On the to-do list for the summer was to improve the Celtic goalkeeping position.

Joe Hart is now 37 with visible signs, especially in pre-season, of physical decline. Scott Bain is a useful third-choice “homegrown” squad member whilst Benjamin Siegrist is miles from the manager’s thinking. That - amongst many other things - did not happen, and Celtic entered a Champions League campaign and vital SPFL schedule where approaching £60 million may be on the line, in this situation.


Following a demoralising defeat to Heart of Midlothian who scored with their only two shots on target - one from Shunsuke Nakamura territory - does the data support the ongoing need to strengthen this position?

I am leading the witness here, but let’s dive in…


Goalkeeping metrics are much less well developed than for example attacking metrics such as expected goals, assists and no end-of-shot analysis data.

Whilst there is no magic bullet for any position, analysing shots that land on target, measuring the expected goals of them and comparing them to expected goals against them is useful. Using this concept, StatsBomb has come up with goals saved above average.

This has been defined as: “How many goals did the keeper save/concede versus expectation (post-shot xG faced)? This is representative of how many goals the goalkeeper's saves prevented within a season.” Let’s look at the league's goalkeepers this season, shall we?

Liam Kelly has conceded 2.7 goals more than expected in a Motherwell side that has not won since September 3. Jamal George plays for the bottom club Livingston. Kelle Roos is having a poor season overall compared to last and was horribly exposed the day Celtic put six past Aberdeen.

Then there is Joe Hart. He is the fourth goalkeeper to be in negative territory by this metric. This means he is costing his team goals as opposed to saving them overall compared to an average custodian. It is more worrying when you consider the amount of work he must do.

This shows the goals saved above average mapped with goals conceded. Celtic concede very few goals by SPFL standards and face relatively few shots. Therefore, racking up such a negative score must be seen in the context of how few overall saves he is required to make versus the other goalkeepers on the list.


How bad is it? Well, probably not as bad as Allan McGregor was for Rangers last season. StatsBomb reckoned he was -8.9 goals below average over the full season which was the worst in the league. Hart was on -3.3 last season and -1.0 the previous. Meaning that when he originally signed at 35 he could be considered competent – he makes the saves you expect but no more. The decline set in last season and is accelerating more this campaign.

The question is: to what extent will it cost Celtic in a vital league race? To see how this stresses under more pressure and against better opposition let’s take a look at the Champions League.

Champions League

Clearly, a six-game sample is very small. Here are all the major ‘keepers amongst the 32 group stage participants.

Hart is bottom ranked by this key metric across all 32 goalkeepers. His goals saved above average is -5.2. In six games! Extrapolate that over a 38-game season and that would be 33 goals conceded over average. It is nearly one goal per match conceded over what is expected had an average ‘keeper been in the net.

You could argue that Celtic’s campaign was doomed from the moment in first-half injury time that Feyenoord scored a free kick from nearly 40 yards out. He made many saves at home to Feyenoord, but all were within his increasingly reducing sphere of influence. He conceded one goal from a post-shot xG of 1.08.


When Celtic have achieved rare notable results in Europe, an eye-catching ’keeper display has been a key component. Think of the number and quality of saves from Fraser Forster at home to Barcelona and in both ties versus Lazio in 2019-20. It is that level of performance Celtic needs to eke out more points at this level. It is the one position where you can extract a marginal advantage compared to the bigger clubs.


The need to better challenge or even replace Hart has been identified many times over the last two seasons. The risks are now manifest and were partly responsible for the Champions League outcomes. Here is his radar comparison between last season and this.

The main differences are declines in expected save per cent and claims (e.g. coming for crosses and high balls). Both declines are further evidence of a reduction in physical capability. We may not be in the territory where this is league-defining, but it is certainly a live risk.