Sportpunter’s new WTA Model – h2h

sportpunter Caroline-Wozniacki-tennisSportpunter have done a major major upgrade to their WTA and ATP models. So major in fact that we started from scratch and rebuild the model from the ground up. Predictions are now available on the website for clients as shown here

H2h bets

Line bets

Set bets

Totals bets

Bet Challengers

But we will go through some of the analysis of the results using a hold out sample for 2013 and 2014.

In this time the model had made 2788 bets, $463,000 bet for $15,731 profit at 3.4% which isn’t bad.

But this is better improved when looking at the table below. This outlines, how the model has gone based on various levels of probability.

00.157 $1,091.49 $921.8984.5%
0.10.2143 $6,050.97 -$3,469.61 -57.3%
0.20.3239 $13,823.27 -$3,182.33 -23.0%
0.30.4339 $27,989.68 -$2,230.19 -8.0%
0.40.5467 $53,962.50 $2,429.85 4.5%
0.50.6492 $79,125.79 $9,453.48 11.9%
0.60.7421 $84,272.81 $2,203.03 2.6%
0.70.8367 $96,200.41 $2,386.05 2.5%
0.80.9211 $74,705.32 $5,547.25 7.4%
0.9152 $25,970.47 $1,671.80 6.4%

As the table suggests, betting on anything less than 40% probability has resulted in a loss. A loss in fact of 16.2%. Maybe we should back the other side here! 😉

But anything above a 40% probability has resulted in a significant profit of $23,691 @ 5.7 %ROI.

So why is this so? Well astute sportpunter tennis fans will already know that the WTA model doesn’t seem to do well with outsiders or low probabilities, and there is a definite reason for this. Shown below is another table which outlines how one would have gone had they blindly bet to win $1000 on every single tennis match.

11.2608 $9,228,294.08 -$12,418.77 -0.1%
1.21.4889 $3,146,868.97 $49,347.57 1.6%
1.41.6823 $1,698,398.43 -$26,843.14 -1.6%
1.621169 $1,566,157.70 -$35,194.66 -2.2%
22.5964 $784,364.07 -$23,487.48 -3.0%
2.53627 $365,560.25 -$10,669.49 -2.9%
34699 $291,394.40 -$16,076.55 -5.5%
45330 $95,464.47 -$32,360.05 -33.9%
58405 $82,019.14 -$16,757.44 -20.4%
8100292 $25,974.82 -$7,340.73 -28.3%
6806 $17,284,496.32 -$131,800.75 -0.8%

Note that when betting to win at odds of less than 1.4, you can actually make a profit, just by betting every outcome. The profit you would have achieved is $36,928 at 0.3% ROI, which no doubt includes a lot of big bets on 1.01 shots. But more interesting is the favourite long shot bias here. When you start looking at betting women tennis players at over odds of 4.00, the results are disastrous. Any blind bet to win $1000 here has lost an incredible 27.7% ROI.

It would appear that the WTA Sportpunter model doesn’t make money with small probabilities because it has to make so much ground up to overcome the favourite long shot bias.

Of course who knows how this bias will unfold in the future, perhaps more people will start betting favourites and hence give more value (or less nonvalue) to underdogs.

Either way, my limiting the betting to players with probabilities over 40% or 50%, one can have a significant advantage in betting the Sportpunter WTA model.

But that’s not all, we will analyse line and totals betting in the next article, and the results are very juicy indeed.


Related Posts with Thumbnails
This entry was posted in Model, Sport Models. Bookmark the permalink.

2 Responses to Sportpunter’s new WTA Model – h2h

  1. nugdan says:

    I wouldn’t have thought the fav/longshot bias to be relevant here. If the odds are always biased to the fav, then your model should identify value more frequently on favs and less on dogs. and thus recommended bets on longshots >$4 should be quite rare. The fact that this is not the case has little to do with fav/ls bias but suggests to me that your model is being overly generous to the longshots chances. Perhaps the spread of the distribution of expected results is too wide…

  2. admin says:

    yep agree nugdan, i think its a combination of fav/longshot bias as well as how the model works. (which with wta has never favoured outsiders)

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.