Friday, 14 December 2012

In defence of Lawro - How one man is beating the market

Quite often on football forums I'll see something along the lines of "Lawrenson's tipped us to win, we're doomed" as if Lawro was some sort of reverse Midas.  Lawro tends to get a lot of stick but is rare among pundits for giving his predictions on a regular basis and also having them held up to scrutiny by being collated and published on the BBC Sport website.

Far too often I'll listen to podcasts where the prediction goes along the lines of "I'd expect Chelsea to win, but it wouldn't surprise me if West Brom got something from this game" where pretty much every base is covered.

What I've done is look at the results of the 158 predictions on Premier League games made this season and the expected return if a notional £10 was placed on each game according to Lawro's predictions (I've used Home/Draw/Away rather than looking at correct score betting).  I've taken the odds from Bet365 as held at Football-Data

Top level figures are:
Overall it can be seen that sticking a tenner on each game based on Lawro's predictions would net a decent profit.  The expected loss figure comes from the fact that the combined odds of Home/Away/Draw for Bet365 come to around 105%, if you were willing to shop around using a dozen or so different bookies and picking best odds for each individual bet this would come down closer to 100%.

How does Lawro achieve this profit?  Breaking his selections down into the type of prediction, it can be seen that all the profit comes from the Draw predictions:
Lawro has been correct with half of his predictions where he's said it'll be a draw where the actual odds for the draw were never below 3 (or 2/1 in traditional odds).

Is Lawro a mystic? Is it that his football brain is smarter than the market? Ultimately bookies odds are a mix of what they think is going to happen and where the money is going, if they feel Man Utd are unlikely to win but are taken significantly higher than expected bets on that happening then they'll shorten the odds accordingly.

It may be just be that selecting draws is for boring people and the amount of money going on a Home/Away win is more than the true probability of that happening, resulting in over-generous odds on draws.  If you were to just pick one kind of outcome and have bet on that each time, below are what you would get back:
If you knew nothing about the matches and just stuck a tenner on a draw for every game then you'd end up with a profit not far off Lawro's.

The above is far too small a sample to say definitively that the easiest way to make money is just to back the draw, there's no guarantee this would happen in the future but is arguably something worth looking at in more detail across historical data for multiple seasons/divisions.

Looking at Lawro's actual score predictions also highlights how he tends to come to these predictions.  Ultimately he has to spend the weekend defending them, so looking at his predictions suggests he likes to give some comfort to any team he's predicting will lose by at least predicting them to score.
In over three quarters of predictions, Lawro goes for one of 2-0, 1-1 or 2-1 and pretty much always goes for more than one goal being scored in a game with 38 of his 40 predicted draws being predicted as 1-1
Picking these three scorelines suggests (either deliberately or subconsciously) he's playing it safe, if you say 1-1 and the actual result is 2-1 or 1-0 or 0-1 then you could easily claim to be "not far wrong" and avoids being even further out when the 5-2, 0-5 or other freak scorelines come in.  It's also interesting being a former Liverpool play that he's never predicted Everton to lose this season.

It could also be argued that given 1-1 happens almost twice as often as 0-0 in the Premier League so far, if you felt the game was too close to call that 1-1 is more likely a result than 0-0.  That doesn't however explain the over prediction of 2-0 and 2-1 compared to 1-0 and to me suggests a relatively straightforward process is happening:

  • Easy home win for a Manchester team: 3-0 (all 6 of these were home wins but at tiny odds)
  • Easy home win: 2-0
  • Tough home win: 2-1
  • Too close to call: 1-1
With a few odd exceptions where a "bigger" team is playing away making up most of the rest of the predictions.

Just because it's simple doesn't make it wrong however, an awful lot of how bookies make money is where people get seduced by "big" odds that are actually poor value.  It may be sitting on the fence is the easiest way to make a profit.

There are 158 rather than 160 predictions as Sunderland-Reading was postponed from its scheduled date and no prediction was made for the brought forward Chelsea-Reading game.  Lawro did pick Sunderland to beat Reading when originally scheduled and I'd be very surprised if he would didn't expect Chelsea to beat Reading so could argue these would have also been called correctly.

Bet365 was chosen as it was first bookie in the supplied spreadsheet, as mentioned above, shopping around can improve odds received reducing expected loss if an average punter.

Twitter: @we_r_pl
Football Stats/Analysis: 
More around Data and Analysis:  Analysis Marketing