Paralysis By Analysis

I came across an interesting article today and before I go any further with it, I thought I’d post the link:

http://www.soccerbythenumbers.com/2012/02/dynamics-of-relegation-in-premier.html

It’s an interesting read, but I’ve got a few issues with it. I’m fairly certain that it’s by an American author and mathematician but whoever wrote it is bringing some unintentional cultural baggage with them. With the exception of the NHL (which still uses two points for a win), the two big American professional sports leagues (the NFL and Major League Baseball) don’t award points for wins but rely on percentages to determine league positions and none of those competitions are based on a promotion/relegation model.

The second aspect I’ve noticed about how Americans approach sporting analysis is that there doesn’t appear to be as much importance placed on differences between home and away records. A team could easily have a decent home record combined with an appalling away record and find itself in danger of relegation. Before they had ten points deducted for going into administration (again), Portsmouth were a good example of this and there are several examples in the less well known European leagues.

The third issue isn’t as important in some respects but the post is written from an overly mathematical perspective – as if displaying mathematical knowledge is more important that the application thereof to the subject matter. There seem to a number of American statisticians looking at association football at the moment and I think they may all be either trying to find some kind of all encompassing ‘Moneyball’ theory or – conversely – over complicating things in an attempt to intellectualize football in order to make it appealing to the middle classes. As a bettor, I look at promotion/relegation issues every week or so and I’m guessing that the author has neither the inclination, opportunity or (and this is the most likely factor) confidence in their assertions to part with their money.

Specific Issues Related To The Article

First of all, it is possible to carry out an analysis of the types of teams that face relegation each season. This method is based on a similar one that Kevin Pullein uses in his book to identify potential FA Cup winners. Here’s how you do it.

* Look back at the final tables for the Premier League for every season since it started in 1991/92.

* Beginning with the final table for 1992/93, make a note of where the relegated clubs finished at the end of the previous season.

* Then work out how many points the relegated teams earned per game.

I’ve used this method to predict both promotion/relegation for a couple of seasons and it’s not far off although I’m still trying to assess how early/late in the season. It’s how I decided that Helsingborgs would win the 2011 Allsvenskan. It’s also formed the basis of a number of unsettled bets I have in the promotion and relegation markets in the Football League this season.

Unfortunately after that the post starts reading like a parody rather than a serious piece of research. The gist of the section entitled Offensive Production is basically that teams that were relegated scored fewer goals. It’s the same with the section entitled Defensive Production which informs us that relegated teams conceded more goals. As those of us that have followed football for a while will have noticed, if you concede more goals than you score, you lose the game and receive no points. If you finish with a low amount of points you tend to get relegated.

The killer blow is that Soccer By The Numbers have only used last season’s data to come up with their hypotheses. Could they explain Derby’s disastrous performance in 2007/2008 or Ipswich Town’s catastrophic decline between May 2000 and May 2001? Or how Birmingham scored 18 fewer goals than Blackpool last season (and conceded one less than Aston Villa) but still finished above them even though both teams were relegated?

Let’s also not forget that Arsenal kept fewer clean sheets in away games than Blackpool did last season. How does that fit? Might that stat explain why Arsenal seem to have declined over the past few seasons?

In the spirit of co-operation…

I don’t normally reveal my methods but I’m going to make an exception here. I compiled the data I mentioned above and these are the general trends, so without further ado, here’s how you can tell if a team might be possible candidates for relegation from the Premier League:

* At least one team promoted from the second tier was relegated the following season. Only once in the last sixteen seasons have all three promoted teams been relegated at the end of the following season.

* Since the Premier League switched to 20 teams (38 games) in 1995/96, only nine of the 48 relegated teams earned an average of one point or more per game. All of the teams that finished bottom earned less than one point. That’s something that can be worked out every week with either a piece of paper and a calculator or a copy of Excel with a knowledge of conditional formatting: it’s also applicable to the last ten seasons in the Championship.

* Ipswich were relegated in 2001/02 despite having finished 5th in 2000/01 – the highest placed club to have been relegated after the following season. As a general rule of thumb, any team finishing in 12th place or lower in the Premier League is at risk of relegation the following season but it’s worth remembering that 7 of the 32 clubs that have been relegated since 1994/95 finished in the top ten at the end of the previous season.

There is another important stat that can gauge how well or badly a team is performing ‘in season’ but I’m not giving that away. Let’s put it this way: if you can outscore your opponents you’re going to perform well regardless of any outliers or friction (von Clausewitz’ definition of chaos and chance) and reduce the chances of Black Swans as well as your chances of avoiding relegation.

So who’s in trouble this season?

It won’t come as a massive surprise that the current bottom five all fit the profile – however, QPR aren’t in the bottom three at the moment but if one promoted club is practically guaranteed a relegation spot then it’s likely to be Mark Hughes’ team rather than Norwich or Swansea unless both of them experience a catastrophic regression between now and the end of the season.

Of course, it’s not out of the question that none of the promoted teams might be relegated at the end of the season. Just because it’s not happened before doesn’t mean it’ll never happen, but as I’m now moving away from the point I’ll end there.

 

>Saturday Round Up

>Well, not bad for a single day considering the less than spectacular January I’m having – this is where good record keeping comes in as there are a number of holes in my game that I’ve identified and I’m going to spend the rest of the month plugging them.

For example, I definitely need to change my approach to goals markets. Half of my losses for this month have been in this area: in fact, only I’ve won only four of seventeen bets on goal markets this month, which is pretty dreadful. I need to stick to over/under 2.5 and not get involved with any of the more ‘exotic’ markets – ‘both teams to score’ markets are a good example of a quick route to the poor house.

In a low scoring league such as the Greek Super League this season – and not counting today’s games – the ‘both teams to score’ numbers look like this:

139 total games, both teams scored in 55 of them – 39.56%. About half of those 55 games ended in draws. It’s therefore reasonable to assume that in about 60% of games in Greece, both teams will not score. Furthermore, only 43 of those 139 games featured over 2.5 goals: that’s 30%. So a ‘typical’ Greek game is likely to be under 2.5 goals with either one or neither side scoring.

In comparison, the numbers for Portugal – where there are slightly more goals per game but the majority of games are still under 2.5 – are as follows: in 123 total games, both teams scored in 58 of them (47%) and only 20 of those games were draws (34%). So once again, it’s reasonable to assume that in a majority of games in Portugal, both teams won’t score although it’s less clear cut than the Greek example.

Armed with that information – even though it’s not comprehensive and comes from two relatively goal shy competitions – it should help my decision making in goals markets.

NFL games previews will be up later.