Sunday, 21 January 2018

Championship Weekend and More on the Draw

The Home team and Overs is the trendy play in the NFL Championship round. From the 2001 season, Home teams are 17-15 ATS (against the spread) while Overs has a 19-12-1 record.

The first game is the AFC's Championship game between the New England Patriots and the Jacksonville Jaguars. The line is currently 7 points, and of the 11 previous Championship games with a spread in the 5 - 9 point range, the home team is 7-4. When New England are hosting the game, the Unders has a 5-1 record, the one Overs being last season's rout of the Pittsburgh Steelers. While this type of team stat is generally useless, it's a little more relevant here given that the head coach (Bill Belichick) has been there since 2000, the same year current quarterback Tom Brady was drafted. 

The second game is the NFC title game between the Philadelphia Eagles and the Minnesota Vikings. The Road Vikings are favourites, giving three points, only the sixth time the road team has been favoured in a Championship game.

In 2016, the Patriots were three point favourites at Denver Broncos and lost the game.

In 2012, the San Francisco 49ers won and covered against the Atlanta Falcons when getting 3.5 points, and in 2010, the Green Bay Packers were also 3.5 point favourites at the Chicago Bears, and both won and covered.

In 2008, the Eagles were 3.5 point favourites to win at the Arizona Cardinals but lost by seven and in 2005, the Patriots were three point favourites in Pittsburgh and won by 14 points.

In the last four seasons, all eight home teams have won, and by some big margins. The average line for the home team has been 4.5 points, while the average margin of victory has been by 17.25 points.

While looking at the numbers this week, I discovered that the Week 18 for Wild Card games, Week 19 for Divisional and Week 20 for Championship wasn't actually the case in the 2001 season, which was extended by a week following the 9/11 attacks. The spreadsheet has been corrected.

I was also corrected by the esteemed Jospeh Buchdahl on my numbers for the EPL Draws for 2013-14 in my post on the EPL Draw last week. The problem was that I had resurrected a spreadsheet that only had results up to February of that season. The numbers by season are tabulated below, although the 2011-12 season was a more exclusive XX-Draws sub-set of the selections from 2012-14. Also worth mentioning that we did not have Pinnacle prices in 2011-12, so I have used those from Bet365. I've also used those for a couple of matches in Serie A where there were also no Pinnacle prices, presumably due to the Dodgy Draw in those games. 

The table shows the XXDraws by league and by season, and for comparison also the results of blindly backing the Draw (not a good idea) and of backing it in matches where neither team is odds-on, which in the EPL at least is a good starting point.

Regarding the numbers for 2013-14, this was a very poor season for Draws across the big leagues: 

Joseph's thoughts on this season were that:
This show that with other European leagues included XX Draws failed to make a profit that year. Does this not illustrate my original point about data dredging?
For me, this is not what "dredging" is about, and in the same way that a managed mutual fund will tend to go down if the indexes go down, and up in an up year, the measure of its worth is how that fund compares with a benchmark.

The benchmark in 2013-14 for Draws performed rather poorly, which isn't surprising given that compared with the previous ten years, Serie A was at its fewest ever, the EPL was at its second fewest ever, one Draw more than in 2005-06, the Bundesliga was also at its second fewest ever, one Draw more than 2010-11, La Liga was at its third fewest ever, and only in Ligue 1 were they close to average, albeit still below average.

There were 47 more draws the following season, a number repeated in 2015-16. 

To me, data dredging is more along the lines of 'discovering' a pattern in data, and trying to back-fit a system for it. For finding value in Draws, there's a logic to it, first mentioned by McGovern, and later expanded upon by David Sumpter. It makes perfect sense logically that closely rated teams should draw more frequently. The trick is whether this is profitable or not, i.e. is the market inefficient here? since Draws by themselves are meaningless. If the percentage of drawn matches shoots up to 50%, and the price in general plunges to odds-on, not a lot of good! 

Joseph's response was that:
It depends how you developed the system. If it was via back-testing a load of data there is always the risk that this is exactly what it is. How did the earlier seasons for the other European leagues perform? Have you got those figures?
There was no back-testing at all. I tried to explain how I stumbled across the draws being (relatively) good value in last week's post. Back testing wasn't even possible because the results were based on constantly changing Elo ratings and goal expectancies.

After the 2013-14 season, I realised that the time necessary to perform the updates each week and determine the selections wasn't worth it. A simple system based on the prices, and the differences between the teams, worked well enough. It works even better if the stakes are additionally weighted by Big 6 as well. 

Incidentally, yesterday's three selections by this basic (exclude odds-ons) method threw up two more winners in the games at Everton and West Ham United, with a loser at Stoke City. 

Joseph also made a comment on Derek McGovern’s book “On Sports Betting …And How To Make It Pay“ which I had mentioned had planted the seed of the idea that the draw might potentially offer value.
Joseph tweeted
It was actually a rather disappointing book which has not aged well as betting has evolved, and as I replied to Joseph, I very much doubt that he is paying Premium Charges to Betfair. We could ask him! He does have an account there, posting on the Forum under the name charwell as recently as Friday. I suspect a lot of self-published authors are better at selling books than they are making actual money. It used to be said that Those Who Can, Do; Those Who Can't, Teach. In the world of betting and trading, it's more along the lines of Those Who Can, Do; Those Who Can't, Sell Books About It.

** 21 January: The table was updated to reflect correct percentages on the All Draws column. Thanks to Joseph Buchdahl for pointing that out.

No comments: