Saturday 13 January 2018

Probability, Difference And The EPL Draw

The Draw result in football has been a mild obsession of mine for several years. 


I’m not sure when I first stumbled upon the idea that the odds on the Draw might not be efficient, but it may well have been when reading Derek McGovern’s book “On Sports Betting …And How To Make It Pay“ maybe ten years ago. 

The book was published in 1999 but my interest in betting wasn’t re-awakened until discovering the Exchange concept in 2004, by which time some of the book’s content had become obsolete.

Much of the chapter on football betting dates back to a time when punters were far less sophisticated than today and bookmaker restrictions meant that for most matches, home teams needed to be backed in accumulators of at least five, and the draw and away selections in minimums of three. The typical over-round was 111%, a far cry from today 102% with Pinnacle and in theory close to 100% if you shop around and have access to several books.

McGovern intriguingly wrote:

Regular soccer punters will know that in football betting the odds are very much in the bookmaker’s favour. In a match in which theoretically all three potential outcomes – home win, away win, and draw = are equally likely, the true odds should be 2-1, 2-1, 2-1 [for younger readers, 2-1 is 3.0 in decimal odds]. But such odds give bookmakers no profit margin, so instead they will offer, say, 6-4 [2.5] the home team, 6-4 the away team, and 11-5 [3.2] the draw, the draw being longest price because bookies know that few punters ever back it. This produces a book of 111% giving bookies a theoretical profit of 11 per cent whatever the result.
Unless the expectancy of any goals is extremely low, or there’s a fix in place, the Draw will never be favourite, but the interesting part was that a book might be offering value on an outcome simply because the bet might not be popular.

Later in the chapter, McGovern talks about bookmaker Ron Wadey, who at the time owned five betting shops in the north of England. On the subject of odds-compilers, which such independent bookies used in those days, Wadey is quoted as saying:

“Their job is to come up with odds for a football match that will attract three-way support – for the home team, the away, and the draw, although I accept that the draw will never be popular with punters”.
Further on, McGovern writes:
Bookmakers freely admit that few punters back the draw in football matches. Punters want an allegiance in a game, to shout on a given team. Backing the draw is almost an admission that they can’t make up their minds.
The football punter in those days was clearly recreational.

After writing that roughly 30% of matches in the top two English Divisions in 1998-99 resulted in a Draw (implied odds 9-4 [3.25]) he observes that:
“Yet very often 12-5 [3.4], 5-2 [3.5] or even 11-4 [3.75] is available about the draw in a top-flight game. Bookies can get away with offering over the odds for the draw because they know it will attract little money, and it allows them to tighten even further odds for the home and away sides”.
This led to the logical conclusion that:
"a loophole could be exploited in matches in which bookmakers could not split the two sides, for instance offering 6-4 [2.5] the home win, 6-4 [2.5] the away win and 11-5 [3.2] the draw.
Bookies were saying, I argued, that there was absolutely nothing between the two sides, that an away win was just as likely as a home win. But by doing that they were also implicitly admitting that the draw was an equally likely result. Indeed you could argue that if the two sides are so evenly matched, a deadlock is the likeliest of all three outcomes. Therefore, the true odds of each outcome must be 2-1 [3.0], translated into 7-4 [2.75]  once the bookmaker’s in-built margin is taken into account.
If you could consistently get 11-5 [3.2] or more about a 2-1 chance, my reasoning went, profits were inevitable”.
What a lovely thought – inevitable profits. Nothing is usually this simple but it was an idea that stuck in my mind.

Around 2009, and in hindsight I was something of a pioneer in the field given the recent surge in interest, I started calculating expected goals (now referred to as xG) and deriving probabilities, and thus odds, for all goal related markets – e.g. match odds, over / under, Both Teams to Score.

My Excel spreadsheet maintained Elo based ratings for teams in the top five European Leagues (English Premier League, La Liga, Serie A, Bundesliga and Ligue 1). After each round of matches I would enter match data, and update the ratings. I would then enter the upcoming fixtures and the spreadsheet would calculate probabilities on various outcomes.

It was while entering data from completed matches that I discovered that for drawn matches, more often than should have been the case, the outcome was that the ratings for both teams would remain unchanged, or almost unchanged.

It wasn’t too difficult to reverse engineer these findings so that by entering an upcoming fixture, it would generate a figure for how much a draw would change the ratings. A number between -0.1 and 0.5 was a strong indicator that a Draw might result, with higher positive numbers indicating a home win, negative numbers an away win.

From this came the hugely successful XX Draws which were profitable for the three seasons (2011-14) I shared them, until a work promotion reduced my free time and I suspect my very basic xG calculations were about to be improved upon by far better resourced enterprises anyway.

The XX Draws included selections from the top five European Leagues, and the results from these years are below:

My model was also rather good at finding the Draw in matches where it was priced at 4.0 or higher, although the effort of finding an average of seven selections a season can't really be justified. 
I mention the 4.0+ (IP 25%-) mark because blindly backing the Draw at this price in the EPL is a losing proposition.
In a recent post, I published the results of backing the Draw selectively from the 2,120 matches played in the EPL since the start of the 2012-13 season. Those numbers used the maximum price for an apples to apples comparison with Soccernomics' numbers, but the Pinnacle Closing price is more realistic, and using these, the following table might be of interest to some of you.

The Big 6 and draws have received some recent attention for good reason - blindly backing the Draw in these fixtures has an ROI of 14.3%, a number that is easily exceeded by being selective on these games. 

To help make the table easier to understand, P is Probability, D is Draw, H is Home, A is Away, W is Win. The 'Difference' is a measure of how far from the favourite the other team is - the lower this number, the closer the two teams are. For examples, the Draw in matches between teams with a difference of 2.5% or less has an ROI of 40.3%, but a low sample size. 
For Premium Charge mitigation, you can do worse than blindly back the draw when it is priced at under 4.0, but a couple of simple additional parameters, such as excluding any matches where one team is odds-on, can improve returns hugely. 

I'll update these numbers in the summer, and may take a look at some of the other leagues in the meantime. 

Looking at today's games, and we have an odd situation with the Albion derby match, West Bromwich v Brighton and Hove. At the time of writing, Pinnacle have:
This 46.5% v 23.2% 'profile' hasn't been seen before. When the probability of the home team winning is .465, the average Draw and Away prices are 3.44 and 3.79 respectively. 

When the Away side is around 23.2%, the Home team has always been odds-on, 1.93 on average, with the Draw at 3.72!  

The Draw has only Closed at 3.1 or less on 29 previous occasions, (since 2012), and was a winner in 14 of them, so don't let that short price necessarily put you off. 

Has word of the EPL draw value finally got out? 

No comments: