In his efforts to perfect a new ratings' system for the Premier League, Jack Houghton has discovered, much to his disappointment (and Einstein's), that goal difference matters, which bodes well for Arsenal away to West Brom on Saturday...
"At the time, I liked to quote Einstein’s supposed (but actually falsely attributed) maxim, “as simple as possible, but no simpler”"
A win is a win, isn't it? Manchester City's 4-1 drubbing of Spurs in October is no more impressive than Stoke's tightly contested 2-1 win over the same team, is it?
A number of years ago now, on this very site, horseracing guru Simon Rowlands and me had an ongoing debate about whether having a method of assessing horseracing trainer form was useful to a punter. Simon had come up with an (as always) ingenious and statistically robust way of determining which trainers were enjoying a purple patch, and which were to be avoided.
I had no problem at all with his method. My concern was whether it represented double counting. If you were already assessing the form of horses in a race using some kind of rating system (in my case, speed ratings), weren't you already taking trainer form into account? By then overlaying another measure of trainer form, weren't you confusing things?
There's no right or wrong answer to this debate, of course. Which side you come down on has more to do with your philosophy on ratings systems, as much as anything else. At the time, I liked to quote Einstein's supposed (but actually falsely attributed) maxim, "as simple as possible, but no simpler". Assessing past, and predicting future, performance was all about finding the purest, unadulterated numerical form that I could.
Creating my own Elo ratings for football this year has seen that underlying philosophy challenged in a few areas, none more so than when it comes to the question of whether you should adjust your ratings based on the margin of victory.
Initially, my instinct was to ignore it. After all, I'd grown up in an era where Arsenal would grind out 1-0 victories and anything more than that one goal was considered gluttonous. At the time, other teams would beat the same teams by bigger margins, and yet it was Arsenal who were clearly the better outfit.
It also didn't make rational sense to me that the margin of victory told you much about real superiority because, after one goal is scored, the tactics of a game change significantly, usually with the losing team taking more risks, which can surely lead to unrepresentative goal tallies.
Unfortunately, my instincts turned out to be wrong. The first clue came in realising that every other football rating system I could find gave extra value to those teams who chalked up bigger goal differences.
Then followed a very simple piece of analysis. In just about every previous Premier League season that I looked at, the top-ranked team also had the biggest goal difference. What's more, the goal difference in most seasons runs fairly sequentially from biggest positive to biggest negative as you look down the table. If goal supremacy wasn't important, then you would expect to see a more random pattern.
Finally, I started playing around with some data which looked at how teams with similar goal differences across various seasons had faired against each other. If margins of victory were important, I reckoned, then you would expect the percentage of home wins, draws, and away wins to be broadly similar to the average percentage of all games. And lo and behold, they were.
Which is why I'll be backing Arsenal away against West Brom on Saturday. Arsenal's indifferent season and weak recent form means they are a generous 1.9210/11, but they have shown an ability to win by big margins along the way.
Back Arsenal to beat West Brom at 1.9210/11