Monday, 31 December 2012

Statistical inference and mathematical modelling (part 7)

7.0 Analysis of results

In the final part of the series, it is time to analyse the results of the NFL math model. Here are each's team predictions for season interceptions, together with the outcomes of the 26 individual bets. 
The confidence level of each pick is in the third column. The teams are ranked in order of their unlisted Kelly % - a measure of hypothetical profitability - rather than their win %. This is because the best odds on offer about each proposition varied.

New Orleans SaintsOVER 12.569.8%WON
New England PatriotsUNDER 22.069.5%WON
Chicago BearsUNDER 20.066.8%LOST
Baltimore RavensOVER 16.567.4%LOST
Philadelphia EaglesUNDER 18.564.7%WON
Dallas CowboysUNDER 18.064.1%WON
Arizona CardinalsOVER 14.063.4%WON
Tennessee TitansOVER 14.062.8%WON
Denver BroncosOVER 13.561.4%WON
Buffalo BillsUNDER 18.562.0%WON
Oakland RaidersOVER 15.560.5%LOST
New York JetsUNDER 17.560.9%WON
Kansas City ChiefsUNDER 18.559.9%WON
Indianapolis ColtsOVER 11.559.2%WON
Jacksonville JaguarsOVER 15.060.1%LOST
New York GiantsUNDER 18.558.6%LOST
Carolina PanthersUNDER 17.559.0%WON
Detroit LionsUNDER 18.058.1%WON
Houston TexansOVER 16.557.0%LOST
Cleveland BrownsOVER 13.558.1%WON
Green Bay PackersUNDER 24.558.9%WON
Miami DolphinsUNDER 16.056.7%WON
Minnesota VikingsOVER 11.557.3%LOST
San Francisco '49ersUNDER 20.555.8%WON
Atlanta FalconsUNDER 19.055.0%LOST
Washington RedskinsOVER 14.554.2%WON

7.1 Summary
  • The 26 Selections went 18-8 overall, a win percentage of 69.2%.
  • The nine best bets went 7-2 (77.8% strike-rate); the next best nine were 6-3 (66.7%) and the remainder 5-3 (62.5%). This is a particularly satisfactory result as far as staking is concerned.
  • The entire portfolio showed a 133.2% return on investment at the listed stakes and odds. Had level staking been adopted (the initial £1000 divided into 26 bundles of £38.46) the ROI would have been 128.3%.
  • There was a total excess of 87.5 interceptions compared with the market's projections, or 3.4 interceptions per team. There is a very large margin and strongly indicates the portfolio was extremely robust. In other words, this is an exploitable market and will remain so - and there are plenty of others like it in sport.

December 31/2012