Lies, lies, damn lies and statistics

[Mundell]

Two fantastic results for the Addicks at Shrewsbury and then yesterday at Portsmouth have completely transformed our chances of making the playoffs, but results elsewhere have helped too.

As others have said on other threads, to all intents and purposes it’s now a three horse race for the two remaining playoff places and Experimental 3-6-1’s analysis confirms this conclusion.

Experimental’s simulations suggest we now have a 75% chance of finishing in the top six with Scunthorpe at just below 90% and Plymouth now a little less than 30%. Portsmouth, Peterborough and Bradford retain mathematical chances, but even in combination these represent just less than 10%.

To paraphrase @lardiman a week is a long time in football. The turnaround in our fortunes since this time last week is extraordinary. That’s the beauty of the game, of course.

[seriouslyred]

Not only did we win and increase our chances of the play-offs up to 75% but this set of graphs show that our competitors are struggling. One shouldn't take just one game to assess how a team is doing but it is noticeable that we created enough to have an expected goals of 2-1 whilst our rivals Plymouth and Scunthorpe struggled to create enough for two goals between them. And that was against weaker teams.

And Shrewbury who may well be our opponents in the play-off semi-finals just about deserved a 1-1 draw with relegated Bury.

It's still very hard to tell what the final points total required will be so we just need to hope that one of the other horses falls before we do. If we need a win at Rochdale on the last day, that is possible but high risk given how well they are playing to avoid the drop. Perhpas they might do us a favour in beating Plymouth on Tuesday?

[Mundell]

Agreed. We’re now at the stage of the season where anything can happen. Randomness can easily dominate over just two, three or four games so that outcomes become very hard to predict.

What’s helpful about Experimental 3-6-1, and indeed the bookmakers’ odds which have the benefit of similar analysis, is that it provides an objective assessment of the probability of different outcomes where our intuition often really struggles.

Plymouth’s chances might seem somewhat low at just less than 30% given that they’ve got two games in hand, but the fact is that statistically they’re not very likely to win those matches. For example, Plymouth’s chances of winning at both Rochdale and Gillingham are probably less than 20%.

It’s unlikely, of course, but we could lose to Blackburn and yet head to Rochdale certain of a place in the top six.

Whatever now happens, it will have been a fantastic end to the season.

[jonkool]

A 46 game League season plus Carabao, Checkertrade & FA Cups is a long slog and as I have mentioned in pre-season’s past our wage budget and Academy must offer the gaffer a far greater squad in comparison with all our tier 3 rivals bar Wigan and Rovers.

Shrews with their run to Wemberlee in the Checkatrade and Plymouth who have given 100% in order to transform their league position from bottom 4 to top 6 may we’ll be running out of legs.

Have to say that if we make the play offs based on instinct as opposed to stats is that I would prefer to finish 6th and play Shrews and let Scunthorpe take 5th and play off against Rotherham even though we did the double over them.

It’s wonderful that we have a possible post season to look forward to - in Bow we must surely trust (but it would be advantageous if Fosu was fit for the remainder of the season).

[reamsofverse]

Apr 22, 2018 10:32:51 GMT seriouslyred said:

Not only did we win and increase our chances of the play-offs up to 75% but this set of graphs show that our competitors are struggling. One shouldn't take just one game to assess how a team is doing but it is noticeable that we created enough to have an expected goals of 2-1 whilst our rivals Plymouth and Scunthorpe struggled to create enough for two goals between them. And that was against weaker teams.

And Shrewbury who may well be our opponents in the play-off semi-finals just about deserved a 1-1 draw with relegated Bury.

It's still very hard to tell what the final points total required will be so we just need to hope that one of the other horses falls before we do. If we need a win at Rochdale on the last day, that is possible but high risk given how well they are playing to avoid the drop. Perhpas they might do us a favour in beating Plymouth on Tuesday?

Why do you need graphs to fathom that. All you have to do is look at the results ?

Obsessed is not the word.

[Mundell]

Completely agree with that jonkool

I too feel that Rotherham are going to be the best of the playoff contenders. I think there is also a chance that Plymouth will find their next three games (at Rochdale on Tuesday, then at home to Rotherham, followed by Scunthorpe away the following Tuesday) really hard going. Who knows? We may neither need nor want to win at the Crown Oil Arena on the last day of the season!! 

[jonkool]

Apr 22, 2018 20:50:28 GMT Mundell said:

Completely agree with that jonkool

I too feel that Rotherham are going to be the best of the playoff contenders. I think there is also a chance that Plymouth will find their next three games (at Rochdale on Tuesday, then at home to Rotherham, followed by Scunthorpe away the following Tuesday) really hard going. Who knows? We may neither need nor want to win at the Crown Oil Arena on the last day of the season!! 

The Crown Oil Arena - home for a bunch of frackers?

The post Easter run in has been highlighted by most of the playoff contenders playing each other in tense late season 6 pointers. I suspect that the attendances of the top 8 or so clubs have risen over recent weeks at a time when EFL attendances have benn on the waine over recent seasons. Maybe Div 1 isn’t a Mickey Mouse division after all!

[Deleted]

Apr 22, 2018 20:50:28 GMT Mundell said:

Completely agree with that jonkool

I too feel that Rotherham are going to be the best of the playoff contenders. I think there is also a chance that Plymouth will find their next three games (at Rochdale on Tuesday, then at home to Rotherham, followed by Scunthorpe away the following Tuesday) really hard going. Who knows? We may neither need nor want to win at the Crown Oil Arena on the last day of the season!! 

I broadly agree with most of this, and other recent posts.
My only concern would be perhaps we should not worry so much about choosing our favoured opponent as keeping momentum.

I'm pretty sure nothing makes the players feel better than winning a match. It might be counter-productive to go into the final game of the regular season against Rochdale not being too bothered if we get beaten.

Though I can see the case for Shrewsbury being preferable opponents in the play-offs because they have played so many games this season and will possibly still be deflated by missing out on an automatic promotion place, to my mind there is not really very much difference in difficulty whether we end up getting Shrewsbury or Rotherham in the play-off semi-finals.

With that in mind, I would rather Charlton go into the play-offs on a high and full of confidence after winning our game at Rochdale.
Also, for the cause of confidence and momentum as much as to secure the points we may need and what a great day it would be for us fans, I am also still very much hoping for a victory against Blackburn Rovers on Saturday.

From a statistical point of view (and depending on other results of course) we may not need to beat Blackburn or even Rochdale, but I believe it would be far better for our play-off prospects if we do.
If Charlton can go into those play-off semi-finals brimming with confidence and enjoying the taste of more victories I don't think it will matter too much who we have to beat next.

[Mundell]

Don’t think anyone could disagree with that @lardiman

[Deleted]

What a fantastic result for Lee Bowyer and his men yesterday.
Our play-off fate is in our own hands now.





The players know a win against Rochdale will guarantee Charlton a place in the play-off semi-finals.

Other results may or may not go our way if we do not win.
We could even lose and still finish 6th regardless of the Scunthorpe V Plymouth game on Tuesday.
But we would be relying on Scunthorpe or Plymouth to also lose on the last day.

However, 3 points for the Addicks on May 5th and we keep our season alive.
Doesn't matter what anybody else does.
That will be a magnificent achievement from our young caretaker manager.

[Mundell]

There are now only four results which matter as far as we are concerned, our own game at Rochdale next Saturday, the six pointer between Scunthorpe and Plymouth on Tuesday evening and then Scunthorpe's home game against Bradford City and Plymouth's visit to Gillingham. 

As @lardiman has said, a win at Rochdale and we are home free, but we could lose that game and still make it. Indeed, while we will probably be no better than evens to take all three points at the Crown Oil Arena, we are now 10/1 on at the bookmakers to make the playoffs. Scunthorpe are 12/1 on with Plymouth now out at 5/2.

Experimental 3-6-1's analysis can be found here and concludes that we have a 90% chance of finishing in the top six. 

Yesterday's League One timelines suggest that our game against Blackburn was one of the most 'chance-less' games of the season. The expected goals were 0.4 v 0.1 in Blackburn's favour. 

Lee Bowyer has clearly made us very hard to beat with just four goals conceded in nine games. Moreover, and despite the quality of the opposition we've played, our expected goals (Xg) against over those nine matches is 0.89 per game. Over the course of the entire season only Wigan have bettered that, with an Xg against of 0.82. That's impressive and augers very well for the playoffs, assuming we get there.

What's perhaps been a little less encouraging has been our attacking threat. While we've scored fourteen goals in Bowyer's nine games, and significantly improved our goal difference, our Xg for is only 1.26 which is 'fairly average', albeit we've been playing strong opposition during this sequence of games.

It will be interesting to see how Bowyer approaches the playoffs. Which system he uses and whether he finds a way of making use of Aribo, Fosu, if fit, Mavididi, Zyro and Kaikai.   

[seriouslyred]

So we have a 10% chance of missing out. Back in January 2017 we had a 10% chance of making the play-offs and in Jan 2016 there was a 10% chance of avoiding relegation. The boot is on the other foot so to speak.

For our xG to improve from zero to a positive 0.45 goals overnight with the appointment of Bow is somewhat spectacular. We will see how the Tuesday game goes and then hope for the best on Saturday.

Isn't it great to be looking for when the play-off games are scheduled and look forward to another 14-15,000 Charlton fans gathering at the Valley. We've had some poor attendances these past two seasons but the average is now back close to 12,000.

And it's clear that it will climb if we are successful.

[Mundell]

Wasn’t sure where to post this article on Burnley which appeared in today’s Times, but this thread seemed the best bet. 

Anyway, I’ve linked it here because I felt some people might find it interesting. I really enjoyed reading it. 

In essence, the article says that last season Burnley’s results were much better than would have been expected given the data on their performances, i.e. possession, shots for, shots against etc. This season it’s been a different story, of course. Burnley sit second from bottom of the table, prompting the question what’s gone wrong? 

I confess, I’m fascinated by the question how do we separate ‘signal’ from ‘noise’? How do we know when a team’s good (or bad) results mean that it’s playing well (or badly) and when, on the other hand, we’re being fooled by randomness? Did Sean Dyche keep rolling sixes last season or was there some good reason, some ‘skill’, ‘process’ or ‘tactic’ which somehow meant that results were, for good reason, better than they should have been based simply on the data? 

And what’s gone wrong this season? Has Dyche lost his touch, the players their commitment or are we simply observing mean reversion? It’s possible even Dyche himself doesn’t really know. 

My Burnley supporting friend says that the fans’ perception is the players aren’t quite as fired up as they were last season. The fans may be right or they may just be struggling to find a narrative that appears to fit the facts.

Either way, their side was fired up on Wednesday against Liverpool, too much so for Jurgen Klopp’s liking, it seems.

Burnley lost anyway though and Dyche remains under pressure. His team are now favourites for relegation. 

[Mundell]

Only my second post on this thread this season and the first to look at Experimental 3-6-1’s analysis of League One. 

Here’s a link to an article which discusses Experimental’s so-called E Ratings and probabilities of final league table positioning. 

At least as far as we’re concerned the results are fairly intuitive. Based on ‘expected goals’ we are rated 8th in the League for both attack and defence and as a good ‘all rounder’ 4th overall, behind Barnsley, Luton and Portsmouth. 

Our chances of winning automatic promotion and of a playoff place are very consistent with the bookmakers’ odds. 

[newstreethill]

Jan 27, 2019 17:10:39 GMT Mundell said:

Only my second post on this thread this season and the first to look at Experimental 3-6-1’s analysis of League One. 

Here’s a link to an article which discusses Experimental’s so-called E Ratings and probabilities of final league table positioning. 

At least as far as we’re concerned the results are fairly intuitive. Based on ‘expected goals’ we are rated 8th in the League for both attack and defence and as a good ‘all rounder’ 4th overall, behind Barnsley, Luton and Portsmouth. 

Our chances of winning automatic promotion and of a playoff place are very consistent with the bookmakers’ odds. 

It's odd to see that our defensive rating hasn't improved seeing as how we've just kept three clean sheets in a row. How do they work that out?
Edi: Ignore that. Our expected goals conceded has actually gone down, as you might expect.

[Mundell]

Haven’t posted on this thread for a while, but with the season entering the business end I felt a comment might be worthwhile.

Many fans seem to believe that we’ll be safe from relegation because there are three teams in the League worse than we are or, for those who like to think more positively, who we are better than. One question we might ask, therefore, is what, if anything, the data tells us about this?

Expected Goals is the most commonly used measure of underlying performance and although the relationship between quality adjusted chances created and conceded and actual goals scored can be noisy, actual goals scored, and hence results, are very likely to mean revert to what’s ‘expected’ in time. That’s the idea, of course.

Two very good, publicly available football stats sites are Experimental 3-6-1 and Infogol. They both put together estimates of Championship expected goals (Xg). Based on Experimental 3-6-1’s data, our Xg For (XgF) so far this season is 36.3. This compares with actual goals scored of 41. Only Huddersfield and Luton have a lower XgF. Infogol’s estimates are very similar. Based on their data our XgF is slightly higher at 37.1, but that’s the lowest in the Championship according to their analysis.

If we look at expected goals against (XgA), Experimental’s and Infogol’s estimates are 54.3 and 57, respectively, compared with 47 goals actually conceded, with only Luton faring worse. This data suggests that we’ve scored more goals than ‘expected’, based on the chances we’ve created, but that more importantly we’ve somehow managed to concede perhaps as many as 10 fewer goals than our opponents might have expected to score given the number and quality of chances they’ve created. The home game against Barnsley is a recent example of this.

Both Experimental and Infogol use their expected goals data for each match to produce an ‘expected’ league table. This is how the table would look if each result was ‘fair’ based on the quality of chances created, based on their data, and making no allowance for quality of finishing etc. Experimental’s table has us in 24th place with just 22 points, i.e. implying we’ve got 14 more points than we should have. Infogol use a different methodology to calibrate their table, but the result is very similar. We’re placed 23rd, above only Luton Town, who are placed 23rd in Experimental’s table. Both models conclude that Charlton and Luton are the weakest teams in the Championship.

Both Charlton and Luton would argue, with good cause, that they’ve been very unlucky with injuries this season and that, when at full strength, they are better than this data suggests. That may be true, but there’s no easy way to test it, unfortunately. However, it’s interesting to look at the eight games Charlton have played this year. We’ve taken eight points from those matches with two wins, two draws and four defeats. However, arguably, that’s more than our underlying performances have justified.

Our XgF during these eight matches has averaged just 0.8 per game, compared to around 1.0 per game over the entire season to date. We’ve become less of a goal threat. Our XgA has been 1.39 which is somewhat lower than over the season in it’s entirely. There is some evidence here that we’ve been playing more defensively and that this has helped to reduce the number of chances we’re conceding, but that it’s come at the price of an even lower goal threat. It’s also worth noting that based on the Xg scores we didn’t ‘deserve’ to win any of these matches. We were very lucky to beat Barnsley and perhaps a little fortunate to beat Forest.

What’s important, of course, is that it’s actual points which count, not points expected or ‘deserved’ based on underlying performances. And based on actuals we’re five points clear of the bottom three with a superior goal difference. The real question then, even if we are one of the two poorest teams in the League, is whether this points cushion is enough? Infogol attempts to answer just this question. Using Xg as a measure of relative strength, they simulate all remaining results to arrive at an ‘expected’ final table. In that table we are placed 21st, above Luton, Barnsley and Wigan. Phew!!

This suggests that we don’t necessarily need to improve to survive, but we do need to hope others don’t. Saturday is going to be a very big game, for both sides. Luton have just won two consecutive matches, both 1-0, and manager Graeme Jones has said he has his entire squad available for the first time this season. A win would be ideal, but avoiding defeat may prove to be essential.

[redoctober]

So many people thinking that we’ll survive with ease. Wrong. It’s going to be an uphill battle and one that in my opinion its no better than 50/50 that we will.

[Mundell]

Feb 17, 2020 13:05:10 GMT redoctober said:

So many people thinking that we’ll survive with ease. Wrong. It’s going to be an uphill battle and one that in my opinion its no better than 50/50 that we will.

It could well be very touch and go and expectations would shift dramatically if we lost to Luton on Saturday.

The SkyBet odds imply that we have around a 65% chance of survival. However, those odds are heavily influenced by the fact they’ve almost given up on Luton (1 to 7 on for the drop) and Barnsley (1 to 5 on). If either of those sides ‘came good’ we’d be in big trouble.

[seriouslyred]

Feb 17, 2020 13:05:10 GMT redoctober said:

So many people thinking that we’ll survive with ease. Wrong. It’s going to be an uphill battle and one that in my opinion its no better than 50/50 that we will.

50/50 is a good call as we dont have enough information to call it to any finer degree.

Still that's better than 25% chance of promotion this time last year!

In all seriousness, pay close attention to how our players do against lower half sides now everybody is available.

What we do know is that the atmosphere will go toxic should we lose on Saturday!

[jonkool]

Well it’s better than mid table obscurity (so long as we survive the drop) 😉

[canterburyaddick]

Feb 18, 2020 21:07:58 GMT jonkool said:

Well it’s better than mid table obscurity (so long as we survive the drop) 😉

I'd settle for mid-table obscurity to be honest JK.

[jonkool]

Feb 18, 2020 22:08:35 GMT canterburyaddick said:

Feb 18, 2020 21:07:58 GMT jonkool said:

Well it’s better than mid table obscurity (so long as we survive the drop) 😉

I'd settle for mid-table obscurity to be honest JK.

maybe I'm a bit of an adrenalin junky Cants

[earlpurple]

If it's only based on how many clear-cut chances your team and your opponents create, it doesn't take into account how good your goalkeeper is (thus making saves) and how good your strikers are at taking their chances, which goes a long way to your success.

[seriouslyred]

Feb 19, 2020 13:47:12 GMT earlpurple said:

If it's only based on how many clear-cut chances your team and your opponents create, it doesn't take into account how good your goalkeeper is (thus making saves) and how good your strikers are at taking their chances, which goes a long way to your success.

Nor does it account for defenders "putting their bodies on the line" by blocking chances and challenging for headers etc. But here's the thing:

Whether one looks at expected goals, playing budgets (Swiss Ramble), playing squad value (transfermarkt) or the bookies odds, all of these datasets hold about 70% correlation with the League table.

Therefore 70% is predictable soon after the season starts... but that vital 30% unpredictability is down to luck, injuries, the ref... and perhaps the ability of the management to make the most of the playing resources available.

That was the difference between Bows and KR, between Gallen and Meire. But there's no place either for blind optimism - the odds, the chances created and the wage bill all point to us being a bottom six club.

Hence the odds are 50:50 until performances and points on the board prove otherwise.

Let's see where we are at 5pm on Saturday!

And let us enjoy the journey.

[Mundell]

I posted this article on Barnsley elsewhere, but thought it would be good to have it here too.

It’s a real insight into what’s going on at one of our rivals in the battle to stay in the Championship. Two things I thought were interesting. First, the stark contrast between the management approach at Barnsley and that at Charlton. Second, there is some insightful performance data which it would be interesting to see more of, both for us and for the Championship as a whole. 

[Mundell]

Almost posted this comment on the v Northampton match thread, but thought I'd do so here instead, if only to reawaken our discussion about data, for those who are interested.

I've linked here Experimental 3-6-1's Expected Goals League Table, updated following yesterday's win. 

As I've always said, data of this type needs to be interpreted with care. It is not fact. Teams can out or under perform, in terms of actual points won, what is predicted by their 'Xg' and even if that's not true, the data can be very noisy. However, neither is this kind of analysis wrong or likely to be fundamentally misleading. On average, and over time, a team's Xg is a good indicator of its underlying performances and, hence, a good predictor of future results. Over the course of a season a team's Xg is likely to correlate fairly strongly, therefore, with final League Table position, though there will always be exceptions. 

So what does Experimental's table tell us? The message is very clear. We've been fairly good at preventing our opponents creating clear cut chances, though nine teams have done better. A key for us has been that we've conceded 2.9 fewer goals than predicted. That might be poor finishing by our opponents, good goalkeeping by Ben Amos, good fortune, or something else, but only one team has done better. 

On the other hand, we are bottom of the table, and by some margin, for the quantity and quality of chances created. This has to be a concern, though its perhaps no surprise. 

Perhaps a bigger concern though is the apparent disconnect between this data point on chances created and Lee Bowyer's rhetoric. Again yesterday Bowyer said (something like) "given the way we play we'll always create chances". To be blunt, that's complete tosh. Perhaps he doesn't really believe it himself? I hope he doesn't.   

Either way, its seems reasonably clear that if we are to challenge for a top six place we are going to need to create a lot more chances. Its early days yet, of course, so perhaps as the team settles down this will come. In the meantime, it looks very much as if we're relying on our very solid defence and on winning ugly.  

[oldred]

Not disagreeing here, but I wonder if LB was speaking in general terms, and looking at the bigger picture of when we we have players like Maddison and Gilbey available again.
With them, Williams , Shinnie and JFC, we really should create more chances than we do at the moment.

Apart from the chance that Inniss created for Smyth, I can’t remember us creating many chances at all, and that will obviously affect the effectiveness of the forwards.
Defensively, we look stronger than for a long time, and in Inniss, a player who is dominant in the air. In the opponent’s penalty area, he set up two good chances, one of which was taken.
It’s not inconceivable, that he might score a few goals himself also.

When all of the midfield players are available, particularly Maddison, our chances created should be much higher. We need those chances to be taken though, because that’s all that seems to be lacking now.

[aaronaldo]

I’ve always thought the Xg stats provide a very good summary of our games. We do lack creativity and chance creation atm. I don’t think anyone will disagree on that.

[seriouslyred]

Oct 25, 2020 12:01:41 GMT oldred said:

Not disagreeing here, but I wonder if LB was speaking in general terms, and looking at the bigger picture of when we we have players like Maddison and Gilbey available again.
With them, Williams , Shinnie and JFC, we really should create more chances than we do at the moment.

Apart from the chance that Inniss created for Smyth, I can’t remember us creating many chances at all, and that will obviously affect the effectiveness of the forwards.
Defensively, we look stronger than for a long time, and in Inniss, a player who is dominant in the air. In the opponent’s penalty area, he set up two good chances, one of which was taken.
It’s not inconceivable, that he might score a few goals himself also.

When all of the midfield players are available, particularly Maddison, our chances created should be much higher. We need those chances to be taken though, because that’s all that seems to be lacking now.

The reasons we are so poor in the table which Mundell has linked are that
1/ we really were poor in the first four games, especially against Sunderland who created enough chances in the first half to score twice - but failed
2/ once certain defenders arrived there was a significant improvement/reduction in our conceding of scoring chances - it was about 1.5 xG conceded per game whereas the last three games this is now down to 0.7
3/ our attack was improving too but yesterday we went backwards with nothing in the first half - there's a long way to go to match the top clubs
4/ the top teams consistently create enough decent chances to score once or twice every game - we don't

Part of the solution has to be Maddison being fit, but it's not clear Bowyer is talking about options in the future. More like he has been talking up our creativity now which isn't as strong as he makes out. Ideally, Bowyer is more realistic about where we are at a day or two after the game. It's all very well to be proud of winning ugly but we won't see progress on creating chances unless that's the real focus at the training ground.

What's happened so far is done and we're fortunate to be in the top half of the table. We all know the back five is sound and ideally the xG will continue to improve as management work out their best midfield combination. However, one senses we might get a rude awakening on Tuesday playing Oxford?

[jonkool]

Yes stats over a longer period paint a far more meaningful picture.

The stats from the first 4 matches are not going to be pretty as we had such a small squad of experienced players to select from.

We were able to improve the squad in the final week of the window but many of the new arrivals lacked basic match fitness.

Therefore the stats May prove to be more insightful from say 3 or 4 weeks time when hopefully we will have increased the number of chances created.

Like many fans the concern is that we just don’t have attackers who will be able to convert sufficient chances.

But in the first place it will greatly help if the likes of Maddison and Gilbey greatly increase the team’s chances per match stats.

Fingers crossed that this proves to be the case.