Wednesday, 23 July 2014

Ratings For Anoraks

I received a couple of comments on my recent post which compared my ratings with Statto's ratings. As I mentioned, I used Stattos' ratings as the starting point for my rather ambitious plan to track the lower English leagues down to the Conference in addition to the top five European leagues.

Fairfranco asked:
Out of interest do you know if Statto's ratings have been of any use over time?
Would be good to see historical results with ratings at the time alongside them to get a grasp on some accuracy.
Matthew T added:
I would also be interested Padrino if you had any experience regarding the validity of the Statto ratings. Once upon a time I used to paper trade a range of ratings system while betting my own but it became too time consuming. If I remember correctly I found the Statto ratings were very weighted to long term quality over short term form and I never could devise a method for incorporating their trend rating in to a betting system. They seem very similar in style to the Racing Football Outlook newspaper football ratings and both seem to have their origin in Elo. They seemed lackluster in the divisions where more complex models were rife e.g. EPL and better in the lower class European leagues. For top leagues a good free option remains Dectech for Poisson style goals for and against ratings.
The short answer is that I do not know if Statto's ratings have been of any use over time. They claim a pretty good record for them, writing:
The Science Behind Statto Index
The Index has been developed and tested over many years, starting in the 1980s before the World Wide Web era. Our testing uses around 25 years worth of results and approximately 10 years of bookmakers odds to test the accuracy and profitability of the system.
Naturally, no system is perfect and we continuously monitor its performance. For the mathematically minded, the correlation between our forecasts and actual results is just over 0.9, which means that approximately 90% of the final outcomes are explained by the prior forecasts.
Naturally, no system is perfect, but 90% isn't too shabby, although I think we need a more detailed explanation of what exactly "final outcomes explained by the prior forecasts means"!

The forecast odds on this site are produced by our Forecasting Engine using our unique Statto Index ratings. The Index currently covers 29 leagues together with all International teams - a total of over 750 teams. Our forecasted odds are based on a true 100% book and therefore do not take account of any commission or projected profit margin - this is left to you based on personal preference. Odds are forecasted for both match and outright markets.
I agree with Matthew that their ratings do appear to move rather slowly, and their focus appears to be more targeted at the anorak / weatherman types rather than someone looking to make money from them.

Having said that, I have to admit that I have a little of the 'anorak' in me myself.

A recent DNA analysis threw up some surprises in my ancestry - along with the alarmingly high Neanderthal percentage shown above, and the less surprising majority British and Irish percentage shown below, there was a small percentage of 'train-spotter' DNA detected. Where that came from is something of a mystery, although a great-great-granddad - in other words a "really fantastic" granddad - did work with the trains on Ryde Pier, Isle of Wight circa 1850.

Anyway, it's my excuse for finding the Historical Ratings from 1888 quite interesting.

The best (highest rated) English team of all-time is Chelsea (2006) who reached 995 points, just 35 points ahead of Aston Villa's historic best, which unfortunately for them was not last season, but came in 1897. Just two League clubs reached a new high in 2014 - Manchester City (960) and Burton Albion (706). At 958, City are in with a good chance of setting a new high this season - they don't face Crystal Palace until December - but whether it is possible to make money from such ratings seems unlikely.

Matthew mentioned Dectech as a good free option, but I remember questioning their sanity when I saw they had Norwich City predicted to finish fifth in the EPL last season with 68 points. They got 33, and were relegated of course.

Admittedly I had Norwich City avoiding relegation at the start of last season too, but didn't go so far as a top half finish, never mind chasing a Champions League spot. Perhaps Dectech reacted too much to Norwich's meaningless last day 3-2 win at Manchester City the previous season?

I mentioned earlier that I agreed with Matthew's observation that Statto's ratings move too slowly, at least from a betting perspective, but the other side of the coin is to be careful not to overreact to a result such as the one just mentioned. Norwich were never contenders for a European spot of any type based on league performance.

Thanks to all of you who have renewed or subscribed for the first time to the Cassini Service. The early-bird special offer of £129 (or £99 for renewals) is good through the end of this month. PayPal to calciocassini @

I'll re-focus on the Friendly Tipster League this weekend, and see if we get some prizes / bonuses from the likes of Ian Erskine. Watch this space.

No comments: