Thursday, 22 January 2015

Let Them Eat Porridge

Marty writes, I hope facetiously:

“If I told you that Doncaster eat porridge for their half-time snack would you retract this criticism? This slow burning food typically sees them underperform their opponents (who typically eat faster burning sugary foods at HT) in the 15 mins after HT. Obviously, after 60 minutes they benefit from higher energy levels as opponents suffer from the sugar crash while the good old porridge keeps burning." 
I would certainly consider retracting my criticism if it were proven that the half-time choice of nutrition had some predictive merit, but this scenario is rather improbable.

The null hypothesis here is that the half-time break has no significant effect on Doncaster’s early second half performance, and rejecting this requires a significance test to show a 95%+ likelihood that the results do not fit the null hypothesis. That's a tall order.

Although Marty uses the unlikely example of the team eating porridge as a possible predictive parameter, the question really is why anyone think that half-time affects one team more than another team?

There are reasons, and one reason, far more likely than the choice of nutrition, might be that the coaching staff are better at their jobs than average, and are able to make positive tactical changes*, (perhaps Milton Keynes Dons) or worse than average, and fail to adapt and hurt their team in the process (perhaps Crewe Alexandra). [Milton Keynes Dons go from a first half goal difference of +5 to +22 in the second, while Crewe go from 0 to -20]. 

However, I still find an arbitrary period of time chosen simply because it divides into 45, somewhat problematic. If you are good at determining value at half-time, at least give it the full 45 minutes rather than expect it to reap rewards straight away. The 59th minute is closer to the 61st minute than the 46th, and it's a poor use of data to ignore this.

Although I took the statistic quoted for Doncaster Rovers as true when I wrote the post yesterday, it doesn’t actually appear to even be correct, or at least not if we are looking at league matches (and surely no one serious about betting would include cup matches, which are quite different in nature). 

It’s also worth mentioning that assuming all data is the same (e.g. that an event away against a team in first place is the same as that event at home against a last placed team) is only going to lead to trouble.

Anyway, the facts are that Doncaster have conceded just four goals all season in the 46’ to 60’ period, a number bettered by just four other League One teams this season. The average in League One is 5.8 goals conceded for this 15 minute period, so if anything, Doncaster Rovers’ half-time ingestion of oatmeal appears to be paying off handsomely.

In fact, Doncaster have actually conceded more goals in the 1’ – 15’, 16’ – 30’, 31’ – 45’ and 61’ – 75’ periods than they do at the start of the second half, and their defence in the second half overall is the best in the division, conceding just 12 goals!

The tip to simply “trade out at half-time if the score is 0:0” never made any sense, and makes even less sense now, if that is even possible!

What about price? Why does the price not come into play here? "Trade out at HT 0:0" – that’s it? Agreed that trading out at 1.01 would be great value, but at 1.80, not so much.

The Tweeter falls into the trap of thinking that having backed at 1.85, laying off at 1.2 or whatever is a good trade. It’s not, and there is no reason why the in-play market will offer value at half-time rather than at any random point in the match. A good trade is one that was a value back (or lay) on taking the initial position, and a value lay (or back) to close out the position. Any other combination, and it was a trade that could have been better, or could hardly have been worse! 

As I have said before, when it comes to football, if you have identified value pre-game, then you will be extremely unlikely to find value on the other side of the bet during the game.

*In my opinion you see this more often in sports such as the NBA or NFL, where there are more breaks than the one half-time interval seen in football, and more frequent changes in team line-ups and player match-ups.

No comments: