Goalkeepers Season Review Part 1: Shot-Stopping

The 2018/19 Scottish Premiership season was one in which goalkeepers contributed heavily to its story-lines.

Alan McGregor made a heralded returned to Rangers, replacing Wes Foderingham as the club’s number one, while also being involved in several controversial moments. Scott Bain replacing Craig Gordon as Celtic’s number one was a clear example of how some managers now value a keeper’s distribution skills at least as much as their shot-stopping abilities. Hearts, Hibs and Kilmarnock all brought in new goalkeepers at the start of the season, St. Mirren recruited successfully during it, and at Motherwell Trever Carson was unfortunately side-lined from December onward due to illness.

Last summer I did a season review of goalkeepers for Modern Fitba, looking solely at shot-stopping. This season, thanks to the data provided by ORTEC sport, we can expand that to also involve aerial command and distribution. Will the numbers back up the ‘accepted’ narratives about which goalkeepers performed well and which ones struggled?

This first part will analyse what most fans will probably see as a goalkeeper’s number one job: shot-stopping.

Expected Saves

For a long time, the one major goalkeeping statistic was how many clean sheets a keeper had achieved through a season. It’s a woefully flawed stat category and tells us very little about the quality of a goalkeeper. Save percentage is better: how many saves a goalkeeper made compared to have many shots on target they faced will give us a lot more information about their abilities. But it is still flawed as it contains no information about where shots were taken from and how big a chance that attempt on goal came from.

Step forward Expected Goals. Used primarily to measure the quality of the chances an outfield players gets to, it can easily be reversed to fit a goalkeeper metric: how big were the chances a keeper faced and how many saves was he ‘expected’ to make from such chances? A more detailed explanation of how a xG model for goalkeepers is set up can be found both in last season’s review and this introductory piece, but the main difference from a standard xG model is that all shots that did not hit the target are removed.

For example, if a type of chance is scored 1 in every 4 attempts, it has a standard xG of 0.25 (as it is converted 25% of the time). However, remove all the shots that did not hit the target and you might find that those attempts that did resulted in a goal 50% of the time. So the xG in a goalkeeper model would for the same type of chance be 0.50.

As with all stat categories, there are still flaws here: the major one is that the ORTEC data does not tell us anything about the quality of the shot taken, which will obviously impact on the difficulty of any save. But a metric where we measure Goals Against (GA) compared to the Expected Goals Against (xGA), will tell us a lot more about a goalkeepers shot-stopping ability than Save Percentage or (shudder) amount of clean sheets. This metric is often also referred to as Expected Saves

Explanation out of the way, let’s check the results.

Keeper GA xGA Table.PNG

Hibs’ Ofir Marciano is the clear stand-out when it comes to shot-stopping in the Premiership this season. In fact, his stats are close to sensational: across his 21 games, he prevented 14 goals more than what could normally be expected, an average of 0.69 goals per every 90 minutes played (p90). He’s followed closely by Daniel Bachman and Mark Gillespie, who saved 0.37 and 0.35 more goals p90 than the average for the type of shots on target they faced.

Below is a map showing all the non-penalty shots on target Marciano faced last season, and where he made (or attempt to make) his saves. What immediately pops out is that he only conceded one goal from outside the penalty box, and that shot was not far outside either. In fact, he only conceded two goals in total form outside the ‘Danger Zone’, the middle area of the penalty box. Shots from outside this area will have a low xG, so conceding from them will have a clear impact on a goalkeepers’ shot-stopping stats. Equally, shots saved from this area will be a positive for goalkeepers, and Marciano has a cluster of saves from especially the right side of the Danger Zone.

Ofir Save Map.PNG

As mentioned earlier, Scott Bain replacement of Craig Gordon in the Celtic goal was seen to be more to do with his ability on the ball, but this season he also trumps Gordon when it comes to stopping the ball, his xGA is 0.25 better than Gordon p90 which is worth one less goal conceded per every four games. Bain did have the highest save percentage in the league, stopping 86% of all the shots he faced, but he also faced on average the lowest quality of chances against him, meaning he is only 4th when it comes to expected saves made.

Allan McGregor was named the SPFL Goalkeeper of the year, but his shot stopping stats are only the 10th best of the goalkeepers who played at least 9 full games this season.

There is a distinct difference in the performances of the two goalkeepers who appeared for Hibs, Motherwell, Kilmarnock and St. Mirren. While Ofir Marciano tops the table, Adam Bogdan is almost at the other end, in 16th. The difference between the two goalkeepers is stark: Marciano prevented 0.78 goals more than expected than Bogdan has per 90 minutes. Jamie McDonald’s shot-stopping performance this season wasn’t bad, making *exactly* the amount of saves expected of him according to the model – but Daniel Bachman gave his team a lot more, 0.37 more goals prevented per game to be precise. Vaclav Hladky was seen as a key part of St. Mirren avoiding relegation this season after he came in during the January window and the stats confirm this: he is 8th in the league compared to Craig Samson in second last place – a difference of 0.43 more goals prevented p90.

Carson shotmap.PNG

The starkest contrast between two goalkeepers at the same club was between Mark Gillespie and Trevor Carson. While he has been generally highly rated and even earning caps for Northern Ireland during his time at Motherwell, at Modern Fitba we’ve sounded alarms about Carson’s shot-stopping stats before: last year he was 11th out of 15 goalkeepers in the league. This season it’s even worse: not only is he at the bottom of the table, he is performing twice as bad as Craig Samson in the spot above him. In the games he played this season, Carson let in almost 7 goals more than was expected based on the shots he faced, an average of 0.57 per every 90. The improvement with Gillespie in goals was 0.94 – in other words, Motherwell was conceding almost a goal a game less compared to the expected rate with Gillespie in goals.

If that wasn’t bad enough for Carson, the quality of chances he did face was the 2nd lowest in the league. His stats could hardly be any worse.

Below it’s the table above illustrated on a graph - there’s also the added information about whether the saves made were held or parried away. Such characteristics of the save itself is what we’ll focus on in the second part of the analysis.

GK Graph.png

Save Action: Catch or Punch?

ORTEC’s data allows us to dig even deeper down in the detail and to start painting a picture of how the goalkeepers make their save, labeling each on with a range of different markers, including whether a save was ‘caught’ or ‘punched’.

Whether a save sees the ball held or parried will of course be influenced by the quality of the shot, but it can provide interesting extra information around what happens after a save is made: a caught ball wins possession for the goalkeeper’s team and ensures no further chances in the attack, while punching a shot into a dangerous area might even turn what was a small chance into a much bigger one.

Colin Doyle’s opportunities at Hearts were somewhat limited this season, but for 75% of the saves he made over his nine league games he caught the ball for – the highest rate in the league. While his sample is a bit on the small side, not a single one of the times he did punch the ball away did the opposition manage to get to a new shot within the next two actions on the ball - Doyle’s saves were indeed very safe.

Keeper Save Action Table.PNG

At the bottom we find Alan McGregor and Adam Bogdan, who respectively only caught the ball 39% and 36% of the times they faced a shot on target. The quality of the chance they faced can not explain such a low %, as those chances were just below average in the league. This is definitely something that would warrant further video analysts of the two goalkeepers, to see in detail why they only managed to hold on to the ball for about a third of their saves.

Daniel Bachman was second in the shot-stopping stats, and he’s also second when it comes to the % of saves that were caught. And while Ofir Marciano was superb when it came to shot-stopping and he was also above average in the % of his saves that were caught, 19% of his punched saves immediately led to another shot by the opposition. This was the second highest in the league, with Vaclav Hladky the clear ‘winner’: over a third of the shots he didn’t catch led directly to another attempt on goal. Is there an explanation for why these two goalkeepers parried so many shots back to the opposition? Well, actually…. 

Save Type: Stand or Dive, Feet or Hand?

One possible reason for Hladky’s and Marciano’s high rates for punched saves leading to another shot can be the fact that they also made the highest % of saves with their feet: Marciano with 19% and Hladky’s with a quite incredible 35%: over a third of the St. Mirren’s goalkeeper’s saves were with his feet.

Jamie McDonald and Craig Gordon had the 3rd and 4th lowest rate when it came to saves made with their feet – these guys like a diving save instead: at 66% and 65% they top the league for the percentage of saves that were made through a diving motion. The contrast to Gordon’s Celtic colleague Scott Bain is very stark: he had the highest rate of saves made while standing up at 65%, with only 35% made while diving.

Keeper Save Type Table.PNG

Again, this is not conclusive evidence in any way that Bain is better than Gordon at positioning himself correctly when facing a shot, thereby mitigating the need to make a dive, but it is interesting to note that 3 of the 4 goalkeepers with the highest rate of standing saves also make up 3 of the top 4 spaces in the overall shot-stopping table.

Conclusion:

Ofir Marciano is the outstanding performer when it comes to shot-stopping in the 2018/19 Premiership season. Hibernian, Kilmarnock, St, Mirren and especially Motherwell saw a significant difference between their two main goalkeepers this season, while Scott Bain more than justified his inclusion in front of Craig Gordon.

While Alan McGregor was seen as a key part in Ranger’s qualification to the Europa League groups stages, his shot-stopping stats in the league do not match up to the acclaim he received throughout the season.

Hopefully Trever Carson can make a full recovery from his illness and be back in time for the 2019/20 season. When he does return he has a lot to prove when it comes to shot-stopping.

The second part of this season review will look at how the goalkeepers make their presence felt in the air above their area.