Exploring Value in Shots on Target

Working on a theory that there is value in Shots on Target due to ridiculous amounts of bias in the Overs.

To do this I had a look at modelling my own Shots on Target this week based on a combination of historical data and current match AGS/FGS prices, and it initially appears to me that the bookies have significant bias in the Overs.

I looked at all shots on target in the top four UK leagues between 2011 and 2019 and compared them against pre-match xG (derived from o2.5 prices).

UK leagues had an average of 8.55 Shots on Target, and an average xG of 2.61.

The two blue dots at the top of the scatter are Liverpool v Watford in 2016 and Alfreton Town v Grimsby in 2014 – both had 25 shots on target, 3 more than the third highest game. There is a direct relationship between xG and xSOT and we can start to model xSOT under our own parameters instead of the bookmakers. Back-testing these modelled numbers against assumed bookmaker lines (generated from regression) showed reasonable profit betting on the Unders in the 19/20 season. As it also seemed to do in 18/19, 17/18, 16/17, 15/16……

Shots on Target seem to be an avenue that bookmakers continue to go down. Fred included them in his daily specials this week. Sky have begun to include them in their headline boosts. It is a market that could be ideal for exploitation in the future if it continues to grow. A long data collection exercise has begun on a simple premise that may suggest some validity in the theory: recording the p/l of a strategy of betting all Under Match Shots on Target.

Notes about nothing.

NB1: 83/1 7-fold lands in Europ last night. Would have been nice if i was on  (+sample size above is meaningless, obvs)

NB2: We published some of our modelled xSOT by team and player numbers, but then locked them to private after something of a backlash. Criticism labelled the numbers as “Ludicrous”, “Wrong” and having “Invented” based on existing bookmaker prices. My position is that we are estimating xSOT following our own models. through our own data-driven processes. There is no “right” or “wrong” answer when estimating expected variables in sport.  Can we beat what the bookmakers are using? Maybe. Maybe not. There can always be room for improvement and precision. But by bench-marking our estimations against existing bookmakers prices they themselves may be applying a logical fallacy as our premise is that the bookmakers are significantly biased. Take any colourful language with a pinch of salt – we will monitor this using empirical analysis and stats.