The headline, untouched by M Salut’s hand, is noteworthy for the absence of Sunderland from the list of contenders for the last of the bottom three slots. We must all (save for visiting supporters of the other affected clubs) take comfort from John McCormick‘s scholarly ways, sincerely hope he knows what he’s doing and be assured he is not tempting fate … it is the latest of his studies of how fluctuating goal differences may affect the outcome of the pressing Premier issue that remains to be resolved following Man Utd’s confirmation of the title
It’s now a while since Christmas and the beginning of my exploration of goal differences. You’ll remember I predicted even then that QPR and Reading were going to be relegated, but they weren’t my focus. What I was looking for was some pointer as to which one of the throng of clubs hovering around the drop zone would be joining them. I reasoned that it would be the one to show a consistent decline in goal difference in the second half of the season, with others at the bottom showing a greater improvement.
This was always speculative and by Mid-March it was clear that my reasoning had been faulty. Goal difference trends are no way to tell who will go down and who will stay up. Why should anyone expect any different? A string of 9 games that consist of narrow losses interspersed with a couple of draws will generate a small decrease in goal difference but only two points. On the other hand, some thrashings interspersed with a couple of wins will generate a worsening goal difference but enough points to give a club and fans hope. Does this sound familiar? It should as we were the one whose goal difference remained relatively healthy while clubs below and above us picked up precious points. Reading, especially, looked for a while as if they might escape and QPR, seemingly rejuvenated under Harry and a bundle of transfers, also staged a revival. Scary times, don’t you think?
And yet, in mid March the goal difference trendlines I was using pointed to both Reading and QPR as still going down and one of Wigan or Villa joining them, much as I had predicted after only five games, when there hadn’t been many results to compare. I resolved to carry on and see how far a discredited method would take me.
So let’s begin, as ever, with Sunderland’s goal difference since the summer:
You can see the unspectacular but jaggedly consistent downward trend had turned into an almost fatal plunge until the transformation brought about by Paolo di Canio. Would we have become the third team to go down? It’s a moot point. No one will ever know but I think few fans are now complaining that the change was made.
Even so, as the next chart shows, our actual goal difference was not too bad in comparison with other clubs at the bottom:
A close look at the right hand side of the chart reveals the four clubs who currently occupy the bottom places – QPR and Reading plus Wigan and Villa – are somewhat adrift of the pack. Then come NUFC and ourselves plus Southampton and Fulham.
Fulham are safe by any reckoning, not because they have played well since Christmas but because they had enough points on the board for even a mediocre second half to generate enough additional points for safety. Southampton, on the other hand, did turn their season around. Stoke and Norwich aren’t shown; their declines started after Fulham’s and after I’d selected the teams for my focus.
However, it wasn’t goal difference since the summer that interested me, it was how goal difference changed after Christmas. So here, once again, are goal differences since then, with all of the clubs reset to zero:
This has QPR, Reading and Wigan still at the bottom, with NUFC replacing Villa, but it’s a bittoo jumbled to pick out patterns of decline. I’ve been using trendlines to simplify the charts so here are the latest trendlines for these clubs, with those for January- March given first for comparison:
Join the Salut! Sunderland Facebook group – click anywhere along this line
And follow us on Twitter: @salutsunderland … click along this line
In March there were three very clear groups of trendlines. NUFC and Fulham’s were almost horizontal and I reckoned those two were safe. Reading and Wigan both had fairly steep slopes and Southampton, QPR, Villa and ourselves were somewhere in between. I decided from the slopes and the R2 values (a measure of how well the trendlines fit the graph points, which is no measure of reliability) that QPR and Reading would not be able to catch up, that Wigan would accompany them although it would be a close contest with Villa, and that Southampton and SAFC would be the next two clubs at the bottom.
The current league table tells me I was wrong about Southampton and not quite right about Newcastle. It’s hard to say whether or not I would have been right about us had things not changed. The current trendlines, shown below, tell a similar story. There are still three groups but they are not quite the the three groups we had in March. Fulham are still almost horizontal, and Southampton have gone from pointing downwards to upwards but NUFC have joined the middle group with a downward gradient. Their position has become more precarious in the last few weeks. Villa, too, are in this group but their trendline is not going downwards so steeply. For me, the trendlines still point to QPR and Reading, and also Wigan (unless they beat Villa) going down, with NUFC and ourselves still not able to say we are safe:
Of course, I could have reached this conclusion by looking at the league table on Sunday, in which case I’d also be able to comment on Stoke and Norwich, but I don’t think I’ve done too badly over the last few weeks. My system might be rubbish, but it has been fairly consistent in mirroring changes in the league table and, with four games to go (five in the case of Wigan) it’s as good as anything I’ve seen on Match of the Day.
Which comparison leads me to a final conclusion. The predictions made by experts on Match of the Day are no better than those made by someone who looks at irrelevant data and applies faulty logic and mathematically incorrect methods to them. But I do acknowledge the MOTD team’s expertise in other aspects of the beautiful game – I remember one of them trying his hand at management and he did a brilliant job.
* Monsieur Salut’s whimsical look at ESPN at how two players from overseas, of different generations and with styles of their own, have been crucial elements in Sunderland’s battles against the drop: Sessegnon is the new Reyna as PDC steers a Sunderland course for survival
[Sess’s goals vs NUFC and Everton] may turn out to worth many millions in Mr Short’s business calculations, and gold to some of the most passionate and least rewarded fans in the Premier League …
If Sessegnon has more natural ability, Reyna is rightly remembered for his tidier, technically sound approach to football. The former Nottingham Forest and Manchester United striker, Garry Birtles, now a television pundit, said during Saturday’s game that Sessegnon, for all his skill, often made the wrong final decision; that could rarely be said of Reyna.
7 thoughts on “Will it be Wigan or Villa (or Stoke or Newcastle) to go down with Reading and QPR?”
That’s a very good point about the 3 points Geoff. If nothing else the GD trend that John has measured is in many ways a better predictor or record of “form” in the purest sense than anything else we have at our disposal.
Far more use than the WDLLW sort of stuff that the papers churn out.
I think this is an excellent way of measuring performance over a whole season. I wouldn’t go into such detail but the point about goal difference is that everything is compared on an equal basis. In the old days of 2 points for a win points were a more reliable factor in this sort of prediction but the problem now with 3 points for a win is that we’ve got away from the situation where each game played was worth the same number of points.
I know that I’ve been desperately hoping AFC Wimbledon stay up. But I’ve been very aware of their goal difference telling me they were always in a slightly false position. Now they are in deep trouble. I was also expecting Huddersfield to fall a bit further than they have. Even so they may still go down which I believe they deserve to do.
Goal difference might not be an accurate way of predicting who will finish where but it’s an indicator that we ignore at our peril.
I reckon Wigan’s luck will finally run out and they’ll go down.
Although Stoke had an extremely poor form run and were looking pretty precarious until their last win at QPR.
Aston Villa just have enough to stay up. Newcastle will be safe.
We’ll pick up points v Villa & Stoke and be OK.
John said “any half decent mathematician would be leaving the site shaking her or his head at what I’d done.”
Maybe so John, but it’s a belting article regardless. We must remember that any statistics are being applied to something which is not scientifically quantifiable in the same way as weather patterns or the effects of topography on precipitation etc.
What the trend will show (particularly as it relates to several months of data) is that performances or team effectiveness (which is what GD provides measure of even if it is distorted) are worsening and who is worsening at the greatest rate.
I don’t think that you answered the question in the title but I assume that was Colin’s header rather than yours. Most people I talk to think that its between WIgan and Villa now, but as your trend shows and my heart believes, the Horse Thumpers are a long way from safety. They have a couple of really nasty home games and what is going to go a long way in the final outcome is the final day fixture between Wigan and Villa. Only one of them can come out of the final day with 3 points. A draw will probably be of no use to either side and the next couple of weeks are going to tee that game up a treat. The nerves will be jangling on Tyneside if they get beat at the weekend. A loss to Aston Villa might not be a terrible result for us on Monday night you know. 😉
Apologies to anyone who wanted to comment, subsequent to those posted above, but couldn’t. The ‘enable’ box had somehow become unticked. Readers posting for the first time, eg supporters of the other relegation zone clubs, may experience a short delay because of the anti-spam filter but one of us usually approves pretty quickly.
One of the problems with my method is that it hasn’t even been looking at raw data but has been using trendlines. These aren’t designed for what I’ve been doing. At one stage NUFC’s goal difference produced a more-or-less U shaped curve. The trendline was horizontal, across the middle from one vertical bar to the other and came nowhere near showing what was going on or what the next few games might bring. And then, on top of this, I used R2values to give an idea of how accurately the trendline was predicting the future. R2 measures how near the trendline is to the actual points on the curve. No wonder it was near zero with NUFC’s and some others.
GD trends were interesting because they did mirror what went on in the table and I enjoyed pursuing the theme – I might even revisit it at the end of the season – but really, any half decent mathematician would be leaving the site shaking her or his head at what I’d done.
Its an interesting piece of analysis but the problem with goal difference is that it doesn’t accurately reflect the potency of a team or how leaky it’s defence might be. It simply compares the numerical difference between the two raw pieces of data. It does yield some bizarre trends though.
Inspired by your article to look at the Championship table John there are some noticeable “anomalies.” Ipswich lie 14th with 57 points but with a -13 GD. Burnley who lie a place beneath them with a point less have a GD of only -1. Even more peculiarly Huddersfield are 19th and have a GD of -22 which is far worse than any of the three teams in the drop zone. Wolves, Barnsley & Bristol City are -11, -16 and -20. Just to add to the confusion, Birmingham who are 6 points short of the play offs have a -4 GD.
So what does all of this tell us? You might think that it tells us nothing, but in fact it tells us a great deal. GD does show a trend. It shows a trend in terms of a team’s “effectiveness” or overall performance because no other stats apart from actual points provides that.
Here’s what I mean. If you start to look for trends elsewhere, they are difficult to find. Sticking with the Championship, bottom club, Bristol City (who I hope go the same way as Coventry City) who are already relegated have scored 39 goals at home this season, which you will be surprised to know is more than either Cardiff who have gone up as Champions or Hull CIty. Where they have fallen down is in conceding a lot of goals at home. They have the worst record for conceding at home letting in a whopping total of 41.
Now where it gets really interesting (and assuming that you have stuck with me this far…..) is that they also have the worst record in the division when it comes to scoring away from home (18), and yet their defensive record away from Ashton Gate is comparable to anyone else in the table including Bolton who are in the play off zone. In fact their defensive record is better away from home.
Goal difference and the trends that John is looking at are very useful indicators and that is why GD (as a raw stat) is used as a real simplification in terms of clarifying what is going on.
The bottom of the PL has been so tight that two weekends can completely transform the relegation scenario, apart from Reading and QPR who are virtually doomed. Reading at one point started to collect a few points but John’s trend analysis shows that it was a false dawn for them.
Looking at this type of thing would drive you mad as a manager though!
Comments are closed.