Is there any value in dropping Jack Stevens?

Two more fabulous goals scored, but three cheap ones conceded too. 

After a fourth 3-2 in six home matches for United, it is clear where their deficiencies lie. 

Robinson's side are scintillating going forward, able to slice through teams at will. 

Quality finishing is helping, too, with Billy Bodin netting a fabulous brace against Bolton, only to see his heroics cancelled out by more disappointing defending. A story which is becoming increasingly familiar.  

Oxford's tendency to concede has inevitably led to questions at the back, and more specifically the goalkeeper. 

For the first time in his short professional career, Jack Stevens' performances are coming under genuine pressure. Declan John's equaliser on ten minutes from a seemingly implausible position only adding fuel to the fire. 

The vast majority of fans would still support Jack Stevens, and his impact since replacing Eastwood in November last year has been undisputed, but Oxford are certainly losing out on points in games where they are the better side. 

But how much of this can we attribute to Stevens, and is it time to give Eastwood another go between the sticks. 

... 

At many points this season, it has felt like United were unlucky to concede as many as they have. 

However, looking at the data, this doesn't seem to be the case. 

Expected Goals Against (xGA) is a measure of the quality of chances a team allows their opposition to have, showing the number of goals a team can reasonably expect to concede given the opportunities they allow. 

The below table illustrates each League One team's xGA difference (the difference between their Expected Goals Against and the number of goals they have actually conceded). A positive number indicates they have conceded fewer goals than we would expect given the chances the concede, and a negative xGA difference means they have conceded more goals than we would expect. 

 

The number of goals a team has conceded compared to what we would expect

The table shows that, although Oxford are slightly below average for xGA difference, they are still in the positive, meaning they concede less goals than they should given the chances they concede. 

The difference is marginal, with Oxford's xGA sitting at 42.56, whereas the number of goals they have actually conceded is just over one and a half less at 41. 

This would seem to point in favour of Jack Stevens, suggesting he has prevented more goals than we would reasonably expect him to. 

However, there are numerous caveats to this. 

The first and most obvious is that Jack Stevens has not played every game this season, with injury and illness ruling him out for considerable chunks of the campaign. 

He has still played more than any other keeper, with 56% of the minutes, but United's stopper role has also been shared with Simon Eastwood (39% of total minutes) and Connal Truman (5%). 

This means we cannot attribute all of Oxford's defensive numbers to Stevens. 

The other caveat to the data is that xGA does not account for the quality of shot made by the opposition. This means a player six-yards-out who scuffs his shot straight at the goalkeeper will still record a high xG given the quality of his position, despite the fact it was not a difficult save to make for the keeper. 

A more accurate measure of an individual goalkeeper's record is Expected Goals on Target Conceded (xGOT), which measures both the quality of chance and the location in the goalmouth a shot ends up. The terminology sounds confusing, but it basically means more credit is given to a shot which ends up in the corners, than one which is straight at the goalkeeper. 

For this we can use Opta's The Analyst Tool, which is a free to use website providing in depth analysis on all of Europe's top divisions, including the EFL. 

Looking at this data, the picture doesn't look as promising for Stevens. 

Stevens has faced 71 shots this season, totalling at 16 xGOT. The number of goals he has actually conceded, though, is 24, meaning he has conceded eight more goals than you would expect from the quality of shots he has faced this season (excluding penalties and own goals). 

This ranks him as the second worst keeper in League One by xGOT difference. 

xGOT difference of all League One keepers to have played at least 50% of the minutes.

It is notable that the majority of goalkeepers register a negative xGOT difference, so Jack Stevens is certainly not alone in allowing more goals than the metric suggests he should. 

The extent of this is worrying, though, and gets worse when the number of shots they actually face is factored in. 

Goals Prevented is a metric which calculates xGOT difference in proportion to the number of shots they face, and gives a rate of goals prevented per shot. A ranking of one for goals prevented means they prevent as many goals as they were expected to per shot, whilst a number below one means they concede more goals per shot than they are expected to. 

Of all goalkeepers who have played over 50% of their team's minutes this year, Jack Stevens has the lowest goals prevented rate. 

The number of goals a keeper prevents per shot

It should be noted that Ipswich goalkeeper Vaclav Hladky does rank below Stevens on this metric, with a 0.57 goals prevented rate, but he has only played 38% of Ipswich's minutes so doesn't qualify for the table. 

Hladky was dropped after 13 matches, at which point Ipswich were conceding 1.54 goals per game, and have since conceded 0.85 goals per game. It is no surprise that some are hoping Robinson takes similar action with Stevens, then. 

However, it would be harsh to judge Stevens purely on his shot stopping ability. Yes, this is the primary role of a goalkeeper, but the way Oxford play means so much more is required of their number one. 

Jack Stevens averages the fifth most passes per game (33.3) of all League One goalkeepers who have played at least 50% of the minutes, and is particularly adept at getting Oxford moving on the attack, with the second most key passes per game of all keepers. 

Despite his tendency to play risky, forward passed, Stevens also has a respectable passing accuracy of 57.8%, ranking him eleventh of the 23 keepers to have played over 50% of the available minutes. 

He is also significantly more comfortable with the ball at his feet than Oxford's current alternative, Simon Eastwood. 

Stevens' average of 33.3 passes per game is well above Eastwood's 27.9.

His range of passing is also far greater than his compatriot, with 5.6 accurate long passes per game to Eastwood's 4.2. 

Another trait difficult to quantify through numbers is a keepers ability to dominate his area, but this is another element where Stevens is on top. From set pieces he is significantly more comfortable coming to claim the ball, relieving a lot of pressure on Oxford.  

The 2021-22 version of Eastwood is also not a notably better shot stopper. 

His goals prevented rate of 0.74 is not much better than Stevens and is still within the lower reaches of goalkeepers (he is not included in much of the data as he has not featured in 50% of the minutes). 

On a very basic level, Stevens also has a better save percentage, at 66.2% to 63.9% for Eastwood. 

The poor numbers registered by both players on shot stopping indicates a fundamental playing style for Oxford; they will largely dominate matches, and when they concede chances, they are usually of a very high standard. 

This means Stevens' performances are probably not as disastrous as the stats suggest, and owe a lot to United's gung ho approach. 

It remains, though, that he has registered well below par numbers this season, and his shot stopping has led to Oxford conceding a few more goals than they should have, but it is not just this that matters. 

Stevens is good on the ball and a commanding presence in the box, whilst Eastwood is not a significantly better shot stopper either. 

It may be a bit early to remove Stevens from his number one spot, but you can certainly see why people are beginning to question his performances.