Friday 30 October 2015

Championship Week 13: Are Birmingham legit?


This week marks the anniversary of Gary Rowett being appointed Brimingham City manager and with the Blues sitting in the playoff spots it seems the Championship has been awash with articles about how successful Rowett’s tenure has been.
To top it all there's a big West Midlands derby against Wolves this weekend too.

But are the plaudits warranted?

When Lee Clark was sacked Birmingham were 21st just one point off the relegation zone. By the end of the season they were nicely in the top half of the table having secured 10th position.
With relegation successfully avoided and now an apparent push for a playoff place on the cards this season it seems case closed – Rowett is a success.

Let’s have a look at the numbers to see if they back this up. (I’m using statistics under all situations in this instance to be able to compare last season and this.)

Clark left with an overall shot share (Corsi/TSR) of roughly 46% - his teams took 46% of all shots in games they played, and therefore conceded around 54% of all shots.
Most worryingly the shots on target share was worse at below 42%.
Add to that Clark had seen shots scored and saved by his team at a slightly below average rate and the place just off the relegation zone was rather obvious.


So what of Rowett’s tenure so far?
Last season Rowett’s overall shot share actually dipped below Lee Clark’s, but crucially, his shots on target share lept eight percentage points to just below average. These figures have been maintained so far this season, indicating it may well be a product of Rowett’s tactics.
If we look at the per game figures last season we can see Rowett’s team took and conceded a near identical amount of shots per game, but managed to get 1.2 more per game on target, while only conceding one more on target every seven or so games.
Certainly that is a happy improvement to make.



This season, Birmingham have been tighter at the back, but have also been similarly conservative up front.

The other significant difference between Rowett and Clark is the rate chances are being scored and, most importantly, saved at.
Somehow Birmingham’s GK got 10 percentage points better under Rowett’s stewardship last year, and it has continued at the same clip so far this term.
While the scoring rate was largely unchanged last season, the increase in shots on target gave 0.32 goals per game (almost 11 actual goals) more than Lee Clark would have expected.

This season the shooting rate has seen a 10 percentage point increase as well, meaning Birmingham have the second highest PDO in the Championship at present.

Given this it’s no wonder Rowett’s team is almost unrecognisable from the one Lee Clark was regularly fielding.

So is Rowett the greatest thing to hit Birmingham since Tim Sherwood?
No.

I would expect this massive run of form to cool off at some point (above average shooting and save rates are pretty tough to maintain over a long term – indeed it already has done over the last few games) and Birmingham to drop back into the middle of the pack.
But Rowett certainly has improved the team’s shots on target generation while seemingly not damaging its defensive performance, and that deserves some praise.

Friday 2 October 2015

Championship Week 9: New beginnings

Welcome to the 2015/16 version of The Only Statistic That Matters.
First an apology for my tardiness at putting this out but sometimes life, among other things, gets in the way.

This season I’m collecting a whole load more data which I’m still working out what to do with, what is useful and what can add value to understanding how this wonderful sport we so love works.
One of the big changes I’ve made with the data I’m tracking this year is to use 11v11 wherever possible.
I noticed last year that on occasions there seemed to be big swings in shot stats involving games with a red card (Brentford’s 41 shot mauling of Blackpool being the prime example).
In ice hockey, where many of these metrics originate, the power play/penalty kill (where one team temporarily has a player less) are separated out from full strength (5v5 in that case) play.
Yes, it will make for a smaller sample size, but I’m hoping the sample I’m left with will be more accurate as to the real skills of the Championship teams.

The good point about starting coverage at this stage in the season is that many of the metrics have started to settle down – at least they were doing so at this time last season.
We are still dealing with a small sample size in terms of shots though.
The average team will have taken (and conceded) 118 shots so far, meaning a couple of goals here or freak deflections there can still make all the difference in the table and in the numbers.

I’m also recording where the shots are being taken from, at least in broad geographic on-pitch locations.
These are the standard ones that I’m sure many of you will be familiar with already, but just for clarification we have: inside the six yard box, the central area of the 18 yard box immediately in front of the six yard box, the sides of the 18 yard box, and outside the 18 yard box.
The six yard box and area in front of it collectively being the Danger Zone.
Headers are being recorded separately from shots: inside the six yard box and outside (the vast majority of these take place in the central 18 yard box area so there seems little point in also using the sides).

Season so far

So now that’s cleared up, or at least some of it, let’s have a look at where we are so far.

Pos Team Played Won Drawn Lost For Against Goal Diff Points
1 Brighton 9 6 3 0 13 7 6 21
2 Middlesbrough 9 6 2 1 17 5 12 20
3 Reading 9 4 3 2 14 7 7 15
4 Hull 9 4 3 2 12 7 5 15
5 Birmingham 9 4 3 2 14 10 4 15
6 Cardiff 9 4 3 2 13 10 3 15
7 Burnley 9 4 3 2 12 10 2 15
8 Ipswich 9 4 3 2 15 15 0 15
9 Derby 9 3 5 1 10 7 3 14
10 Wolves 9 3 3 3 12 11 1 12
11 Nott'm Forest 9 3 3 3 9 9 0 12
12 Sheffield Weds 9 3 3 3 12 13 -1 12
13 QPR 9 3 3 3 14 17 -3 12
14 Fulham 9 3 2 4 15 14 1 11
15 Leeds 9 2 5 2 9 11 -2 11
16 Huddersfield 9 2 4 3 10 11 -1 10
17 Charlton 9 2 3 4 9 12 -3 9
18 Blackburn 9 1 5 3 10 10 0 8
19 Brentford 9 2 2 5 12 17 -5 8
20 Rotherham 9 2 2 5 11 17 -6 8
21 Bolton 9 1 5 3 6 12 -6 8
22 Preston 9 1 4 4 6 10 -4 7
23 Milton Keynes Dons 9 2 1 6 8 13 -5 7
24 Bristol City 9 1 3 5 11 19 -8 6


Middlesbrough being in the promotion spots will not be a surprise to many. Brighton being top probably will be, but shouldn’t be. I was cautiously optimistic about the Seagulls before this season given their very good shot metrics last year.
With very good shot numbers again it seems this season they’re actually scoring and saving shots that last year they should have been.
At the other end, I had much higher (mid-table) hopes for at least two of the promoted teams.
But while it may make pretty grim viewing for the trio right now, as we’ll see shortly, only MK Dons are making a particularly bad start, with Preston and Bristol City bitten by horrible luck and variance in and in front of goal.

For the eagle-eyed among you, Reading  (0.681 shot share, 97.78 PDO) and Middlesbrough (0.530 shot share, 126.38 PDO) do not fit on the chart.

Again, Reading  (0.714 shot share, 97.78 PDO) and Middlesbrough (0.531 shot share, 126.38 PDO) do not fit on the chart.


Shooting stars

The early statistical stars of the season are without doubt Reading.
They started off at around an 80% shot share but that was never going to be sustainable and are now in the region of 68-72%, depending on which flavour of metric you prefer.
The secret of this is defence.
The Royals have conceded just 67 total shots – almost half the league average.
And while doing that, they’ve also been very nicely above average at taking shots (137, 3rd best).
That is quite simply superb.
There may be other reasons for this such as the fixture schedule so far, but they have played (and outplayed) Burnley and Ipswich, while they matched Derby in shots despite being down to ten men for half the game.
In fact, Ipswich (in that memorable Friday night hammering) are the only team to reach double figures in shots taken at 11v11 and are the only team to out shoot Reading.
I suspect much of that was due to the massive score difference leading to Reading sitting back.
So certainly at the moment, Reading look like legitimate promotion contenders.
There is still a long way to go, but boy do the early stats make positive reading.


Rotherham

Somewhat lost amid the row about a certain West London team was the news that Steve Evans left (by one method or another) the manager’s post at Rotherham United.
While his personality can rub people the wrong way, last season he did a fine job keeping the small budget Millers up and it was only a horrible PDO score that saw them slide towards the relegation zone.
Their shot metrics were give or take mid-table obscurity.

Unfortunately for Evans, in the summer he was forced to sell some key players without being able to replace them sufficiently and the results this year have been rather ugly.
Rotherham are bottom in overall 11v11 shot share (taking just 43.3% of all shots in their games), second bottom in unblocked shot share (again 43.3%) and third bottom in shots on target share (40.7%).
Getting burned a bit by PDO as well (93.94) meant there was nowhere to hide and so Evans left the club.
Perhaps the real killer for Rotherham has been the quality of shots taken by opponents.
The Millers have conceded 40 danger zone shots (most) including nine from within the six yard box.
The next worst teams have only conceded six close in, with the average being three and a combined 25 from the six yard box and danger zone respectively.

Whether anyone else will be able to improve performances given the talent drain at the club it will be tough to do, but I’d suggest they look at the defensive side of things first.


Let’s talk about Brentford


Well nobody else is, so I may as well give the developments at Griffin Park a bit of coverage, eh?
I had actually really hoped to give Brentford a wide-ish berth this season and just let the club get on with what they are doing and see how it goes.
I had not expected to be writing about sacking the manager in my first post this early in the season.
There's been a lot of heat and not much light in the comments that I've seen about the Brentford situation this week. Hopefully this will add some much needed light to the discussion.

Like just about everyone else in the world I have very little idea of the exact reasons why Marinus Dijkuizen was sacked on Monday.
But it seems fair to look at the numbers and see what’s been going-on on the pitch.
And the numbers do not make great viewing.

At the time of Dijkuzien’s sacking, while the all situations metrics looked just below average, these were buoyed by a positive piece of play when up a man against Bristol City.
In 11v11 it was not pretty. In the bottom eight in both overall shot share and unblocked shot share (47.4% and 45.4%), the Bees were third worst in the Championship with a shots on target share of just 39.7%.
That is a quite remarkable turnaround from last season.
When we look down into the raw shot numbers there seems to be a likely dearth of quality scoring chances.
Although there were 27 shots taken centrally in the 18 yard box (above the 22 average at the time), there was just one solitary shot form inside the six yard box. And no headers from inside the six yard box too.
While not damning, those numbers are not exactly warm and fuzzy either.

However, if we look at the shot shares when scores are level in the game, this makes a bad situation look even worse.


Pretty horrific yeah?
Taking 40% (or even worse just 30%) of shots when the score is tied is an awful position to start a game from.
Yes, the PDO score is kind of bad as well, but a team would need a mammoth PDO up around 120 to be surviving that kind of shots performance.
As a result, by the completion of week eight, Brentford had spent as much time at one goal down as they had at level scores.

I cannot say with anything other than guess work that these are the sorts of figures the management team at Brentford would be using to make the decision, but it seems a likely path.
Brentford were not in their lowly position by an awful run of bad luck. This league position was not a quirk of fate. Shots-wise at least, it was fully deserved.

And although it may make some fans or media happy to see the experiment with analytics in Griffin Park apparently fail so quickly, that Dijkuzien was let go at this early point probably says loud enough that there was a realisation that his role and input was not working out.
That is what the analytics should be doing; helping make informed decisions at the appropriate time without reacting to isolated events.
Tuesday’s home defeat by Birmingham was another shots horror show, so it will be interesting to see how Lee Carsley beds-in and if he can turn this poor start around.