Italian translation at settesei.it
Unless seeds withdraw at the last minute, every 2nd round match at Indian Wells and Miami is between a seed and a non-seed. While byes are by no means limited to these two events, Indian Wells and Miami are the only ones that offer us 32 matches pittting a seeded favorite against an unseeded underdog.
Of course, for a variety of reasons, from surface to health to lucky bounces, the favorites don’t always win. But over the last two days at Crandon Park, it has felt like they do. All 32 seeds showed up ready to play, and 29 of them advanced to the third round. Only Juan Ignacio Chela, Feliciano Lopez, and Marcel Granollers lost.
Cue the chorus: That’s got to be some kind of record, right?
Indeed it is, at least back to 1991, which is the current extent of my database. Miami has had the 96-player draw with 32 seeds (and 32 byes) back to 1986, while Indian Wells got into the act in 2004. That gives us 30 past tournaments in my database, including last week’s event at Indian Wells, for the 2012 Miami Masters to measure up against.
On average, seeds win approximately two-thirds of their 2nd-round matches in these 96-player draws. (At tour-level events in general, seeds win 70% of their matches against unseeded players.) In a typical event, then, 21 or 22 seeds advance to the third round. As it turns out, that’s what happened last week at Indian Wells–21 wins, 10 losses, one withdrawal.
This week’s 29 seeded winners doesn’t just set a new record–it blows away the old mark. Three years ago, 25 seeds advanced to the third round in Miami. In 2008, the same number advanced in Indian Wells, and that’s the best the seeds have ever done. Five other times (including last year at Indian Wells), 24 seeds advanced. At the other extreme, the 1997 Miami event was a bloodbath, with only half of the seeds advancing.
It’s remarkable enough that this many seeds won for the first time in 31 tournaments. But the odds are far lower than that. Using my projections for the second round–which, of course, aren’t perfect, and may slightly underestimate the odds of the top few players advancing–there was only a 0.37% chance that 29 or more seeds would win their first matches. That’s roughly 1 in 270.
So, if you were watching yesterday, you were witnessing history. Rather boring history, but a rare event nonetheless.
Do your round-by-round projections make use of the “upset scores” you created for individual players? I’m looking back at this post –
http://tennisabstract.com/blog/2011/08/16/the-most-and-least-consistent-players-on-the-atp-tour/
Also, something like an “upset score” is what I think I meant by “variance” in tennis terms – not mathematically, but in terms of considering short-term swings, i.e. upsets either way. In poker an aggressive player (e.g. Doyle Brunson, back 50 years ago when most players in his circle were conservative), would have much more variance. In theory, he might also enjoy better expectation in the long term, despite all that variance. There’s no direct equivalent in tennis, but an “upset score” seems to come close to a common theme.
My projections don’t make use of upset score, no. I’m not sure that it would make sense to do so — if one were actually using analytics to bet, one might want to have some kind of confidence interval on an individual projection, using upset score, as well as the sample size. (e.g. tough to predict how Roddick will do in a clay match, since he plays on clay so irregularly.)
One thing I haven’t looked at, that might have some predictive value, is whether certain players are more likely to create streaks within tournaments — in other words, whether they are likely to get hot and play above themselves for several matches in a row. But in all sports analytics, streakiness is easy to spot, well-nigh impossible to predict.