Friday, 16 January 2009

It's The Quality Of The Statistics Stupid - A Flashback

Once per month when the avalanche of projects simply becomes overwhelming, we resort to chicanery and reproduce an old post rather than undertake anything so time consuming as creativity.

But, we try to make the choice pertinent.

Today we reprint a post that was originally put out on December 30th 2007, just prior to the onset of the January Transfer Window Twilight Zone.
Parts of the post are fucking dire, but there's enough stuff of interest to give it another run out...

The current Twilight Zone is even more twilit.
Manchester City are buying anything that moves via a financially depleted, yet evidently still robust, sovereign wealth-fund operation called Abu Dhabi United.
Nobody else is buying anybody for anything.
The market has collapsed.

Excellent.
This will lead to yet more rigged matches as agents are forced to further focus on corrupting match outcomes rather than similarly corrupting the Twilight Zone.

A Wayne Bridge to Nowhere...
_____________________________________________________________________________________

Today, there are two posts in that broadsheet bastion of the bookmaking fraternity called The Guardian that focus on Sabermetrics. This prompts us to produce an overall assessment of the value of such "science" in modern day football.
To the uninitiated, Sabermetrics is a statistical evaluation of a sportsplayer based on selective and historical match data together with current training ground measurements. These analyses are extensively used in English football by the likes of Sam Allardyce, Aidy Boothroyd and Damien Comolli. As ever when the non-scientist approaches scientific data, there is much tomfoolery around. But, correctly used, Sabermetrics is an extremely valuable tool.
Sabermetrics was originally the analysis of baseball through objective evidence, especially baseball statistics. The term is derived from the acronym SABR, which stands for the Society for American Baseball Research. It was coined by Bill James, who was among its first proponents and has long been its most prominent advocate known to the general public. In the US, a leading advocate of such analyses is Billy Beane, the general manager of the Oakland A's baseball team who repeatedly manage to finish in the higher reaches of the tournaments while having the lowest wage bill in the league.
As Sabermetrics has two core constituencies, the value of a player and the current form of a player, we will take these two areas in turn. Lets start with the transfer market.
"We too often reward the lucky strike and overvalue randomness" states Beane and, indeed, there are a whole load of players who live on one "magic" moment - think Michael Owen's pre-gambling-career goal against Argentina or John Barnes' genius against Brazil. The key issue here is value. Players are effectively commodities in the financial market that is football. As with all market prices, the value is based on historical information. But the value of a football player differs from the value of a stock in one very important manner - nowhere near all the information is in the price. Market efficiency, as this is known, is a vital component of market analysis. If financial markets were strongly efficient, no analyst would be able to make a profit as all the information would be fully incorporated into the market price. Fortunately for my bank balance, financial markets are only weakly efficient allowing skilled operators to detect value and arbitrage opportunities in the marketplace. The football transfer market does not even reach the benchmark of weak efficiency. Players used to be assessed solely by eye or on the word of trusted scouts but the evaluation now utilises the statistics generated by companies like ProZone, Opta and Sports Universal. And herein lies one of the problems. The statistics are simplistic. There is no differentiation within the data. So, Opta measure a Ray Wilkins-style 1.5 metre square ball as equivalent to a Juan Sebastian Veron 50 metre switching quarterback delivery. When coaches, agents and managers make an assessment of a player based solely on pseudo-statistics, they are using the data to create invalid structures. The outcome is contrary to the declared aim - value is destroyed rather than created. For example, Comolli used Sabermetrics in determining that Darren Bent was worth £17 million in the summer - although adequate, Bent's real value is probably around one-third of this price. The highly skilled operations, like Arsenal, utilise proprietary analyses based on Sabermetrics together with the intuitive assessments of the management team. This approach undoubtedly is the preferable route to success and Wenger has spent a nett £5 million in the same window that Abramovich has forked out a quarter of a billion pounds. And which of these two teams has the more sustainable squad? Evidently, not Chelsea...
Sabermetrics has a further problem with regard to the transfer market. We could, if we chose to, name around twenty players in the Premiership who incorporate gambling interests into their playing performances. To correctly evaluate such an individual, one is required to use enhanced statistics, gut feeling and the correlation between performance and betting patterns in the global markets. Once again, the slickest are already scratching the surface of this holistic approach as was demonstrated by the total lack of interest from any of the major Mediterranean teams in ############# when he was being offered around last summer or when all the top English sides decided to pass on ##############'s availability after Real Madrid - a player at the peak of his career rarely makes a move from the biggest club in the world to the biggest club in ############### willingly.
Additionally, Sabermetrics fails to pick up on the solo rogue operator. Whereas the most criminalised players are retained by the leading bookmakers, the loose cannons corrupt to a personal agenda. Step forward ####################################
############################################################################
#############################################################################
################################################################################
################################################################################
#############################################################################
#########################################################################
Just ask yourself, what value are Opta's (or indeed anybody else's) stats on these games? The player evaluation must be holistic otherwise the idiocy of Comolli and his like will continue to be the norm away from the very top outfits. An additional benefit of the correct utilisation of Sabermetrics is created by arbitrage over a time window. This is Wenger's core competency - buy cheap, sell rich as he has repeatedly shown with the likes of Henry, Petit, Overmars, Vieira, Anelka etc.
The second area where Sabermetrics are widely abused is in team selection. Although there is value in measuring a players training performance cycles, it should merely form one tool in the box. Morale, motivation, private life, grudge, interpersonal issues etc etc are also relevant but the core problem with such naive assessment is that it individualises a team game. The core competency in a team sport is producing a scenario where the complete team performs at a greater level than the basic sum of its parts - team spirit cannot be measured nor created by measuring a cycling machine. Furthermore, when players realise that their place in the starting eleven is dependent on running around like a blue-arsed-fly in training, they effectively undermine the team effort by putting in too much effort in training in midweek with the result that the side selected for the weekend game is physically more tired than the players who took it easier in training. This impacts negatively on results. Such methods also require a hierarchy sympathetic to the use of Sabermetrics. Allardyce undoubtedly had such an environment at Bolton but the ego's, gamblers, cliques and general dysfunctionality at Newcastle is an altogether different beast. If I were Big Sam, I'd sell ############################################################ and ############## in the January transfer window and use Sabermetrics to bring in young committed and non-corrupt players at a fraction of the cost (both in transfer and wages). I have no doubt that results would improve.
Of course, the media plays a role in this misuse of pseudo-statistics. As does aesthetics...
We are all persuaded by a credible source rather than a credible argument and, so, when journalists wax lyrical about the sublime touch of Dimitar Berbatov, sending his value ever upwards in the process, they choose to ignore the fact that the man often selects to be a liability (as Martin Jol discovered to his cost). Additionally, the media promotes those who it is in the interests of The Game to be promoted - the number of column inches associated with the 20 Premiership players who we know to be working alongside bookmakers is far out of kilter with the performances of these players on the pitch. And, as we have stated before, the Guti and The Beast Syndrome results in a homoerotic element to the evaluation - we simply do not believe that being square-jawed, or having symmetrical features or flowing locks has any influence on match performance. Of course, the beautiful are known statistically to succeed more than the aesthetically challenged but, when making the assessment of a player's true value, such factors should be sidelined. Who would you rather have in your team, Paul Scholes or Guti, Franck Ribéry or David Beckham, Carlos Tevez or Berbatov? And, although a true great, Paolo Maldini was never quite as good as the length of his eyelashes would suggest. There is, indeed, considerable evidence that Black and red-haired players are grossly undervalued when compared to the standard anglo-saxon features...
Around 50% of the teams in the top two English leagues utilise Sabermetrics. As with all analyses, it is the inputs which produce value. Give a neural network to a monkey and you'll get Watford; give it to Wenger and you'll get Arsenal. If Arsenal were to incorporate the machinations against their interests by the English institutional powers into their holistic Sabermetrics, they would actually have had a chance of the title this year - as we showed in our last post, Arsenal would have been eight points clear at the seasonal midway stage without the effect of corrupt match officials. But, in the criminalised world of the Premiership, there are many necessary skills that have too little to do with football and too much to do with the global gambling markets.
There is one other prime problem with Sabermetrics. By applying an analysis that aims to effectively play the percentages, the coaches who develop simplistic Sabermetrics are unable to adapt to the tactical requirements necessary to win the really big games. And it is these matches that determine trophies. Although he is past his sell-buy date now, a couple of years ago, Barcelona would rather select Ronaldinho, even though he had spent the night samba-ing around the Ramblas, than any other individual who had happened to have had a particularly good training session on the rowing machine. Barça won the Champions League and it is interesting that Beane's baseball team always falls short of ultimate success as Sabermetrics, while perhaps improving the chances of avoiding massive underachievement, similarly fails to engender overachievement.
Dietrological intend to launch a new business next summer. Provisionally entitled Holistic Sabermetrics, we will be offering a thorough and complete statistical assessment of both individual and team realities based on our proprietary data. We are confident of our success in this venture - there is nobody else out there with our databases.
Someday soon, somebody is going to be persuaded by the pseudo-scientific data to pay £30 million for Berbatov. They should probably chat with us first...

Unfortunately, the blocked out parts of the posts and all posts labelled # are available in full to Dietrological or Football Is Fixed subscribers only.
But, please enjoy the free snippets that we are posting on this site.

© Football Is Fixed/Dietrological