John Harrison Profile picture
University of Cambridge Astronomy PhD graduate ☄️🔭🌟, Freelance GK analyst specialising in statistical analysis, Work with @gkintelligence & @GKUnion🧤⚽️📊
18 May
What a save from #DeGea

Unbelievably calm GKing in a crucial moment

It may look like a simple save but that’s all because of his exemplary decision to use the wait & react 1v1 strategy as the 1v1 was from long range!

Not many GKs could make such a good decision

The striker was ~16 yrds from the centre of goal & at a pretty wide angle (hard to see from the photo) so #DeGea’s decision to stay deep made the finish very difficult as he didn’t rush out & gift a simple dink/rounded finish
The great thing about #DeGea’s positioning here is that if the striker takes a touch rather than shooting he’s far enough off his line & prepared & ready to engage then, only once he knows the striker is so close to goal that he can no longer trust his reactions to make the save!
Read 8 tweets
7 May
After showing us how a GK can maximise their chances of saving a close range 1v1, #Schmeichel has just show how us how a GK can minimise their chances of saving a long range 1v1!

By engaging early he allows Wilson to easily round him with 1 touch!

He needed to wait for longer!
If #Schmeichel waits inside his 6 yard box for #Wilson’s touch, #Wilson then has a decision to make:

1. Shoot from 17yrd knowing defenders are track back.

2. Take a touch knowing he will bring it closer to goal but also closer to the GK.

Regardless of what decision #Wilson makes #Schmeichel will have a higher probability of making the save than he would have if he rushes out & engages early like he does.

1. If #Wilson shoots #Schmeichel’s deep position will maximise his reaction time & thus his save probability.
Read 6 tweets
7 May
What a save by #Schmeichel!

There are 2 crucial decisions which made this save possible:

• The decision to engage early -> minimised the distance to the ball thus maximised the goal area covered!

• The decision to spread -> spreading further maximised the goal area covered!
#Schmeichel quickly realised the 1v1 was going to be from close range & he wouldn’t be able to react to the shot if he stayed deep & thus he needed to rush out & form a premeditated barrier

Without this quick decision he wouldn’t have gotten close enough to ASM to make the save!
Even once he gets out to ASM it could still be a goal but #Schmeichel’s excellent spread technique means he fills a huge portion of the goal & manages to get a toe on the ball & deflect it out for a corner.

Using a different barrier technique would’ve likely resulted in a goal.
Read 4 tweets
4 May
I’ve always thought the discourse surrounding #Pickford is odd

Twitter would make you think he is the worst GK to ever play in the #PremierLeague which isn’t true, he’s an avg #PL shotstopper!

Yes he had a bad 2020 but his recent good form is starting to make up for that

Pickford’s #PremierLeague Save% for #Everton = 68.9

Expected Save% for an avg #PL GK who faced the same shots as #Pickford in that time = 69.0

I understand why people think that #England’s #1 should be better than a middle of the road PL GK but the abuse he gets is just crazy!
Also regarding the #England #1 shirt, the main reason he has it is because currently #Southgate believes that being an average shotstopper is outweighed by the fact he is a level above any other English GK when it comes to distributing the ball, especially when under pressure!
Read 5 tweets
4 May
#PremierLeague Shotstopping Performance Graphic:

#Ederson #Areola #Martinez & #Sanchez have been class 🔥

#Henderson has outperformed #DeGea when it comes to shotstopping 🔄

Despite criticism #Ramsdale #Pickford & #Guaita have performed at a roughly average top flight level 👀
Explanation of ExSaves:

ExSaves are calculated by looking at every shot a GK faces & using the shots distance, angle, trajectory, final location, the number of defenders in the way + other things & using historical data to find the probability that an avg PL GK makes the save...
...Therefore as a rough example a shot from 40yrds at a tight angle that goes down the middle of goal with all the teams defenders behind the ball may have an ExSave of 99% whereas a central 1v1 from 10yrds fired into the bottom corner may have an ExSave of 10%.
Read 4 tweets
13 Apr
Excellent 1st half from #Neuer which more than makes up for last weeks error!

Saving 1v1s continues to be his forte!

He’s putting on a 1v1 decision making & 1v1 technique masterclass

His 1st 1v1 stop was a lovely block which left #Neymar with no gaps to exploit!

His 2nd 1v1 save was a nice spread!
I loved the fact he backed off & didn’t commit himself when #Mbappe ran in 1v1 as he knew #Mbappe was too far from goal to trouble him then as soon as he squared it to #Neymar he realised the shot was now in the engage zone & he then rushed out
If you are wondering what I am talking about in terms of “block”, “spread”, “engage some” etc.

Please read this thread:
Read 5 tweets
12 Apr

#Henderson’s shot prevention has been class in his first 8 #PremierLeague games & its played a big part in #ManchesterUnited’s success!

He is preventing 3.07 chances occurring per game corresponding to ~0.44 ExG per game!

That’s double #DeGea’s ExG prevented per game!
Here I will compare #Henderson’s cross claiming & sweeping (the two branches of shot prevention) so far this year to #DeGea’s & discuss how this has affected #ManchesterUnited’s defence.

I will use my shot prevention model as sadly mainstream models miss many important actions!

#Henderson’s sweeping has been highlighted after his tidy display vs #THFC & it wasn’t a one off!

So far this year for #ManchesterUnited he has prevented 11 opposition chances occurring by sweeping which corresponds to preventing 0.15 ExG/90 (#PL avg is 0.10 ExG/90)
Read 12 tweets
11 Apr
For me #Martinez has been the #PremierLeague GK of the year

He’s shown class shot stopping, shot preventing & distributing skills & been excellent in 1v1 situations!

#Martinez has saved 11 1v1s this year & my model predicts an avg PL GK would‘ve only managed to save ~8 of them!
I don’t think it is unsustainable form either, especially when talking about 1v1s, because my model shows his excellent decision making is regularly making the saves easier for him!

His Post Decision Ex1v1 Saves is 1.42 saves higher than his Pre Decision Ex1v1 Saves!

Read 5 tweets
10 Apr
Ah #Alisson won’t want to see this again!

There’s 2 mistakes here:

1. Positioning: he’s 6yrd out for a central 15yrd shot thus he cuts his reaction time greatly without reducing the effective goal area a lot.

2. Technique: he goes with his hand rather than his foot!

Research has shown the optimum GK position for none 1v1 long range shots is roughly 1.5-2yrd off the line

Once a GK starts positioning themselves further out than this the reduced goal area they have to cover doesn’t outweigh the loss of reaction time hence goals like this occur
In this situation though #Watkins’ finish is actually a poor one as he smashes it down the centre, so #Alisson could’ve got away with his poor positioning but he compounded his first mistake by going with his hand rather than his foot!

#PremierLeague #LIVAVL
Read 4 tweets
10 Apr
#Ederson’s biggest weakness costs #ManchesterCity again!

His habit of rushing out & engaging during long range 1v1s consistently costs goals!

He needs to back pedal into his 6yrd box & force a proper finish

By rushing here he allows a central sidefoot (the easiest finish) in!
My 1v1 model shows that in this situation if a GK back pedals into their 6yrd box they will save this 1v1 91% of the time!

Whereas if they engage & spread as #Ederson did here they will only make the save 48% of the time!

His decision making made a simple save into a hard one!
Also in this situation it should be noted that #Dallas cannot dribble closer to goal due to the immense defensive pressure on him however even if he could dribble closer #Ederson should still wait & lure him in as engaging 1v1s gets more successful the closer the CF is to goal!
Read 5 tweets
26 Mar
Here’s a nice example of neutral set positions making saves easier

I’ve already discussed #Henderson’s propensity to set quite low this season & here even though he’s taller than #Pickford he tips an almost identical shot into the roof of the net while #Pickford catches it

Another key detail from this clip is how #Pickford moves back to his line while #Henderson moves forwards.

This tiny difference in distance thus reaction time can impact on a GKs ability to make the save as it can be the difference between getting 2 hands to the ball & 1 hand.
In this clip #Pickford & #Henderson also make lots of saves with their feet when the service is lower & #Henderson actually struggles with these too as his lower set position means he needs to get his hands out of the way so it doesn’t even give him an advantage for low shots
Read 6 tweets
9 Mar
#Henderson’s cross claiming has been exceptional in his first 5 #PremierLeague appearances

#Henderson boasts a 100% claim success rate while being twice as active the average #PL GK!

His claiming is preventing #ManchesterUnited conceding chances worth ~0.4 ExG per game!

For comparison, #DeGea cross claiming numbers this year are:

Chances Prevented = 0.81
(Only 41% of DH)

ExG Prevented = 0.15
(Only 35% of DH)

Average Claim Distance = 4.5yrd (Only 78% of DH)

Claim Success = 95%
(95% of DH)

Catch Fraction = 58%
(104% of DH)

For more information on the metrics I quote in the graphics above please see this thread, it explains how I calculate which cross claims prevent chances & which don’t & how I assign ExG values to each claim & why it’s important to look deeper than “total high balls claimed”!
Read 6 tweets
7 Mar
Surely @OptaJoe have to change this definition?

How can you classify a cross-shot as not a shot on target if it’s stopped by the GK but as a shot on target if it goes in!

Surely this definition screws up their ExG model? It’s artificially removing nongoal shots?!

#FPLCommunity Image
Surely any ball with a trajectory such that it would be a goal without a GKs intervention must be a shot on target & a save or at least a save & a “cross on target” if for some reason you want to make two separate categories to distinguish the two.

#PremierLeague #FPL
I didn’t realise Opta had such an odd definition until #MartinezGate

I understand potentially why you wouldn’t want to call it a “shot” but it has to be consistent


if saved = save
if goal = cross on target

You cant have them be shots on target if goals but not if saves!
Read 5 tweets
21 Feb

Have #DeGea’s cross claiming & sweeping numbers changed dramatically this season in comparison to last season?

Below I will look into #DeGea’s sweeping & cross claiming performance in the 19/20 #PremierLeague season & compare it with his performance so far this year! Image
Claiming 19/20:

#DeGea prevented 0.77 chances per 90 occurring by claiming crosses

Average #PremierLeague GK prevents 1.10 chances per 90 occurring by claiming crosses

Key points

• Attempts to claim infrequently

• Mostly successful when he does attempt to claim

#MUFC Image
Claiming 20/21:

So far #DeGea has prevented 0.74 chances per 90 occurring by claiming crosses

#DeGea has performed in a very similar manner against crosses this year with his willingness to claim & his success when doing so remaining roughly the same.

#MUFC Image
Read 7 tweets
3 Dec 20
#DeGea’s ability to stop close range 1v1s has dropped off greatly recently:


He often chooses the optimum 1v1 technique he just gets the execution wrong!

Last night he failed to ground his foot & thus the ball squirmed under him

It’s these fine margins which stop #DeGea replicating his form of 17/18.

Again with the #Neymar goal he does the right thing trying to use the block but he drops into it too far from #Neymar he needs to get closer to him & fill more of the goal area if he wants to make the save. Image
The 3rd goal is on #Maguire.

#DeGea again gets the decision right to try to block but this time he also nails the execution & gives the PSG player nothing to hit.

But as #Maguire does not trust #DeGea he drops back to the line rather than marking #Neymar giving him a tap in. Image
Read 5 tweets
29 Nov 20
JWP’s FK is class & it would have been difficult for #DeGea to save regardless but yet again he gives himself no chance by getting his body shape wrong given his positioning

He needs to start higher off his line so his body angle doesn’t force him to dive into the goal!

If #DeGea stands ~1yrd forwards he will be able to cover the wall side without diving backwards into his goal + it won’t hinder his ability to defend his side as while it will cut his reaction time slightly the distance to the far post is such that it won’t make a big difference.
If #DeGea moved too far forwards (ie 2 to 3 yards) then obviously it would make high saves to his side difficult & cut his reaction time by a substantial amount thus make any save more tricky but I’m not suggesting such a big change.
Read 4 tweets
27 Nov 20
Last year #Guaita only saved 36% of the close range 1v1s he faced (only #Kepa & #DeGea were lower)

This is because he rarely forms a premeditated barrier before the shot is struck so gives himself no chance of making the save!

This weakness was exposed again tonight.

When shots are from close range there is simply not enough time for a GK to react therefore they have to get close to the striker & form a premeditated barrier & shrink the effective goal area & force the striker to “just hit the ball at them”

As the images below show
I analysed 800+ 1v1s in the #PremierLeague last year & using a logistic regression model found that the engage & react strategy that #Guiata regularly employs gives a GK a far lower chance of making the save vs the engage & block & the engage & spread premediated strategies.
Read 8 tweets
19 Sep 20

How important is a goalkeeper’s ability to prevent shots?

Is it of comparable importance to shot stopping ability?

In this thread I will investigate these questions & outline my ExG prevented model using #Alisson as an example.

#PremierLeague #LFC #GKUnion #EPL
Every action a GK performs can be separated into 3 categories:

Save-> here the GK attempts to block the ball going into the goal

Distribution-> here the GK attempts to play the ball to a teammate

Interception-> here the GK attempts to stop the ball from reaching an opponent
Evaluating GK intercepting from public data is difficult as most stats providers split intercepting actions between recoveries, tackles, interceptions, punches, & claims, & the differences between the categories are unclear & not relevant to interception difficulty/importance.
Read 28 tweets
3 Sep 20

Over the past year has Dean #Henderson staked a realistic claim to be #ManchesterUnited’s number 1 #Goalkeeper ahead of David #DeGea?

I will mainly investigate an area of GKing that receives little statistical analysis:

Cross Claiming & Sweeping

#MUFC #GKUnion
Firstly before that, I think it is well known that #Henderson had an outstanding shot stopping season while #DeGea didn’t.

Here are the numbers:

#Henderson Save%-ExSave +5.88%
#DeGea Save%-ExSave% +0.78%

2 important conclusions should be drawn from these numbers...

1. #DeGea did not have a below avg season, yes he make shot stopping mistakes but he made enough worldy saves to out weigh the mistakes.

2. #DeGea has only once (17/18) had a better shot stopping season in his career (according to PSExG) than #Henderson had this year.

Read 7 tweets
17 Jun 20

1v1s are incredibly important situations for GKs to master as they correspond to 35% of all goals scored in the 19/20 #PremierLeague!

In this thread I will show how effective various 1v1 strategies are at making saves.

I’m sure #GKcoaches will be v interested in this!
I’ve analysed every single 1v1 situation in the 19/20 #PremierLeague (600+ situations) & calculated using logistic regression the optimum strategy a GK could use to maximise their chance of making the save depending on the situation location & the build-up leading to the chance.
I found that GKs use the following 5 strategies to save 1v1s:

1. Wait deep & then react to the shot
2. Engage & then react to the shot
3. Engage & smother the ball
4. Engage & spread
5. Engage & block

#PremierLeague #GKUnion
Read 16 tweets
29 Mar 20

Dean #Henderson vs David #DeGea
#PremierLeague Shotstopping 19/20

As requested I will compare the shot stopping performances of the two potential #ManchesterUnited number 1’s.

There will be statistics & 1v1 plots!


#MUFC #GKAnalysis #GKUnion
Firstly let’s look at post shot expected goals conceded.

@StatsBomb’s PSExG model takes the position of the shot, the position of the defenders, the speed of the shot & the placement of the shot into account in order to calculate whether an Avg GK would save it.
Using this model #DeGea has played almost exactly to the level of an average #PremierLeague goalkeeper & has conceded -0.2 goals more than expected.

#Henderson on the other hand has saved #SheffieldUnited +6.5 more goals than an avg GK would be expected to!
Read 15 tweets