They may well be the most admired players in the world. Players serving high quality goal scoring chances to their team mates virtually at will. Be it by sliding through crowded defenses, with cutting edge though-balls, by smashing in sharp crosses from the wing, or whatever method works. This post will look at the best way to identify creative talent. Players that allow other players to score goals.
We’ve started off this scouting mini-series by looking at strikers, and should you have missed that one, I urge you to read that first. It makes digesting this one a lot easier. Oh, and just like always, data is collected by OPTA and acessed through Squawka, If it were not for them, no scouting pieces here!
The problem with assists
If you’d ask anyone in football, they would probably tell you that creative talent should be identified by assisted goals. No assists, not good. Lots of assists, good. But just like in striker scouting, this methods only looks at output and it may prove very misleading, and therefore quite costly. Some open doors…
1. An assist is only an assist if the striker finishes the shot, no matter how high the quality of the key pass. Looking at assists underestimates players that provide high quality key passes to poor or unlucky strikers. In the video it’s Twente’s Luc Castaignos that needs a rebound to put the ball past the keeper. No assist for Gutierrez, but oh man, that pass…
2. A poor key pass may count as an assist if the striker somehow turns a goal out of it. In the example video it’s a defensive header that earns Jon Obi Mikel an assist. This video also features on an excellent post by @MacAree on this very point.
3. Assists are less frequent than goals, and therefore much more susceptible to fluctuation. Unfortunately you can’t link a video to that point…
A workaround to these problems is to apply the Expected Goals idea to key passes. Our Expected Goals (ExpG) formula assigns each shot a chance of finding the back of the net, but we could do likewise with each pass leading to a shot. After all, if your creative midfielder creates three one-on-ones, he should be rewarded higher than a winger creating five headers off crosses, right?
So we take the ExpG for each shot, and if the shot comes off a key pass, the player providing the key pass is rewarded with the Expected Assists, or ExpA. Long-term, this ExpA value should align with the total number of assists. This method has some important advantages, as has also been shown earlier by Colin Trainor in this sublime piece on Stats Bomb.
1. ExpA is not susceptible to fluctuation like assists are, because a player may note 0.2 ExpA per match, rather than the odd assist.
2. ExpA removes the finishing from the equation. Players who regularly provide key passes to top strikers (Dani Alves anyone?) get more assists than players serving poor strikers. ExpA corrects this by assuming an average player to apply the finish.
ExpA in the Eredivisie
So here’s the top creative talent of the Eredivisie. Feel free to click on the chart for the full size version. In line with our intuition, Twente’s Dusan Tadic is the absolute number one creative player around, with over 0.5 ExpA per 90 minutes played.
I’ve limited the sample to all players that have played at least half of all available minutes . The second filter is set at players that provide at least one key pass per 90 minutes, in order to end up with players in creative roles. The green bars represent players with an ExpA over the 95% confidence interval. In other words, these players provide significantly more ExpA than the others
It’s interesting to see some players with low assist numbers high on the ExpA list. Roda’s Mark-Jan Fledderus is 4th on our list, with around 0.4 ExpA per 90 minutes, despite having just two assists. ADO’s winger Jerson Cabral should have around 0.35 assists per full match, but he doesn’t even have a single assist. Convertional measure would have criticized these players, while the ExpA method shows that the issue lies not at their feet.
The reverse is also true for some players on the list. Take AZ’s main creative outlet Maarten Martens. He’s got 9 assists already this season, despite producing ‘just’ 0.28 ExpA per 90 minutes, which should lead to 3 or 4 assists if we assume average finishing. I’m writing ‘just’ because 0.28 is still quite good actually. However, it looks like Martens is at the positive end of some statistical variation with his 9 assists this half season.
The good, the bad and the ugly
If we say ExpA is the good, then wasted passes are the bad and ugly. The ideal player creates loads of ExpA, and scores very low on incomplete passes. Looking at ExpA in isolation may provide the wrong picture, as we would want our creative talent to create a maximum of threat with a minimum of costs. If ExpA is the threat, incomplete passes are the costs.
Here’s the both of them in one plot, where I would encourage you to click on it for the full size version. The horizontal axis has each player’s incomplete passes per 90 mins, while the vertical axis has ExpA per 90 minutes. The dotted black and red lines represent the 95% and 99.9% confidence interval. Players beyond these margins have a significantly different pattern from the others in the graph, and these are the only players labeled by their name, to keep the chart accessible. Where would the best players be?
The very best players should be in the top left corner of this chart: a maximum of ExpA and no incomplete passes. However, from the slope of the black line you can see that there is a positive relation between ExpA and incomplete passes. Which makes complete sense, offensive players have a more difficult job completing passes, and they are also the players in the position to score high in terms of ExpA.
The vertical axis is identical to the bar graph earlier, with Dusan Tadic a class above the rest. The trailing group is spread out horizontally, with Heerenveen’s playmaker Hakim Ziyech being the most wasteful, and Vitesse’s Chelsea loanee Piazón and Feyenoord winger Boëtius the most tidy names. N.E.C. striker Soren Rieks is also in a nice spot, being more tidy than the average player, but still creating an elite ExpA in a poor team.
One Ziyech, or two?
This is not to say it’s not good to have Ziyech in your team. Actually it is quite good to have one of the best creative talents of the Eredivisie, but with his amount of incomplete passes, the rest of the team should play a very tidy game, to prevent the tactic from tipping over. One Ziyech wasting some 12 to 13 passes per match is okay, but put two Ziyechs in your midfield and your team loses 25 possessions there already.
Teams could use concepts like this to balance their first eleven out. It’s a risk and reward analysis, where having too many Ziyechs may prove very costly, while having too little of them won’t work either. It’s this kind of analysis that may help to find the ideal balance, without surrendering yourselves to the dangers of random variation. It may also help in selecting the optimal first eleven, as it is not always advisable to start the 11 best individual players outright, without looking at the mix of qualities needed.
Obviously, this concept could be used as a scouting tool too. I suppose that if you successfully trade players from below the thick black line for players above it, you add more creative talent to your team. And if your supposedly creative midfielders are found below the black line, this indicates that they create too little ExpA for the amount of incomplete passes they use.
Measuring creative output in ExpA helps to quantify what type of players you have, at what costs in terms of incomplete passes they operate and what alternative names you could be looking for. Obviously, in composing the midfield, there’s more than just ExpA and incomplete passes, but we’ll get to the other midfield positions later. For the creative part, it’s ExpA versus incomplete passes!