Read on Twitter
Evaluating the usefulness of Key Passes as a metric:
[THREAD]
(All stats from http://fbref.com and http://understat.com )
](https://pbs.twimg.com/media/EryleY7XIAcl8uN.jpg "Evaluating the usefulness of Key Passes as a metric:[THREAD](All stats from http://fbref.com and http://understat.com )")
[THREAD]
(All stats from http://fbref.com and http://understat.com )
In this thread I will:
- Explain what a Key Pass represents
- Discuss how Key Passes are used when evaluating a player
- Illustrate the flaws behind the Key Pass metric
- Provide alternatives to Key Passes for evaluating a player
- Explain what a Key Pass represents
- Discuss how Key Passes are used when evaluating a player
- Illustrate the flaws behind the Key Pass metric
- Provide alternatives to Key Passes for evaluating a player
What is a Key Pass?
A Key Pass is awarded to the player who makes the pass right before a shot is taken. Really, an assist is actually just a Key Pass which results in a goal, and in that case, a player is awarded both a Key Pass and an assist for their contribution.
A Key Pass is awarded to the player who makes the pass right before a shot is taken. Really, an assist is actually just a Key Pass which results in a goal, and in that case, a player is awarded both a Key Pass and an assist for their contribution.
How are Key Passes used to analyze a player?
Key Passes are an indicator of how "creative" a player is. Common sense would tell you the more Key Passes a player accumulates, the more scoring opportunities they generate for their team, and thus, the more "creative" they must be.
Key Passes are an indicator of how "creative" a player is. Common sense would tell you the more Key Passes a player accumulates, the more scoring opportunities they generate for their team, and thus, the more "creative" they must be.
Let's take a look at two players, both of whom are 20, English, and play in the Premier League.
Player A:
- 333 minutes played itl
- 2.97 Key Passes/90
Player B:
- 324 minutes played itl
- 1.67 Key Passes/90
Just looking at Key Passes, Player A is 75% more "creative" than B.
Player A:
- 333 minutes played itl
- 2.97 Key Passes/90
Player B:
- 324 minutes played itl
- 1.67 Key Passes/90
Just looking at Key Passes, Player A is 75% more "creative" than B.
Now obviously, using one stat to compare players doesn't tell the whole story, so concluding that Player A is more "creative" than Player B based purely on Key Passes is logically flawed.
However, even if using one stat wasn't logically flawed, using Key Passes is.
However, even if using one stat wasn't logically flawed, using Key Passes is.
Why is the Key Pass metric flawed?
There are two major reasons why this metric is flawed:
1. It is dependent on a teammate actually getting a shot off
2. Not all Key Passes are created equal
There are two major reasons why this metric is flawed:
1. It is dependent on a teammate actually getting a shot off
2. Not all Key Passes are created equal
The Key Pass metric's dependence on teammates:
As stated earlier in this thread, a player will only get credit for a Key Pass if the player who receives the pass takes a shot. There are two scenarios in which the Key Pass metric can sell a player's creativity short.
As stated earlier in this thread, a player will only get credit for a Key Pass if the player who receives the pass takes a shot. There are two scenarios in which the Key Pass metric can sell a player's creativity short.
Scenario 1:
In this clip, Granit Xhaka really creates the goal, as he spots the run of Tierney and plays the line-splitting pass, but he would not receive credit for a Key Pass in this scenario.
In this clip, Granit Xhaka really creates the goal, as he spots the run of Tierney and plays the line-splitting pass, but he would not receive credit for a Key Pass in this scenario.
What metric should be used in place of Key Passes in this scenario?
In the case of pre-assists like this one, Shot-Creating Actions (SCA)/90, is really useful, as it accounts for the two actions prior to the shot, which would give Xhaka credit for his pass.
In the case of pre-assists like this one, Shot-Creating Actions (SCA)/90, is really useful, as it accounts for the two actions prior to the shot, which would give Xhaka credit for his pass.
Scenario 2:
This clip, which incorporates some of Scenario 1 with Xhaka's pass, shows Smith Rowe playing a low cross into the box, which would surely find an unmarked Lacazette if not for Bellerin. Since no shot is taken, this doesn't count as a Key Pass.
This clip, which incorporates some of Scenario 1 with Xhaka's pass, shows Smith Rowe playing a low cross into the box, which would surely find an unmarked Lacazette if not for Bellerin. Since no shot is taken, this doesn't count as a Key Pass.
Really, any situation which involves a great pass which results in a teammate not being able to get a shot off, whether they can't get the ball out of their feet, they get dispossessed, or it's an enticing cross that isn't met, falls under this category.
What metric should be used in place of Key Passes in this scenario?
Honestly, there isn't really a way to quantitatively measure these "near" misses. Just watch football. Tierney has missed out on loads of Key Passes because our forwards don't get on the end of his crosses.
Honestly, there isn't really a way to quantitatively measure these "near" misses. Just watch football. Tierney has missed out on loads of Key Passes because our forwards don't get on the end of his crosses.
Not all Key Passes are created equal:
The biggest drawback of the Key Pass metric is that a player either records a Key Pass or he doesn't, and there's no variability in the metric to account for the quality of each Key Pass.
The biggest drawback of the Key Pass metric is that a player either records a Key Pass or he doesn't, and there's no variability in the metric to account for the quality of each Key Pass.
Given the way that Key Passes are awarded, the "creativity" of a player can be under or overvalued, depending on the "quality" of Key Passes they play.
Let's look at two clips, featuring Emile Smith Rowe, that show this.
Let's look at two clips, featuring Emile Smith Rowe, that show this.
Clip 1:
This sequence finishes with ESR playing a cross to an unmarked Saka for a tap-in. Smith Rowe fully deserves the Key Pass he records here, as he had great off the ball movement and made an extremely effective pass.
This sequence finishes with ESR playing a cross to an unmarked Saka for a tap-in. Smith Rowe fully deserves the Key Pass he records here, as he had great off the ball movement and made an extremely effective pass.
Clip 2:
In the same game against West Brom, ESR drives inside from the left wing and passes it back to Ceballos, who takes a hopeful shot from 25 yards out. ESR records his second Key Pass of the game in this sequence.
In the same game against West Brom, ESR drives inside from the left wing and passes it back to Ceballos, who takes a hopeful shot from 25 yards out. ESR records his second Key Pass of the game in this sequence.
If you haven't figured it out already, the problem here is the difference in the quality of scoring chances in these two clips. The Key Pass metric doesn't differentiate between the two, and doesn't accurately quantify "creativity."
What metric should be used in place of Key Passes in this situation?
Whenever the recipient of a pass gets a shot off, xA will always be a more accurate metric than Key Passes, as it is more precise and flexible.
Whenever the recipient of a pass gets a shot off, xA will always be a more accurate metric than Key Passes, as it is more precise and flexible.
What is xA?
Expected Assists, or xA, measures the quality of shot created from a pass. It is measured on a scale of 0 to 1, and based on previous shots in the same exact situation, tells you the probability of scoring from that position.
Expected Assists, or xA, measures the quality of shot created from a pass. It is measured on a scale of 0 to 1, and based on previous shots in the same exact situation, tells you the probability of scoring from that position.
Applying this idea to the two clips above:
Clip 1 (Saka Goal):
- Worth 0.88 xG
- ESR is awarded 1 Key Pass and 0.88 xA
Clip 2 (Ceballos Shot):
- Worth 0.02 xG
- ESR is awarded 1 Key Pass and 0.02 xA
Clip 1 (Saka Goal):
- Worth 0.88 xG
- ESR is awarded 1 Key Pass and 0.88 xA
Clip 2 (Ceballos Shot):
- Worth 0.02 xG
- ESR is awarded 1 Key Pass and 0.02 xA
As shown by the two examples above, not all Key Passes are created equal, and being able to account for the difference in Key Passes is what makes xA a much more viable metric.
In short, there are two major flaws with the Key Pass metric: It's dependent on a player's teammates to get a shot off, and when they do take a shot, the Key Pass matric provides no way to glean the value of each Key Pass.
Let's go back to those two unknown players, and add their xA's to the stats.
Player A:
- 333 minutes played itl
- 2.97 Key Passes/90
- 0.30 xA/90
Player B:
- 324 minutes played itl
- 1.67 Key Passes/90
- 0.29 xA/90
Player A:
- 333 minutes played itl
- 2.97 Key Passes/90
- 0.30 xA/90
Player B:
- 324 minutes played itl
- 1.67 Key Passes/90
- 0.29 xA/90
When looking through the lens of xA, it's clear that A and B generate a similar total quality of chances each game. Let's do one more thing with these stats.
By calculating the xA/Key Pass, we can determine the average quality of shot a player creates for his teammates.
Player A:
- 2.97 KP/90
- 0.30 xA/90
- 0.10 xA/KP
Player B:
- 1.67 KP/90
- 0.29 xA/90
- 0.17 xA/KP
Player A:
- 2.97 KP/90
- 0.30 xA/90
- 0.10 xA/KP
Player B:
- 1.67 KP/90
- 0.29 xA/90
- 0.17 xA/KP
Player A, although he creates more chances for his teammates, they are lower quality than Player B, who creates less, but higher quality chances.
This leads to the argument between quantity and quality, and clubs just have to determine which profile of player they like more, or which profile fits the squad they have in place.
For the people who are wondering who the two players are, Player A is Chelsea's Callum Hudson-Odoi, and Player B is Arsenal's Emile Smith Rowe. Imo both are extremely bright prospects, especially in terms of their "creativity."
Conclusion:
To wrap this thread up, I'd just like to talk about stats as a whole. Many people will tell you that if you use stats to evaluate performances of players and teams, you don't understand football.
To wrap this thread up, I'd just like to talk about stats as a whole. Many people will tell you that if you use stats to evaluate performances of players and teams, you don't understand football.
However, I don't believe this is the case. For me, people who don't know anything about the stats they use, don't understand football. It's always important to understand what each stat shows, as well as the context behind each stat.
Ultimately, you need to both watch football and use meaningful stats to evaluate players. The point of this thread was to show how one stat might not be as meaningful as people think it is.
Overall, the Key Pass metric is useful at a simple level, but in every situation, there is something better that can be used.
Thanks for reading!
If you enjoyed this thread or learned something, a like and retweet would be very appreciated.
If you enjoyed this thread or learned something, a like and retweet would be very appreciated.
