Monday, 26 October 2015

Reductio Ad Absurdum

My post yesterday included a few lines about meaningless data, and another example reared its head on the Simple Soccer Stats blog. In a post titled Corner Odds User Question And How They Are Calculated we see (edited for readability):

This question was received this morning

Example: Aston Villa v Swansea for the Line 9.5 corners in Match. Your site says for under 9.5 the odds are 1.33 so when I get under 9.5 on bet365 at 1.83, there is a big value or is the 1.33 not the odds?
The reply given was

Yes that is saying there is value. Aston Villa at home have had 14, 9, 9 and 8 corners and Swansea away have had 7,4,7 and 12 so that means 6/8 matches have been below 9.5 corners.
6/8 = 0.75 (75%)
1/0.75 = 1.33
Using the past games the price should be 1.33 but obviously there are lots of other factors to consider and it is not as simple as that. The price is just a guide to convert past results into a decimal price.
Indeed, it really is not as simple as that, as the author does at least recognise, and his blog is certainly honestly named. I just find it incredibly irresponsible to even put that 1.33 out there. It's complete nonsense, based on a sample size of four matches, and no context whatsoever - even a beginner would recognise that a Home game versus Manchester City is not comparable to a Home game versus a (for example) manager-less Aston Villa. A 37.6% edge is huge, and indicative that someone is missing something. If it's not the market, it's you.

Are you really confident that your "model" is better than everyone else's? 

Clearly the person asking the question was excited about finding such 'great value'. That the bet was a loser is really neither here nor there. Value bets can and do lose all the time, (well, not ALL the time), but this bet never had anything like 37% in value. 37% on an EPL market? Highly unlikely.  

As Shapeshifter said back in 2014:
I know some excellent sports data analysts, some who have worked with firms both here and America. It is a whole different head-space. One I have known since high school, originally he dove heavily into the stock market then taking his love of numbers and incorporating it into sports. Conversations with him have taught me the one thing that is of utmost importance: each piece of information has to be put into context.
The big problem with Simple Stats model is the reductio ad absurdum that such a complex and random problem as total corners can somehow be be turned into something easy. 

2 comments:

leon pidgeon said...
This comment has been removed by the author.
leon pidgeon said...

Apologies for deleting the original comment, which was to do with it linking to my account.

I normally enjoy reading your blog as it's regularly updated, interesting and intelligent. I also have a chuckle at your sharpness but not this one as I see my site is on the receiving end. The one plus point is I know someone actually reads what I post on the blog.

Yes, 4 games is a ridiculously small sample size but all these stats are taken from pages on the website and they provide these stats whether a team has played 1 match or 45 matches and it's automatic. For me the important phrase in my post was

"The price is just a guide to convert past results into a decimal price."

That is the reason I provide these equivalent odds on my site so a user can see what a percentage is in terms of odds. If something has happened 57% of times then I think it is useful to know that equivalent odds for that would be 1.75. It probably helps when making decisions on whether to bet or not and saves time for site users. None of the stats are recommended bets and I did point that out to the user. I don't see how that is irresponsible.

Back to enjoying your other posts now.