#Karius made 8 saves vs #MUFC, but GKing is about quality not quantity!
5 of the saves had xSave>99% thus #PremierLeague GKs save them every time!
The 2 interesting & difficult saves were the 1v1 vs #Bruno & #Weghorst’s long range shot! Which beat #PL GKs 41% & 29% of the time!
It was these saves whose difficulty outweighed the 2 goals #Karius conceded meaning #Karius saved an above expected amount of shots in the #LeagueCupFinal
#Karius could do nothing with the #Casemiro goal (xSave probability<5%) but the #Rashford goal was a little more interesting
Our model at goalkeeper.com finds this shot is expected to beat a Premier League goalkeeper 41% of the time.
Thus, it’s not an error but is saveable.
#Karius’ timing and weight distribution were the major areas the German would change if he could replay the situation
Crucially, while we can dissect the details of the deflected goal #Karius made 2 big saves meaning that over the course of the game he was still a net positive for #NUFC
We should never abandon a player based on 1 #UCL game!
#Karius still has the skills to play at the top level
• • •
Missing some Tweet in this thread? You can try to
force a refresh
If #DeGea had rushed out to engage this touch he would’ve turned a 38% goal probability chance into a 61% goal probability chance as MA could’ve easily rounded him or chipped him!
By waiting he dictated the 1v1 to MA & made him make a decision, MA chose to take another touch…
Rather than continuing to stay deep #DeGea realised that now MA was close enough that his reaction time may not be enough to save the shot so he rushed & formed the premeditated block barrier reducing the goal probability to 34%!
#DeGea made the save look easy & made the save far easier for himself due to his top class decision to hold deep & then use his top hand. This made a difficult situation comfortable!
The GK xG model shown above looks at every shot faced, pass received, cross faced & through ball faced & calculates the probability of a goal occurring for & against the GK’s team before the event & after the event occurs!
Thus it measures & evaluates every action a GKs makes!
This allows all GKs to be given a single number measured in goals called
“Total Value in Goals”
Which describes the GK’s value to their side vs having a league average GK & takes into account everything they do meaning GKs of vastly different styles can be fairly compared!
When receiving the ball under pressure, both #Chelsea GKs have been below the #PremierLeague standard!
#Mendy has cost #CFC 0.52 xG more than the avg #PL GK would be expected to if they received the ball in the same situations (this is heavily influenced by his mistake vs Leeds)
Similarly #Kepa has cost #CFC 0.59 xG more than the avg #PL GK would be expected to if they received the ball in the same situations as he did in his last 15 #PremierLeague games, #Kepa’s numbers here are heavily influenced by his mistake vs #Liverpool which #Mane capitalised on!
A GK’s handling ability is vital & often under analysed!
Here #DeGea showed how handling can determine a game’s outcome when he parried a 49mph central shot into danger & the rebound was scored!
xRebound xG = 0.09
Actual Rebound xG = 0.47
xG added due to poor handling = 0.38
My handling model uses historic shot data to gauge how often certain shots should be expected to be caught, parried away from danger, & parried into danger in combination with a none shot xG model which assesses how likely rebounds are to be scored based on the parry location
My model found that due the shot’s lower than avg velocity & central trajectory an avg PL GK would expect to parry the ball into an area which resulted in a goal from the rebound only 9% of the time whereas #DeGea pushed it into an area which results in a goal 47% of the time!
My 1v1 model also finds the smother to be very vulnerable when the 1v1 is central as it often allows the GK to be rounded or chipped plus it does very little to cover the side the GK does not have their hands!
A spread wouldn’t drastically increased the Rambos save probability!