Thread by @Jhdharrison1

THREAD:

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.

To avoid having this issue I categorise every ball touch which is not a distribution or a save into either an intercepted cross or an intercepted throughball depending on the pass trajectory.

This allows the 2 key intercepting action areas to be analysed completely & separately.

Another issue with public GK intercepting data is the lack of information of whether the action actually prevented a chance occuring or not.

For instance many, but crucially not all, recoveries GKs make are just collecting over hit throughballs which would go out for goal kicks.

In this thread for each intercepting action I utilise pass trajectory & potential recipient location to evaluate whether the GK prevented a chance from occurring.

I define any action which prevents a potential chance occurring as a “pressurised” claim or a “pressurised” sweep.

The final issue with public intercepting data is that as they don’t include info on whether a chance is prevented they also don’t include any info about the quality of the chance prevented & obviously preventing a 15yrd 1v1 > preventing a pressurised possession 30yrd out.

In this thread I will overcome this by (as well as discussing pressurised claims & sweeps separately) utilising various NSExG grids along with each passes trajectory & potential recipient location to estimate the ExG prevented for each interception a GK makes.

I can then look at every cross & throughball a GK intercepts over a season & sum the individual ExG prevented values to reach a total ExG prevented due to intercepting actions.

I can then look at this for various #PremierLeague GKs & compare shot prevention to shot stopping.

In order to present how the model works I will go through a few actions #Alisson made in the 19/20 #PremierLeague & highlight their classification.

Here are 2 crosses:

vs #AFC if he doesn’t come possible free header -> pressurised

vs #MCFC no possible recipient -> unpressuired

Here are 2 throughball sweeps:

vs #WHUFC if he doesn’t come & pinch the ball off #Antonio’s toes #Antonio gets a clear 1v1 shot away -> pressurised

vs #AVFC the throughball is massively over hit & there is no possible recipient hence no chance is prevented -> unpressurised

Here are 2 pressurised claims:

1st: if the GK did not claim it it would have potentially given a free header close to goal between the goal posts -> ExG prevented 0.35

2nd: if the GK did not claim it the trajectory is such that only #Aguero could reach it -> ExG prevented 0.08

Here are 2 pressurised sweeps:

vs #NCFC if the GK did not sweep up this pass there would have been a clear 1v1 ~10yrd out -> ExG prevented 0.45

vs #WWFC if the GK did not sweep, a shot under defensive pressure from just inside the area could have occurred -> ExG prevented 0.05

Hopefully you have a feeling of how the model works so here is a graphic of every cross #Alisson claimed.

circles = unpressurised
stars = pressurised

⚪️ caught
🟡 punched away from danger
🌸 punched into danger
🔴 missed

Note: unpressurised crosses always have ExG prevented=0

A key part of the model I have not yet mentioned is what happens if the GK does not catch.

After every punch I track the punch trajectory & opposition player locations to estimate an ExG provided number.

The total ExG prevented is then the ExG prevented - the ExG provided.

Here #Alisson punches a cross into danger. My model suggests his decision to make contact prevented 0.08 ExG however his poor punch presented #BFC with 0.15 ExG therefore this action net provided #BFC with 0.07 ExG.

This is one of the red -ve ExG prevented stars on the graphic.

Punches into danger can still be net positive actions.

Here #Alisson punches a low cross back into danger however my model finds the initial intervention prevented 0.82 ExG while it only presented 0.22 ExG thus even though it went into danger the action net prevented 0.60 ExG!

Here is a graphic of every throughball #Alisson swept up last year, he’s so active!

circles = unpressurised
stars = pressurised

⚪️ retained possession
🟡 cleared away from danger
🌸 cleared into danger
🔴 missed the ball

(unpressurised throughballs always have 0 ExG prevented)

When throughballs are not retained they can also present ExG

Here #Alisson tried to intercept a ball aimed for #Son but #Son beat him to the ball & rounded him. My model suggests he prevented 0.15ExG but as the goal is open presented 0.32ExG thus net providing #THFC with 0.17ExG

Also poorly cleared throughballs can have net positive effects.

Here is an example, #Alisson only clears the ball to the edge of the box presenting 0.08ExG however as mentioned earlier he stops a clear 1v1 & prevents 0.45ExG thus this action it is still net 0.37ExG prevented!

So that’s the model!

What all does this mean for the importance of prevention?

The range in shot prevention numbers I’ve found = ~10 ExG per season

The range in shot stopping numbers = ~20 PSExG per season

So the best stopper & worst preventer > worst stopper & best preventer

In the extreme case stopping is more important but for individual cases this isn’t always true:

#DeGea saved 0.08ExG/90 more than #Alisson last year. But #Alisson prevented 0.16ExG/90 more than #DeGea thus looking at stopping alone would not indicate which GK had the better year

CONCLUSION

-I made an ExGprevented model which takes into account pass trajectory & player location

-The model evaluates how useful/effective intercepting actions are

-The range in ExG prevented is ~10ExG per year

-Prevention is important but not as important as stopping

EXTRAS:

If you are interested here are #Alisson’s top 3 & bottom 3 throughball sweeps + crosses claimed last year according to my model.

Top 3 throughball sweeps

vs #Brighton +0.44 ExG prevented
vs #Norwich +0.37 ExG prevented
vs #Villa +0.35 ExG prevented

Top 3 crosses claimed

vs #Burnley +0.65 ExG prevented
vs #Watford +0.60 ExG prevented
vs #ManchesterCity +0.35 ExG prevented

The catch against #Burnley really wasn’t praised highly enough at the time he makes up ~8yrds to dive at the feet of #Wood & prevent a huge chance!

Bottom 3 throughball sweeps

Vs #Tottenham -0.17 ExG prevented
vs #Arsenal -0.15 ExG prevented
vs #Watford -0.05 ExG prevented

#Alisson does get beaten to the ball when coming for throughballs quite often however it’s hard to get it right every time when you sweep so often.

Bottom 3 crosses claimed

vs #Burnley -0.15 ExG prevented
vs #Brighton -0.11 ExG prevented
vs #Burnley -0.07 ExG prevented

#Alisson noticeable struggled with crosses far more since the restart potentially it was just due to rustyness, luckily VVD was on hand to bail him out.

That’s everything folks!

If you have any further questions or what to know anything else about my cross claiming & throughball sweeping model just ask away!

#PremierLeague #GKunion #GKanalysis #Alisson #DeGea