common in the earlier days of competitive Dota 2, but out of all the competitors going into TI2, iG were least likely to win based on their past performance.

In the months leading up to The International 2013, Na'Vi had a winrate of over 73% and the highest rank in the world on the eve of the tournament. In fact, Na'Vi's Elo rank was essentially indentical to Secret's rank now. In contrast, Alliance had huge loss streaks throughout the season leading into the event, and while they won an impressive series of titles they also lost lots of games due to a variety of reasons in the lead-up to TI. Alliance had an Elo just below 1300 to start TI3 while Na'Vi was ranked just above 1500. Alliance were expected to win by many, maybe even most, but they didn't have the best win/loss record going into the event.

Newbee had only played four games against Western teams and had lost 75% of them in the 2014 season, also losing the large majority of their games against high-ranked Chinese opponents. Estimates for their success before the tournament typically placed them between third to fifth for TI 4.

Of course, if you look at the runup to TI 3, many fans correctly expected Alliance to win due to their strong performance at major events. This is key:

in the face of The International, game-by-game or even tournament-by-tournament performance is not a reliable indicator of future success.

This is a tournament that takes more than just odds to win.

Every team that won leading into The International peaked at just the wrong moment and lost a place in Dota lore because of it.

The question remains:

will Secret make history again or will this year be a repeat of the winner's curse?



Note: For all conclusions regarding comparing tournaments where the winner was the favorited team, I ran a T-test with two groups: tournaments where the expected team won and tournaments where the expected team didn't win. Only findings which come back with a P value below .05 were reported as significant.

Elo provided by Noxville (the official English statsman of The International 2015). Special thanks to him.


What do you think? Did Secret peak too early?


Comments