US Open Men’s Draw Predictions

Get ready for a shock: I’m forecasting Novak Djokovic as the winner of this year’s U.S. Open.  I give Djokovic a 27.8% chance of winning the tournament–a higher probability than I gave him at Wimbledon.

There’s a marked difference between Novak and the rest of the pack, in part because Juan Martin del Potro could wreak havoc with the bottom half of the draw.  I give Rafael Nadal a 14.6% chance, Andy Murray 9.2%, and Delpo 6.6%.

Federer comes in fourth behind Murray, at 8.9%.  Making his road tricky is a likely quarterfinal matchup with either Jo-Wilfried Tsonga or Mardy Fish.  Fish does better in my hard-court rankings than on the ATP computer, and is sixth-most likely to win the tournament at 4.2%.  Tsonga comes in 7th at 3.8%, as he shines on hard courts.  Also, my algorithm takes into account Tsonga’s wins over Fed.

Seeded Americans Andy Roddick and John Isner do better than their rankings would suggest, in large part due to their hard-court prowess.  Roddick has a 2.1% chance, and Isner 0.2%.  The overall chance that an American wins the event is 6.6%–just a tick above the combined probability of Fish or Roddick winning, and equal to Delpo’s shot.

The unseeded player my system favors is Nikolay Davydenko, at 0.7%.  Recent disasters aside, he is one of the few players who has proven he can beat the best players in the game.  As I recently wrote, his inconsistency may actually be a good thing.

There are several first-rounders that figure to be extremely tight matches.  Here are all the opening matchups where the favorite (in bold) has less than a 55% chance of getting through to the second round:

  • Granollers vs Malisse
  • Kukushkin vs Montanes
  • Bellucci vs Sela
  • Kohlschreiber vs Stepanek (almost dead even)
  • Bubka vs Haider-Maurer (also nearly even)
  • Dancevic vs Ilhan (two qualifiers)
  • Baghdatis vs Isner (maybe that would change if I re-ran my rankings through Winston-Salem)
  • Young vs Lacko (Kubot pulled out, making both Donald and Lacko lucky)
  • Matosevic vs Chela (actually 58/42, Chela being the least-favored seed)
  • Rosol vs Pospisil (another 50/50, I’d probably bet on the young Canadian)
  • Istomin vs Sweeting
  • Roger-Vasselin vs Muller
For the full breakdown, I’ve published the chart here.  That page will update automatically throughout the tournament.

7 thoughts on “US Open Men’s Draw Predictions”

  1. Great stats Jeff! John Inser’s rank can’t go up much with the 250 series because he already has one win and a final in that category. But, the match toughness and W.S. QF win over Marcos is a big plus heading into the open.

    Lucky Loser Lukas Lacko = LLLL !

    1. Thanks.

      Isner’s rating will go up in my system, though, which is based entirely on *who* you beat, not at what tournament or in what round. Sela/Nieminen/Benneteau doesn’t count for a lot, but five more wins, esp. those against Baghdatis and Roddick, will help him out.

  2. Thanks for the stats, but you give Potro way too much credit. What has the guy done lately? Imo Isner/Baghdatis is just as much a favourite to reach R16 as Potro is…

    1. There’s more to a player than his last three months. Delpo has two titles this year, and won sets in the last two slams against Nadal and Djokovic. He’s not as good as those guys, but he’s arguably had as good a year as Fed or Murray.

  3. Your system resembles that used in chess, Jeff. Roughly speaking, you get your opponent’s grading plus a certain number of points if he you beat him, less the same number if you lose to him, and the same if you draw.

  4. This is great! Do the predictions update themselves just based on who wins and loses or are the most recent matches taken into account in terms of how well everyone is playing. Also I was wondering if you had the original forecast saved somewhere in order to compare original versus updated probabilities.

    1. Sorry for the late reply — didn’t see your comment until now.

      The predictions don’t use matches from the current tournament to adjust each player’s rating. I’ve experimented with that a little bit, and haven’t found anything useful that comes from it. Don’t think i have the original forecast anywhere, but I can probably generate a new one using pre-Open assumptions.

Comments are closed.