Thursday 25 August 2016

An Odd Look At Even Money Shots

Some rather sloppy 'research' (I use the term loosely) over at Pete Nordsted's site, looking at "even money shots" in football's top leagues.

The first error was in expanding the "even money" definition and:

Classifying teams priced between 1.9 and 2.1 seemed very enough as it included teams at the 9/10 level through to 11/10.
What "seemed very enough" means I'm not sure, but I'm guessing the writer (Jimmy) means "seemed good enough" but it really isn't.

1.9 (or 9 to 10) means the selection has an implied probability of 52.63%, with the inverse of that bet being 2.11.

2.1 (or 11 to 10) means the selection has an implied probability of 47.62%, with the inverse of that bet being 1.91.

So presuming you want your range to be fair and balanced, it should either be from 1.9 to 2.11 or from 1.91 to 2.1.

Unless you are into your statistics, this may seem picky, but by using a range of 1.9 to 2.1, you are including more odds-on selections in your sample than you are odds-against.

To clarify, by Jimmy's logic, and by extension, more expansive ranges would then be those from 1.8 to 2.2, from 1.6 to 2.4, from 1.2 to 2.8, and ultimately from 1.01 to 2.99. But the average is still evens... No it isn't.

You can see this logic is clearly flawed, so it is hardly a surprise that the findings would be:
If we use Pinnacle's Closing Prices to determine selections, a balanced range of 1.91 to 2.1, over the last four seasons in the EPL, gives numbers that are a little different -
  • Home wins 60 of 134,
  • Away wins 26 of 55
  • A win ratio of 45.5% overall - and a loss of 16.13 points from the 189 matches.

Missing from Jimmy's post is any mention of profits or losses, which is quite an oversight. I would hope this isn't intentional, with the reader expected to assume that the odds average out at evens and backing "even money shots" is thus profitable, because they don't and the strike rate at 1.9 is quite different to that at 2.1.

"But Cassini, that's a big difference", I hear you say, and you're quite right - it certainly is.

The reason may be partly that Jimmy is looking at the past five seasons, (versus my four), and rather than use the Closing Prices at kick-off to determine selections, his 'analysis' (I use the term loosely) uses the "Best" price to determine qualifiers:
Please see below for a breakdown of the four major leagues with a home and away split of results for teams with a best price between 1.9 and 2.1 at kick off.
At kick off? Admittedly, it is possible that Jimmy has access to kick-off (Closing) prices from several bookmakers, but I am not aware that this data is readily available. Hence my excitement when Football Data updated the past four seasons with this data.

Anyway, including matches where one "best" bookmaker is offering a generous (but probably unavailable in practice) price of 1.9 is another reason why this 'study' (I use the term loosely) is worthless rubbish.

1 comment:

Smithlondon said...

"At kick off? Admittedly, it is possible that Jimmy has access to kick-off (Closing) prices from several bookmakers, but I am not aware that this data is readily available."

This is a bugbear of mine with Betfair. Since 2014 they have had SPs on Premier League games but unlike all their other SP prices they do not publish them. I have asked in vain. I have taken to recording them this season myself. Would be interested if you know of anyone who has recorded them since they started and would be willing to make them available.