Read on Twitter
THREAD:
Re-Introducing Possession Value Pass Maps.
Re-Introducing Possession Value Pass Maps.
OK. Firstly, what is Possession Value (PV)? It's either called Expected Possession Value (EPV) or Possession Value but it's essentially a model that looks at how much scoring value a player (or team) brings to a possession based on their actions (e.g. passing, dribbling, etc).
This is determined by looking at the probability a goal will be scored at the end of the possession in which the action occurs. Example: a pass that ends up in the 6-yard box has a higher scoring value than a pass between CBs. PV just puts a number to it based on historical data.
Is it perfect? No. It does miss a few things such a passes during transitions, some passes that aren't valuable in terms of scoring but valuable in other areas, etc. As a result, it doesn't measure the value of each possession and so shouldn't be used in small samples.
The value can be anywhere from 0 to 1. 0 means it added no value to the probability of a goal being scored and 1 means it added 100% probability to a goal being scored — near enough impossible. The highest PV pass is 0.567/56.7%. I usually use % as it translates better.
~ = roughly
Let's say Player A passes to Player B twice. The first pass was sideways (slightly forward) in the middle of the pitch. The PV is ~0.001 which is 0.1% — most common. They complete another pass but this time from DM to AM. The PV is ~0.01 which is 1% — really good.
Let's say Player A passes to Player B twice. The first pass was sideways (slightly forward) in the middle of the pitch. The PV is ~0.001 which is 0.1% — most common. They complete another pass but this time from DM to AM. The PV is ~0.01 which is 1% — really good.
Someone may ask how 1% is good. Well, it's because a lot of passes in football are ~0.1%. Football isn't a game with a lot of high-value passes. It's a lot sideways and backwards passes to create space with the odd progressive one. It's a create-space-exploit-space sport.
So, a pass with a PV of ~0.1 (~1%) is really good all things considered. Then, of course, we have the highest value passes which mainly come from dangerous crosses, cutbacks, through balls, etc. These all end up with PVs of 0.1+ (10%+) which is fantastic.
This matches observatory analysis (a.k.a. “the eye test”), too, as these passes often create the best chances in football.
PV, like xG, is based on averages so bigger samples = better results. It's not a perfect metric and has its flaws of course.
PV, like xG, is based on averages so bigger samples = better results. It's not a perfect metric and has its flaws of course.
You can read up more on PV here:
https://www.statsperform.com/resource/introducing-a-possession-value-framework/#:~:text=Key%20takeaways,on-the-ball%20events.
https://www.statsperform.com/resource/introducing-a-possession-value-framework/#:~:text=Key%20takeaways,on-the-ball%20events.
Now, what about pass maps? A pass map is just the average positions of players and the most common passes plotted. Different people use different minimums to qualify a pass which is fine. I use 8 as a minimum for reasons I can explain later.
Explore the below.
Explore the below.
Here, I use 8-10, 11-13 and 14+ as my ranges. They are completely arbitrary as values but there *is* a benefit in them as the higher up I go the less passes are seen which means the more predominant ones end up showing clearly which is always a positive.
Likewise, I use five colours to indicate PV value which I've just labelled ‘high’, ‘mid-high’, ‘mid’, ‘mid-low’ and ‘low’ to avoid having to write the math on there as it's quite lengthy.
The bottom left has the highest combination and the right has some territory data.
The bottom left has the highest combination and the right has some territory data.
Here, we can see how Williams became out early-phase outlet down the left with Maguire, McTominay and Pogba all using him whilst Wan-Bissaka wasn't used despite being an outlet with more space. This is because the players aren't yet comfortable with AWB on the ball.
As we were trying to make a comeback we can see the red and orange all at the top. Why? Because we were aggressive in our passing which resulted in a lot of direct, high-value passes. Usually, a map doesn't look like this which is interesting.
There is more analysis I could do but I wanted to just re-introduce these maps so that we could get used to the things I used to produce before. I'll be producing these for every game.
Well, there you have it. Possession Value Pass Maps.
Note: I'll definitely be making little changes to them here and there but the base will remain the same.
I hope you enjoy them.
Bless.
Note: I'll definitely be making little changes to them here and there but the base will remain the same.
I hope you enjoy them.
Bless.
