Model trends: League 1, 12 Mar 2017

Regular readers will know that I churn out updated model predictions after every round of matches but I’m aware that there’s quite a lot of information in them and they can be difficult to track. I’m pretty keen to keep tabs on how well the model performs myself – not least because I’m constantly tinkering with it – so I’ve come up with a simple template that tracks how its assessment of each EFL club has evolved over the season.

The way it works is pretty simple: for pre-season and after every round of games it shows the results of my simulations for a specific club in terms of both:

  • How their average predicted final league position has changed (the solid line in the top chart)
  • How their predicted probability of ending up in each section of the table has changed (the coloured columns in the bottom chart)

The point of this is to show how the model’s assessment of a club’s prospects has changed after each round of games, but I also wanted some idea of how good a predictor it is. I’ve therefore added (as a dashed line) each club’s actual league position and briefly assessed (under each graphic) how well it’s predicted this so far.

I’m not expecting a 100% accuracy rate for a variety of reasons, including:

  • As Leicester recently reminded us, football is notoriously difficult to predict and the strongest team doesn’t always win
  • The rating system that drives the predictions can take a little while to adjust to sudden changes (e.g. a big tactical shift, replacing the manager or a lot of transfer activity)
  • While I’m pretty happy with the rating system and model, there’s limited data available in the lower leagues and therefore it may miss some subtleties in the way certain teams perform

Anyway, onto the graphics. There’s one for every team along with a brief summary and my view of how well the model’s predicted their fortunes.

Based on their disappointing finish to last season, the model thought that the play-offs were a more realistic target for Bolton than the top two. This looked to have been borne out when the Trotters faded after a bright start, but recently they’ve found another gear and their win at Fleetwood has put them in a strong position to claim second place.

Model performance: Not far off, but perhaps a shade too pessimistic

Bradford impressed last season and were therefore expected to challenge for automatic promotion this time around. With the Bantams just two points off second place the pre-season prediction looks to have been fairly accurate, with their attack having improved to match their strong defence. The model has cooled on them slightly but still expects them to challenge for a return to the Championship.

Model performance: Pretty good right from the start

I was surprised that the model was so positive about Rovers in pre-season, given that they’d been promoted twice in succession, but they’ve pushed on and established themselves as a top half team just as predicted. For a while it even looked like they might break into the play-offs but this might now be too big an ask with Millwall having two games in hand.

Model performance: Pretty close overall

Bury have a habit of starting strongly but fading and the model was suspicious of their early good fortune this time around. Even when they were sitting in the top two it expected them to eventually return to the bottom half and from November onwards they’ve been hovering around the relegation zone. A vulnerable defence has undermined a respectable attack and while they’ve had some better results recently they’ll be hoping Port Vale don’t win too many of their games in hand.

Model performance: Rightly suspicious of their good start

Given their poor performances in the Championship, the model expected that mid-table was more likely than an immediate promotion push and so it has proved. Charlton have been hovering around the middle of the division for most of the season and look set to remain there.

Model performance: Very close all season

Chesterfield were expected to struggle against relegation and that’s unfortunately proved to be the case. The model’s assessment was unaffected by their bright start and since mid-October the Spireites have mostly been sat inside the bottom four. For most of that time they’ve looked in trouble and in recent weeks the gap between them and safety has grown forbiddingly wide.

Model performance: Unfortunately spot on

Coventry’s disappointing end to last season meant that they were expected to loiter in lower mid-table and this still looked likely after a poor start was overcome in November. However the wheels soon came off again and since mid-January the Sky Blues have looked destined for League 2. In the model’s defence, a lot has been happening off the field including multiple managerial changes.

Model performance: Too optimistic, albeit with mitigating circumstances

The model thought that Fleetwood were far better than their lowly finish last season and tipped them for a top half finish this time around, but underestimated just how much would go right for the Cod Army. To the model’s credit, it thought they were better than their league position in the first half of the season and had them entering the play-offs in November, but their surge in the second half of the season came a bit out of the blue.

Model performance: Correct to be optimistic, but perhaps didn’t go far enough

The Gills were assessed as being set for a lower mid-table finish in pre-season and that expectation has barely changed since then. A bright start quickly evaporated and they had fallen in line with the model’s prediction by mid-October, remaining on track ever since. While they were challenging for promotion for much of last season, their underlying performances had been ebbing for some time and that’s continued this term.

Model performance: Very close for most of the season

If it wasn’t for some inexplicably leaky defending early in the season I’m pretty sure that Millwall would be sitting a few places higher. They’ve looked consistently solid since the back end of 2015/16 and still look to have done enough to fulfil the model’s pre-season expectation of a play-off finish.

Model performance: Pretty close, if a shade optimistic

It’s been a frustrating season for MK Dons, who were expected to challenge for a play-off place by the model. They’ve dominated matches consistently without getting the breaks and have consequently slid towards mid-table obscurity and the bottom half of it at that. Since mid-January they’ve bounced back somewhat as the model predicted, but overall it’s not gone as the data indicated.

Model performance: Didn’t see such a prolonged slump coming

Northampton defied the model – and probably most other predictions – in storming to promotion last season but were far easier to assess this time around. Lower mid-table was expected in pre-season and when a promising start fizzled out that’s looking almost certain to be the final outcome.

Model performance: Correctly predicted a lower mid-table finish

Oldham’s attacking output has been worryingly low all season and is still getting worse, with the pre-season expectation of a relegation struggle has been proven correct. While they currently sit two points outside the drop zone they have played three more games than Port Vale and therefore aren’t out of the woods yet.

Model performance: Unfortunately foresaw a struggle against relegation

The model was sufficiently impressed with Oxford’s strong showing during last season’s promotion campaign to tip them for a play-off spot this time around. It maintained its optimism despite their inconsistent start and has begun to look prescient as they’ve gradually climbed the table. While they may have left it too late to break into the top six, it looks like they’ll come pretty close.

Model performance: Optimism has been justified after a tricky start

Peterborough look to have had “upper mid-table” stamped across them right from the beginning this season. While they’ve tightened up the defence that was undermining the efforts of their swashbuckling attack, their ratings have consistently fallen short of the top six overall and their incursion into the play-off zone was consequently short-lived.

Model performance: Looks to have been pretty accurate throughout

Despite Port Vale sitting in the play-offs in mid-October, they were creating so few chances that their success looked unsustainable and the model’s pre-season expectation of a relegation battle had barely budged. Reality eventually caught up with them and they now need to make the most of their games in hand to haul themselves out of the bottom four before it’s too late.

Model performance: Correctly anticipated a relegation battle

Rochdale endured a horrible start to the season but their underlying performances remained relatively stable, so the model’s prediction barely budged. Since then they’ve flirted heavily with the play-offs after an impressive run before another big dip in both form and performances has seen them drift back towards the pre-season prediction of upper mid-table.

Model performance: Pre-season prediction looks pretty accurate

While the Iron were sufficiently impressive last season to be tipped as likely play-off contenders, some insanely – and unsustainably – clinical finishing propelled them into a title challenge. Despite leading the division for a long time the model remained slightly cool on their chances and their recent dip in form looks to have justified that outlook.

Model performance: Looks correct to have remained cool on title prospects

The model saw plenty that it liked in Sheffield United even before the season began and consequently didn’t alter its prediction of automatic promotion when the Blades took one point from the first 12 on offer. By the start of November it was 90% certain that they’d finish in the top two and that has continued to look more and more likely.

Model performance: Looks to have called their title challenge early

Shrewsbury went into this season as the division’s worst-rated team and there was little to suggest that this would change until about a month into Paul Hurst’s tenure. Since late November their revival has been impressive and while they’re not out of the woods yet their survival prospects are far, far healthier than just a few months ago.

Model performance: Didn’t see their impressive turnaround coming

The model didn’t think much of Southend last season and assumed a lower mid-table finish would be the likeliest outcome this time around. However – while it remained optimistic during their poor start – the scale of their improvement from mid-October onwards caught it by surprise. The ratings still suggest that a play-off finish would be a step too far, but overall the Shrimpers have massively exceeded expectations.

Model performance: Underestimated the extent of their recovery

The model’s pre-season prediction was that Swindon would have a relegation battle on their hands and that’s proven to be the case. The Robins have been hovering near the bottom four since early October and have recently been dragged into it despite some recent improvements in their performances. The difficulty of their remaining schedule means that the four-point gap separating them from safety may prove unbridgeable.

Model performance: Foresaw the current relegation struggle

In the early part of the season, Walsall’s ratings were in free fall as the players brought in to replace those poached over the summer struggled to adapt. Some uncertain displays depressed their predicted finish further into the “lower mid-table” range but things eventually began to click and the Saddlers’ recent surge in both performances and results has caught the model by surprise. A top half finish suddenly looks likely, which would cap an impressive recovery.

Model performance: Didn’t anticipate their recent surge

Wimbledon looked sufficiently impressive in gaining promotion that the model pegged them as being capable of a mid-table finish. Despite an uncertain start – which didn’t perturb the model all that much – that prediction has proved pretty accurate, with even a brief cameo in the top six along the way.

Model performance: Pre-season expectation looks pretty accurate