Week 7 By The Numbers
Last week was far better than the last few for me as I got back into triple figure returns with 111 points. This week however we are plunged into a very tricky fixture list where there’s no real “giants vs. minnows” matches and I’m nowhere near as confident as I was seven days ago. The matches are so balanced in fact that the goalkeeper most likely to “win to nil” according to the bookmakers is Mignolet – That’s right, a goalkeeper that has shipped 1.5 goals on average per match looks to be the strongest one – hmmmm……
This even weighting, coupled with the fact that some goalkeepers are now ridiculously cheap, means that that there’s loads of goalkeepers thrown up by the system; Mignolet, Hart, Schmeichel, Courtois, Fabianski, Adrian, McGregor, Green and Krul. That’s nearly half the keepers in play and this week that means that the matches are all equally difficult, not that all of these goalkeepers have improved! I’ve decided to limit the systems selections going forward in situations like this to a top 5 selections and each goalkeeper must have over a 31% probability of winning to nil. Our new list then includes five players; Mignolet, Hart (swap for Caballero if necessary), Schmeichel, Courtois and Fabianski. Personally I think that every manager is going to need to be lucky to choose a goalkeeper who achieves a clean sheet this week, but if I have to put one above the rest its Fabianski because he’s well priced, at home, he’s proven his ability and he’s facing Newcastle (with or without Pardew?).
Predicting Football Results
The majority of weeks this stats based column will focus on ideas I’ve had however this week I wanted to touch on an excellent study by a chap called Joe Buchdahl. It’s a brilliant piece of analysis and why would I want to reinvent the wheel? For those of you who would like to read the paper itself, you can find a link to it here (Predicting Football Results).
In a nutshell, Joe went back through 16,402 games worth of data to compare the performance of teams based on their current match form. Current match form is decided by subtracting the number of goals a team has scored over the last six games by the number of goals conceded in the same period (goal difference over the last six games basically). Once this is done, the match is given a match ranking by subtracting the current match form of the away team from that of the home team. Let’s give you an example:
Chelsea have scored 19 goals over the last 6 matches and conceded 7 giving them a current match form of 12. Arsenal have scored 11 goals in the same period and also conceded 7 giving them a score of 4. The match ranking then is 8 (12 minus 4).
Joe then went on to model the results of those matches from the 16,402 where the match ranking was 8 and he found that those matches ended with a home win 59% of the time, a draw 25% of the time and an away win 16% of the time. For the geeks who read this, the formulas for calculating the percentages are (x = match ranking):
Home: 1.56x + 46.47
Away: 0.03x2 – 1.27x + 23.65
Draw: -0.03x2 – 0.29x + 29.48
These formulas are based on lines of best fit and whilst I’ll admit that the distribution gets a little wild towards the ends, that’s to be expected as there’s a far larger population towards the centre of the bell curve.
For those non-geeks out there that have managed to get through the last few paragraphs (I hope most of you), here’s this week’s games and the probabilities of each result shown graphically. You’ll note that I’ve overlaid a conditional format where the greatest chance of success is the darkest green and the strongest chance of defeat the darkest red.
Looking at these predictions it would appear that Swansea, West Ham, Chelsea and Leicester have the strongest chance of success and should be good places to find fantasy players likely to perform. I’m sure many of you would be surprised by Arsenal’s extremely low (16%) probability of success but I hate to say it, Chelsea are just that good. I know many will say that this format strongly favours the home team but I don’t think that many managers realise that you’re roughly twice as likely to win at home than away (50%H 25%A 25%D).
For those of you who like a bit of fun, the absolute best match-up we could have hoped for this week would have been Chelsea at home to QPR with this approach forecasting a whopping 79% chance of a Chelsea win.
That’s it for this week, let me know if this forecasting tool is useful and I’ll look to update it on a weekly basis. Oh and before everyone says it, I know that it isn’t updated on my site :-)
See you next week.
