Lies, lies, damn lies and statistics

[Deleted]

Good posts, but pretty depressing.

A sustainable business model for a football club is a myth, like a perpetual motion machine.

[seriouslyred]

Oct 25, 2016 21:57:19 GMT Mundell said:

Just watched the video again. Thanks for posting.

Fairly obviously, Katrien Meire could have been more careful with her choice of words and with the way she chose to make certain points. However, I have to say that the hysteria this panel discussion created seems even more absurd now than it did at the time.

Once you cut through the bluster and some of the unhelpful language to identify the core messages, it becomes clear that much of what Meire said actually makes sense. Moreover, it's also clear that if Duchatelet could have achieved his objectives he'd have succeeded in building a sustainable business model for the club, rather than destroying it.

Once again, we're back to the question of execution. Appalling execution has been the problem, not malign intent.

To be clear, I'm not defending Katrien Meire. Appointing her as CEO is probably Duchatelet's biggest mistake. It was a really poor decision.

I've now concluded that the club is indeed facing an existential crisis. At the heart of that crisis is a major disconnect between the expectations of fans and the reality of the financial nightmare the club and others like it faces. The wage bill needed to fund a playing squad which can compete in the Championship, let alone one that could be expected to compete for a play-off place, implies significant losses and a major funding gap of up to £10m p.a. Or more. That's a very expensive hobby. The club is simply not sustainable in the Championship.

However, if the club's revenues are too small to fund Championship football, the club's infrastructure is too expensive to enable anything close to breakeven in League One. Moreover, as we've already witnessed this season, the club's fans simply won't 'accept' League One football. The club is not sustainable in League One either.

There may be no way out. The only real hope, beyond a sudden improvement in the current owner's execution of strategy, is the arrival of a very wealthy white knight. We may get lucky, or we may not, but even if we do the underlying challenge the club faces is not going away. Our white knight will need deep pockets and almost certainly a large slice of luck if he or she is to escape the nightmare and make it to the promised land of the EPL. More likely, I fear, is another disappointment and more of the same. How will our fans react to that?

It's complex and it's not just about the owner. Are our fans saving the club or does the club need saving from its fans?

One great quote about Charlton over the years: "It's not the disappointment, it's the hope I can't stand".

I remember in the mid 90s when the introduction of three points for a win and the play-offs meant that we were forever just one or two wins away from the top six - feel familiar!

And yet the gates kept climbing by 1,000 every season.

Forgetting the execution for five minutes, what Duchatelet and Meire have stated is basically forget the hope of the FAPL and instead focus on improving players to earn fees which in turn pay for the club losses.

Does it take a white knight to save us or is it that there is a simple business case for Duchatelet (or his successor) to invest in the squad to get us promoted once and then aim for the top eight in the Championship.

Slater, Jiminez and Powell completed the first part within 18 months of arrival. But the rumour is that their backers baulked at the £10M a season estimated cost to go to the next level.

And that cost of buying a top six slot may well have risen by 50% or more since 2012?

There are a plethora of examples of clubs which have succeeded and those which have failed in their bids to become yo-yo clubs receiving FAPL and parachute monies:

Norwich, Burnley, Watford and dare I mention Palace have all tasted success with the odd relegation. While the likes of Fulham, Blackburn, Cardiff and QPR have accumulated hundreds of millions of debt even when in the top tier.

Nobody can guarantee success at Charlton but if we don't even bloody try then fans will drift away. We will move further and further from the top 30 and down sizing will become an attractive option. Particularly as the Valley ages.

So just two things to add. Firstly Swiss Ramble tracks the Championship finances and his next release once he gets his paws on last seasons numbers will inform the debate. Just how much did Sheffield Wednesday lose in their pursuit of a losing play-off final berth.

The other point is that we should not dismiss the fact that the current owner is pumping in an estimated £1M per month to keep our club going. Some struggle to comprehend that or challenge it as pure fabrication. But the numbers were in the programme in black (red) and white back in October 2015. From January 2014 to August 2015 RD had put in £21M to cover losses, capex, player acquisition, capex and paying off the North upper.

Hard to tell what that figure is now but the club is down on media revenue due to relegation. And we don't know what has been spent on the training ground.

By 2018 that £21M might have grown to £35-40M? I think perhaps some of that money should go on improving league prospects and that requires the right people in charge. But I won't dismiss the tens of millions as easily as some appear to. My approach would be to request better value.

[coloradocafc]

I just read the following, which I thought was mostly nicely done and interesting. You may know the story, but I didn't.


bad math and football analytics


I'm interested in this sort of analysis in sport. My affinity is colored by the fact that it's revolutionized some other sports I follow like baseball and basketball. It has its limitations, of course, and football is a harder sport to describe mathematically, but I think the analysis has explanatory and predictive value.

Ultimately, analysis is as good as the analyzer. I guess the same can be said for football scouts, scientists and bricklayers.

[Mundell]

In case there is interest in the latest Experimental 3-6-1 update I've linked it here - Experimental 3-6-1

[jonkool]

Nov 5, 2016 11:58:35 GMT Mundell said:

In case there is interest in the latest Experimental 3-6-1 update I've linked it here - Experimental 3-6-1

I lurv this thread Mundell - the thread that keeps on giving m8

[Mundell]

I continue to believe that the Experimental analysis provides a useful guide to how we are performing. We are now ranked ninth in the table with the ninth best attack and the seventh strongest defence.

In this blog the author uses his ratings to estimate the probabilities of automatic promotion, a play-off place and of relegation for each team. In round numbers he estimates that we have around a 2% chance of automatic promotion, a 15% chance of a play-off place with 3% chance of relegation.

It is very important to bear in mind that these probabilities assume no improvement in underlying performance. They imply that even if we don't improve and remain the 'ninth best team' we still have a chance of automatic promotion, albeit a very small one, and a realistic chance of making the play-offs. The element of randomness in these outcomes is hard to understand intuitively, but it is, nevertheless, representative of reality. The six best teams do not necessarily finish in the top six.

There are two implications of this analysis. First, if our underlying performances continue to improve our chances of promotion will also improve relative to Experimental's current estimates. Second, there is no escaping the fact that we are going to need the breaks. We need the randomness to work in our favour, not against us. The same is true for every other team. The 'four best teams in the League', based on Experimental's analysis, each have a 10% chance or more of failing to make the top six. That's hard for us all to relate to intuitively. We are all repeatedly fooled by randomness.

[Mundell]

Blogger Ben Mayhew has updated the so called E Ratings which enable him to rate and rank each of the teams in League One. He has also used the ratings to estimate the probability of each team finishing in the top two, in a play-off place, being relegated etc.

This kind of analysis may not be of interest to everyone, but I do feel it gives us an indication of the task ahead for Karl Robinson. Perhaps of more interest is the fact that Experimental 3-6-1's analysis will provide a useful way of assessing how quickly the new manager is able to improve the side's performances.

The link to Experimental 3-6-1 is here.

[seriouslyred]

Interesting that we still have a 15% chance of the play-offs and Bradford dropped from 45% to 40% chance of automatic promotion after our draw.

There are clubs within a couple of points of us who are given a 40% chance of the play-offs.

Let's see how they rate us at the end of the month.

[Mundell]

Experimental 3-6-1 blogger Ben Mayhew has now posted his analysis of the first two rounds of fixtures in League One. His so-called timelines record the 'expected' goals (or score) for each match as well as plotting how the balance of power, measured by chances created, evolved during the game.

As I've said previously, Mayhew's team rankings, based on his expected goals methodology, correlate very strongly with league table position over the course of a season and so can sometimes provide a better guide to how well teams are playing than short-term results which can be 'noisy' and misleading.

It's obviously very early days yet, but the results of his analysis of our matches are, perhaps, a little surprising. They are certainly disappointing. I had expected to find that based on our underlying performance we had 'won' the game at Plymouth and had come unstuck simply because we failed to take our chances, but it turns out that this is only half true. It appears that we did create enough chances to score one or two goals and that we had the better of the first half. However, Plymouth turned things around in the second half and overall it appears to have been a very even game in which a draw would probably have been a 'fair' result.

There were fewer chances in the home game against Bristol Rovers, but that was also a very even game it seems which, if anything, Bristol slightly shaded. That's perhaps not so surprising given that we played for ninety minutes with only ten men, but it means that after two games our 'expected goal difference' is actually slightly negative, which is, of course, inferior to both our expected and actual goal difference for the whole of last season.

We are clearly going to need to play much better if we are to mount a serious challenge for a place in the top six. Unfortunately, there are a number of sides which, in contrast, have come flying out of the blocks. Wigan, for example, have been dominant in both their matches and have fully deserved their two wins. Oxford, with their highly regarded new Spanish manager Pep Clotet, have also been very impressive. Blackburn, on the other hand, have struggled in their opening fixtures, losing both deservedly based on the balance of chances created.

I'll keep this thread updated during the next two to three months. Win, lose or draw in individual games, what I'll be hoping to see is a consistent improvement in our expected goals for and against. If that doesn't happen, then Karl Robinson is going to need to find that elusive Plan B.

Here is a link to the August 12th timelines

[Mundell]

And here are the League One timelines for the weekend of August 5th

[Deleted]

How did Sheffield Utd compare last season?

[seriouslyred]

Aug 15, 2017 7:34:28 GMT jeddez said:

How did Sheffield Utd compare last season?

Here's a scatter graph for last season showing us bang in the middle:

experimental361.com/2017/04/30/scatter-graphics-league-1-201617/

As you can see Sheffield United created 14 chances per game whilst conceding only 8.

We are bang in the middle of every graph. An improvement of an additional two shots per match whilst conceding two less would place us close to the top six.

[Mundell]

Aug 15, 2017 7:34:28 GMT jeddez said:

How did Sheffield Utd compare last season?

Based on the expected goals model Sheffield United were comfortably the best team in League One last season with the best attack (most expected goals for) and the best defence (fewest expected goals against).

The other teams 'rated' in the top six were Bradford, Millwall, Bolton, Oxford and MK Dons, in that order, with MK Dons the real outlier, obviously. Fleetwood and Scunthorpe, the other clubs that made the playoffs, were ranked 8th and 9th respectively.

Note: This data is end of April not end of season though it's unlikely rankings changed.

[Deleted]

Doubt it would have been after the first 7 games thats what I was getting at...

[Mundell]

Aug 15, 2017 8:32:27 GMT jeddez said:

Doubt it would have been after the first 7 games thats what I was getting at...

I agree that that's the right question. I'll check when I get a moment, but I wouldn't be surprised if Sheffield United's performances were better than their results early in the season

[Mundell]

Aug 15, 2017 8:35:31 GMT Mundell said:

Aug 15, 2017 8:32:27 GMT jeddez said:

Doubt it would have been after the first 7 games thats what I was getting at...

I agree that that's the right question. I'll check when I get a moment, but I wouldn't be surprised if Sheffield United's performances were better than their results early in the season

jeddez

I've now taken a look at the data. It's interesting and perhaps a bit surprising, but I can confirm that your intuition about Sheffield United's start to the season is bang on the money.

By way of background, and as discussed previously on this thread, the objective of the expected goals model is to find a measure of underlying performance which is a better indicator of future results and final league table position than the result of any one match. The idea being to 'look through' the inevitable randomness which can impact the outcome of individual games or even a sequence of matches.

Mayhew's model works very well, as confirmed by the comments in my earlier post and, more importantly, by some insightful analysis he did at the end of last season. Results correlate very significantly with 'expected' goals and, on average, teams with a higher 'expected' goal difference finish higher up the table. However, this thread isn't titled 'Lies, lies, damn lies and statistics' without good reason. We need to be very careful when we interpret data of this type.

When a team 'outperforms' (or underperforms) based on expected goals the most likely outcome is that it's results will mean revert relatively quickly. Last season this happened with Millwall, who started slowly, and Bury who started well, for example. However, sometimes this does not happen for some reason. It might simply be (more extreme) randomness at work again, but after a period of time this becomes less plausible so that eventually it is clear that there must be something else going on. For example, teams with outstanding strikers might be able to maintain a higher than 'normal' conversion rate, while a brilliant shot stopper might be able to limits goals conceded versus chances created, relative to the average for the league.

Another reason why Mayhew's expected goals based rankings may not predict final league table position accurately, of course, is that a team's underlying performances might improve as the season develops. It appears that this is what happened with Sheffield United.

United lost three of their first four matches, 'drawing' one of those on expected goals (a match they lost) and 'losing' the other three on the same metric, two quite heavily. After ten matches the eventual champions had a 5-2-3 record and seventeen points. Not a bad start, but they'd actually 'outperformed' their expected goals, winning three games which the opposition shaded and which could, therefore, just as easily have been drawn, or even lost.

One crude statistic that illustrates the extent of United's improvement after those first ten games is goal difference. Based on Mayhew's model the Blades would have had a goal difference of just over two at the end of the season if we simply extrapolate their performance in the first ten matches. Based on their performance over the entire season their 'expected goal difference' was close to forty, compared to an actual goal difference of forty five.

In conclusion then, while Chris Wilder's side eventually dominated the division in terms of points won and expected goals for and against, their start was much less impressive. Indeed, the seventeen points they managed to get from their first ten matches may well have been more than they 'deserved'.

Let's hope that Karl Robinson's side is also able to build strong momentum as the season develops and, just like Sheffield United, that we are able to grind out points early on in the season, even if not playing that well, both to relieve pressure and build confidence.

[stonemuse]

Thanks Mundell, very interesting.

I may be wrong but I guess the main difference between us and Sheffield Utd is that they already had a complete squad whereas we are still looking.

[Deleted]

Thanks for that Mundell.

I do think we will create many chances with the squad we have and a fit Magennis will be a very good striker to have, but when games aren't going out way a lookman/mavididi type player would be great to bring on to offer an alternative. This is the type of player I think we need whether as a purchase of a season long loan...

[Deleted]

Aug 15, 2017 21:17:09 GMT stonemuse said:

Thanks Mundell, very interesting.

I may be wrong but I guess the main difference between us and Sheffield Utd is that they already had a complete squad whereas we are still looking.

They actually signed 2 towards the end of August 1 free in September and 3 in January. 

[Deleted]

Aug 16, 2017 5:00:04 GMT @jondon47 said:

Aug 15, 2017 21:17:09 GMT stonemuse said:

Thanks Mundell, very interesting.

I may be wrong but I guess the main difference between us and Sheffield Utd is that they already had a complete squad whereas we are still looking.

They actually signed 2 towards the end of August 1 free in September and 3 in January. 

Haha oops! These people should check their facts out!

[stonemuse]

Aug 16, 2017 5:00:04 GMT @jondon47 said:

Aug 15, 2017 21:17:09 GMT stonemuse said:

Thanks Mundell, very interesting.

I may be wrong but I guess the main difference between us and Sheffield Utd is that they already had a complete squad whereas we are still looking.

They actually signed 2 towards the end of August 1 free in September and 3 in January. 

Thanks for that. Interesting to see that of the August signings, one was a goalkeeper and one a forward, with another forward in January.

[Mundell]

The article linked below might be of interest to fellow nerds and others who have been following this thread.

The piece appeared in today's Times and explains what 'Expected Goals' is all about. 

The Experimental 3-6-1 analysis I have been reporting here is constructed using the ideas explained in the article. It should be noted, however, that the data available in League One is not as rich as that described in the piece linked so that blogger Ben Mayhew uses a less refined methodology. The principles are exactly the same though. 

Expected Goals

[Mundell]

Here, with all the usual caveats, is the League One expected goals scatter plot, updated following yesterday's win at Rotherham. 

Based on this evidence we've made a solid start, but have, if anything, 'outperformed' our expected goals. Our performances probably need to improve if we're going to stay in the top six. That said, in Bristol Rovers and Rotherham, we've beaten two of the League's better sides by the looks of it. 

It is is hardly surprising that Wigan are the early favourites to win promotion. Their underlying performances suggest that they've been out on their own in the first four games. Oldham, on the other hand, have been the 'worst' team in the League so far, and by some margin, suggesting we have a real chance of another three points next Saturday.

Expected goals scatter plot for League One

[Mundell]

   .... and here are the so-called Timelines from yesterday's League One fixtures. In our case the analysis seems to be consistent with eye witness reports that Rotherham started strongly and that we managed the game more effectively after half-time.

League One timelines

[jonkool]

Some great data Mundell but it comes with the qualification that our stats indicating that we represent the mean average for tier 3 represent less than 10% of the fixtures we will play this season.

A 4 game sample can easily be distorted and in this case by the loss of a red carded player just 6 minutes into a game.

This will potentially impact on a home team by the lowering of the goal scoring potential of the team or the increasing potential to concede.

This early season dismissal will be ironed out by the increasing sending off stats of opponements.

[Mundell]

Completely agree jonkool The data is interesting, but needs to be interpreted with care.

[Mundell]

Another interesting piece of analysis by Experimental 3-6-1 which fellow nerds might enjoy. Once again, it's important to note that it's way too early to draw any firm conclusions. However, this analysis of expected goals by striker (we might call this 'quality adjusted chances') shows that so far we have created plenty of opportunities for Josh Magennis and that his conversion rate is very acceptable. If we can keep this up our number '9' will get to twenty league goals with a few to spare. 

We are creating chances for Josh Magennis

[seriouslyred]

Indeed, Magennis has a good conversion rate. And the brains behind that site promises more data on more players once the EFL clubs have played ten of more games. Perhaps we get to see how more attacking players at CAFC fair compared to their counterparts elsewhere.

Seeing as we play 4-2-3-1, it is vital that Magennis plays 30+ games this season and that his cover is competent. It is also key that the attacking three plus Bauer create and finish chances for that will determine whether we can stay as a top four side or whether we have a challenge to make the play-offs.

The point of these graphs and stats is that we can see a clear comparison to last season irespective of who we have played. This post from that site shows where we were last season - bang in the middle on most of the graphs:
experimental361.com/2017/04/30/scatter-graphics-league-1-201617/
The expected score at our games was a draw with 1.3(!) goals on each side. Top four sides need a higher expected goals scored and lower conceded. We still have a way to travel in that respect.

[seriouslyred]

Here are the timelines from yesterday:
experimental361.com/2017/09/03/league-1-timelines-2-3-sep-2017/
It appears our next opponents should have won. What is interesting is that the line for Charlton follows our understanding of the game: first 25 we batter them - then Fame comes on and their pen brings them back - then Fane picks up a second yellow and "normal service" is resumed.

We are certainly expected to score more than last season but it will probably take a few more games before a reliable pattern for the whole division emerges.

[roburwash]

Sept 3, 2017 9:01:29 GMT seriouslyred said:

Here are the timelines from yesterday:
experimental361.com/2017/09/03/league-1-timelines-2-3-sep-2017/
It appears our next opponents should have won. What is interesting is that the line for Charlton follows our understanding of the game: first 25 we batter them - then Fame comes on and their pen brings them back - then Fane picks up a second yellow and "normal service" is resumed.

We are certainly expected to score more than last season but it will probably take a few more games before a reliable pattern for the whole division emerges.

If experiment361 is a reliable indicator we could shorten the season to about 15 games :-)