Categories

Tuesday, November 3, 2015

Profiting off of Clemsoning

Originally posted Tuesday, November 3, 2015 at probabilis.blogspot.com

In recent years, the term "Clemsoning" has become synonymous with a team shockingly losing a game in which they were large favorites. While the origins of attributing the phrase to Clemson aren't concrete, Clemson has experienced more than its fair share of mind boggling losses, and it may have started with their 2 TD loss to Georgia Tech in 2011.

Related: West Virginia just scored again!

But this is the ACC! There's wealth to be spread! This is the conference that gave us this:



And this:


And this:


That last game (NC State over Florida St.) is an example of Clemsoning, and is one of the inspirations for this post: NC State was a 7/1 underdog in that game. With the ACC's consistently maddening mediocrity and tendencies for upsets, could you profit off of ACC teams "Clemsoning" in conference play?

Thanks to Prediction Machine's Trend Machine, I was able to easily pull all ACC conference games since 2011 (the year "Clemsoning" appears to have started) and look at how often the underdog won, and what the payoff would've been had you bet on that team. If you had simply bet underdogs against the spread, you would've won 52% of the time, which would've been a losing strategy (you need to win 52.38% of spread bets to break even). But we don't care about that! We want straight up victories, the kind that lead to celebrations like this!



The underdog won the game outright 29.13% of the time, which is quite a bit more than the average of around 20.43% (see: "Line (updated)"). Had you blindly bet the underdog in each game, you would've returned a profit of +910 units, which equates to a 3.96% ROI over the past 4 years. So not that much. However, the emotional return is priceless:



Wednesday, May 27, 2015

2015-2016 NFL Win Totals

Originally posted Wednesday, May 27, 2015 at probabilis.blogspot.com

Here are the MDS Model's win total over/under picks for the 2015-2016 NFL season. The calculations are pretty simple: I used the Composite component my model (which is a composite (thus the name) of the Matrix component (only takes into accounts wins and losses) and the Pyth component (strength-of-opponent adjusted Pythagorean expectation)) to determine a win probability for each matchup, and then summed each team's schedule to predict their number of wins. The standard deviation in a team's wins is 2, which was determined both mathematically and via a short simulation (which then allows me to estimate the probability each pick (over or under) is correct).


Win totals were gathered from this article, and are updated as of May 8.


DivisionTeamProjected 1stTeamProjected 2ndTeamProjected 3rdTeamProjected 4th
NFC EastDALPredicted10.26PHIPredicted9.56NYGPredicted7.17WSHPredicted5.31
O/U9.5O/U9.5O/U8O/U6
PickOverPickOverPickUnderPickUnder
Prob64.76%Prob51.24%Prob66.16%Prob63.43%
NFC WestSEAPredicted11.69ARIPredicted8.71SFPredicted7.33STLPredicted6.79
O/U11O/U8.5O/U7.5O/U8
PickOverPickOverPickUnderPickUnder
Prob63.56%Prob54.23%Prob53.42%Prob72.68%
NFC NorthGBPredicted10.39DETPredicted8.50MINPredicted6.37CHIPredicted5.62
O/U11O/U8.5O/U7O/U7
PickUnderPickOverPickUnderPickUnder
Prob61.92%Prob50.09%Prob62.38%Prob75.55%
NFC SouthNOPredicted8.49CARPredicted7.86ATLPredicted7.85TBPredicted4.81
O/U9O/U8.5O/U8O/U6
PickUnderPickUnderPickUnderPickUnder
Prob59.99%Prob62.61%Prob53.06%Prob72.33%
AFC EastNEPredicted11.75BUFPredicted9.31MIAPredicted8.52NYJPredicted5.84
O/U10.5O/U8.5O/U9O/U7
PickOverPickOverPickUnderPickUnder
Prob73.47%Prob65.75%Prob59.44%Prob71.91%
AFC WestDENPredicted10.91KCPredicted9.54SDPredicted8.32OAKPredicted4.24
O/U10O/U8.5O/U8O/U5.5
PickOverPickOverPickOverPickUnder
Prob67.49%Prob69.82%Prob56.29%Prob73.52%
AFC NorthBALPredicted9.29PITPredicted8.64CINPredicted8.28CLEPredicted6.25
O/U9O/U8.5O/U8.5O/U6.5
PickOverPickOverPickUnderPickUnder
Prob55.79%Prob52.88%Prob54.43%Prob54.89%
AFC SouthINDPredicted10.21HOUPredicted9.26JACPredicted4.56TENPredicted4.35
O/U10.5O/U8.5O/U5.5O/U5.5
PickUnderPickOverPickUnderPickUnder
Prob55.73%Prob64.76%Prob68.06%Prob71.76%

For what it's worth, I went 16-15-1 two years ago, when I only used each team's straight Pythagorean expectations. The MDS Model takes into account a lot more factors than that of two years ago.

Saturday, July 27, 2013

Picking 2013-2014 NFL win totals

My methodology uses each team's strength of schedule and win probabilities for each game to determine their expected number of wins in the upcoming season. In making each pick, I consider both value on the betting line (acquired from Bovada) and also making the strict over/under pick (assuming even money on either side). The following picks for each team are ranked from most likely to least likely; note that a moneyline pick is only included if it has value.

Team: Kansas City Chiefs
O/U: 7.5
Pick: Under (90.83%)
Moneyline: U @ -125

Team: Philadelphia  Eagles
O/U: 7.5
Pick: Under (84.95%)
Moneyline: U @ -125

Friday, July 26, 2013

Caveats on sports picks

A few additional points on my picks:

• I will never pick for or against Carolina. In any sport. You just don't do that.

• Inversely, I'll never pick for or against Duke. I cheer for two teams: Carolina and whoever's playing d00k.

• I'll never post a pick for any of my other favorite teams either, which really only include the Tampa Bay Rays.

• The picks are rated 1-5, but 4- and 5-star picks may not necessarily be "locks". Instead, it indicates a higher expected payoff versus the moneyline. While this often does imply a higher probability of winning, it may not pick the huge favorite every single time. 

Instead of thinking in terms of "for sure wins", look at things in terms of probabilities. Even if a favorite has a 90% chance of winning, that still means 10% of the time they're going to lose. The goal is to find when everyone else thinks that underdog won't ever win: knowing that in fact sometimes they will. We're concerned with outcomes in the long run and consistency.