How I scout for players
I've seen a few posts recently asking about the scouting process, the "how do you judge a player" part is interesting. If it is easy then everyone would get every transfer right. If it is pure fortune then some clubs wouldn't be consistently better at it than others. So it must be a skill of some sort either in having an "eye" or of knowing which data to rely on to make a good judgment.
Like many people in this corner of the internet, I have always enjoyed being able to lay claim to have spotted a talented player before everyone else. My earliest memory of this was sticking an "FAO Joe Royle" note in with my Evertonian magazine subscription cheque when I was about 14. It advised him to look at Patrick Vieira (then captaining Nice) and Dynamo Kiev's Andrei Shevchenko (and Marlon Beresford but I tend to leave that one out!) as I'd read about them in World Soccer.
There are three types of player scouting; top 5 league regulars, players with data and players without data. Data helps immensely. Wyscout, Opta, Statsbomb, and others, provide data we could only have dreamed about a few years ago. Watching players with your own eyes is fine, but has all those problems everyone who has read Moneyball will tell you about. Even if you haven't asked.
If you are looking at a player in a top 5 league you can be pretty sure they have already been through a fairly thorough scouting process to get to the point they are at. They are already playing at a high level against highly skilled opponents. With a few allowances for systems, positions, and style you can be reasonably confident that a player dominating the central midfield area in any of those league has at least a very strong chance of carrying that form across leagues if they move to a team of similar relative dominance. I wrote about this for Statsbomb here and it seems to hold true.
For me, that part of relative dominance is important. If you are scouting a player on a bad team who is still producing good data you can expect that they will have better output on a more dominant team. The reverse is true though, if you are signing a player from a dominant team and drop them into a relegation battling team you would expect their stats (relative to league average as most are) would drop off.
So if we can assess top 5 league players reasonably efficiently through data, and the wisdom of the crowds meaning that the player has been assessed all through their career to date as having a top 5 league skill level what about other leagues?
The classic example is the Dutch league. A league that has both produced some of the biggest transfer successes and worst transfer flops in recent years. How can you tell a Van Nistelrooy from a Vincent Janssen without just being wise after the event? On the data side you can look at their performances just against the "big clubs", it is no good scoring 3 every time you play Go Ahead Eagles if every time you play PSV you don't get a kick. Assuming they pass this test it moves on to the eye assessment.
The eye test always needs to be involved, nobody would sensibly suggest otherwise, and I've never heard anyone involved in analytics say it.
I like to keep three things in mind:
Look at the processes not just the outcomes
Scout on attributes - Fellaini
Remember your own biases - you want certain players to be good, first impressions matter
Processes not outcomes
Data is great but there are a number of gaps still to be filled. This is where scouting comes into it. Let us imagine there are 3 players in the Dutch league with similar basic statistics, they type of thing you can pick up for free on Whoscored.
The players are all strikers, all have 0.5xg per 90.
How they generate that 0.5xg is likely to be the key to assessing how high their potential is.
Player A may be a 6ft 5 target man who is fed a large number of crosses per match. He physically outpowers his opponents and scores the majority of goals from headers.
Player B is a shorter player with blistering pace. All his goals have come from very accurately played through balls from deep against very high, very slow defensive lines.
Player C has scored a good mix of different types of goals, he has pace but has also scored headers, volleys with either foot and strikes the ball powerfully and accurately towards the corners.
I've made it very obvious that Player C would be the player with the highest potential but even then it doesn't rule out Players A and B from being good as long as there are plans to use them in a certain way. If you want an aerially dominant impact sub or pace to play against tiring defenses then Players A or B might be affordable options.
Which leads me on to....
Scout on attributes
Who is a better central midfielder; Gerrard, Lampard, Scholes, Fabregas, Vieria, Keane?
You, being a clever person, refuse to enter the debate. They are all different types of player. It depends entirely on who else you have in midfield with them and what you want their role to be. Don't you remember the Gerrard and Lampard combination for England*?
*although being a clever football person you think the problem was overstated in the media and the evidence shows that the problem was more to do with blah blah blah
Different players have different skills and play different roles.
Fellaini being exceptional at retaining balls in high areas of the pitch was a huge advantage, his unorthodox style was the reason Moyes, Van Gaal, Mourinho, and all the Belgium managers all played him.
Therefore on your checklist of "things you want from a central midfielder" you always need a space for other attributes they may have which may allow them to adapt to other roles.
Remember your own biases
Everyone wants to discover that wonder kid. When you see some random 18 year old from the Bulgarian league scoring a 30-yard volley the temptation is to label them the new Stoichkov. However, any player at any level can score the goal of their life at any time. Anyone who witnessed my scorpion kick volley as a one-off in around 1996 might have thought I was some wonder kid. The fact the subsequent attempts to recreate it lead to a mild concussion, as I kept volleying the ball into the back of my own head, would have persuaded them otherwise.
We also have playing styles we are more naturally drawn to. I know I tend to overrate midfielders who drive through the centre of the pitch with nice line breaking runs, and underrate the metronomic players who keep things ticking over through boringly effective use of the ball. These players are dull to scout but often end up forming midfielders because effective is generally what people want over 90 minutes.
So how to do it
First of all, I have to acknowledge I am not an expert, I have no track record to boast of, other than numerous articles showing players I've picked out. I am only now a few months into a proper scouting/data project and am using a system developed by others that I am not going to reveal the secrets of!
All I would say is that you need to consider what you think the important attributes of a player are for the role they are in. So defenders need to be able to defend. But what does that mean? Come up with your own definition, test it with data and see if it looks right. If Van Dijk is near the top it is probably a good starting place but then look how it ranked him at Southampton and Celtic (if the data exists). Is there a pattern forming?
Then to the eye test. You need to watch a lot of football. What can you not see with basic data?
Perhaps for a centre back you might want to see how often they are beaten by pace/movement. This is where you need to both use your judgment and have "calibrated" your eyes to see what is good or not. It may only be a small percentage of difference between a top player and an average. Look at how they use the ball, are they good at coping with all different types of opponent? A large enough sample will take a lot of viewing. Remember strikers A, B and C from earlier, you want to see all of those against the defender. Look at the 7-0 losses as well as the 3-0 wins. For the amateur scout, this isn't worth it, you aren't going to spend 8 hours on a tweet, but in the professional game clubs will want to build up a lot of knowledge before committing to £30m in wages/fees.
For players with no data, such as youth players you really need to have both calibrated your eyes to what is good and be aware of the development changes in players. That 6ft bearded 17 year old with 50 goals might be nowhere near a good as the 5ft 2 player on for the last 10 minutes but who can spot a pass blindfolded. This is the hardest type of scouting, anyone can spot a Rooney, few can say for sure about a Lampard type who kept improving and improving up until about 25.
Proper due diligence is vital with huge money transfers. On and off pitch scouting work is really important when you consider £50k a week 5-year contracts.
Nobody would use data alone, nobody would ignore data altogether, if you want to go into that world think about improving on the offering available from the data companies with your own eyes. My DM's are always open if you have anything interesting you are working on and want any input.
Like many people in this corner of the internet, I have always enjoyed being able to lay claim to have spotted a talented player before everyone else. My earliest memory of this was sticking an "FAO Joe Royle" note in with my Evertonian magazine subscription cheque when I was about 14. It advised him to look at Patrick Vieira (then captaining Nice) and Dynamo Kiev's Andrei Shevchenko (and Marlon Beresford but I tend to leave that one out!) as I'd read about them in World Soccer.
 |
| Dear Joe, I know you can't have got my last letter as you tried to sign Tore Andre Flo and got sacked. |
There are three types of player scouting; top 5 league regulars, players with data and players without data. Data helps immensely. Wyscout, Opta, Statsbomb, and others, provide data we could only have dreamed about a few years ago. Watching players with your own eyes is fine, but has all those problems everyone who has read Moneyball will tell you about. Even if you haven't asked.
If you are looking at a player in a top 5 league you can be pretty sure they have already been through a fairly thorough scouting process to get to the point they are at. They are already playing at a high level against highly skilled opponents. With a few allowances for systems, positions, and style you can be reasonably confident that a player dominating the central midfield area in any of those league has at least a very strong chance of carrying that form across leagues if they move to a team of similar relative dominance. I wrote about this for Statsbomb here and it seems to hold true.
For me, that part of relative dominance is important. If you are scouting a player on a bad team who is still producing good data you can expect that they will have better output on a more dominant team. The reverse is true though, if you are signing a player from a dominant team and drop them into a relegation battling team you would expect their stats (relative to league average as most are) would drop off.
So if we can assess top 5 league players reasonably efficiently through data, and the wisdom of the crowds meaning that the player has been assessed all through their career to date as having a top 5 league skill level what about other leagues?
The classic example is the Dutch league. A league that has both produced some of the biggest transfer successes and worst transfer flops in recent years. How can you tell a Van Nistelrooy from a Vincent Janssen without just being wise after the event? On the data side you can look at their performances just against the "big clubs", it is no good scoring 3 every time you play Go Ahead Eagles if every time you play PSV you don't get a kick. Assuming they pass this test it moves on to the eye assessment.
The eye test always needs to be involved, nobody would sensibly suggest otherwise, and I've never heard anyone involved in analytics say it.
I like to keep three things in mind:
Look at the processes not just the outcomes
Scout on attributes - Fellaini
Remember your own biases - you want certain players to be good, first impressions matter
Processes not outcomes
Data is great but there are a number of gaps still to be filled. This is where scouting comes into it. Let us imagine there are 3 players in the Dutch league with similar basic statistics, they type of thing you can pick up for free on Whoscored.
The players are all strikers, all have 0.5xg per 90.
How they generate that 0.5xg is likely to be the key to assessing how high their potential is.
Player A may be a 6ft 5 target man who is fed a large number of crosses per match. He physically outpowers his opponents and scores the majority of goals from headers.
Player B is a shorter player with blistering pace. All his goals have come from very accurately played through balls from deep against very high, very slow defensive lines.
Player C has scored a good mix of different types of goals, he has pace but has also scored headers, volleys with either foot and strikes the ball powerfully and accurately towards the corners.
I've made it very obvious that Player C would be the player with the highest potential but even then it doesn't rule out Players A and B from being good as long as there are plans to use them in a certain way. If you want an aerially dominant impact sub or pace to play against tiring defenses then Players A or B might be affordable options.
Which leads me on to....
Scout on attributes
Who is a better central midfielder; Gerrard, Lampard, Scholes, Fabregas, Vieria, Keane?
You, being a clever person, refuse to enter the debate. They are all different types of player. It depends entirely on who else you have in midfield with them and what you want their role to be. Don't you remember the Gerrard and Lampard combination for England*?
*although being a clever football person you think the problem was overstated in the media and the evidence shows that the problem was more to do with blah blah blah
Different players have different skills and play different roles.
Fellaini being exceptional at retaining balls in high areas of the pitch was a huge advantage, his unorthodox style was the reason Moyes, Van Gaal, Mourinho, and all the Belgium managers all played him.
Therefore on your checklist of "things you want from a central midfielder" you always need a space for other attributes they may have which may allow them to adapt to other roles.
Remember your own biases
Everyone wants to discover that wonder kid. When you see some random 18 year old from the Bulgarian league scoring a 30-yard volley the temptation is to label them the new Stoichkov. However, any player at any level can score the goal of their life at any time. Anyone who witnessed my scorpion kick volley as a one-off in around 1996 might have thought I was some wonder kid. The fact the subsequent attempts to recreate it lead to a mild concussion, as I kept volleying the ball into the back of my own head, would have persuaded them otherwise.
We also have playing styles we are more naturally drawn to. I know I tend to overrate midfielders who drive through the centre of the pitch with nice line breaking runs, and underrate the metronomic players who keep things ticking over through boringly effective use of the ball. These players are dull to scout but often end up forming midfielders because effective is generally what people want over 90 minutes.
So how to do it
First of all, I have to acknowledge I am not an expert, I have no track record to boast of, other than numerous articles showing players I've picked out. I am only now a few months into a proper scouting/data project and am using a system developed by others that I am not going to reveal the secrets of!
All I would say is that you need to consider what you think the important attributes of a player are for the role they are in. So defenders need to be able to defend. But what does that mean? Come up with your own definition, test it with data and see if it looks right. If Van Dijk is near the top it is probably a good starting place but then look how it ranked him at Southampton and Celtic (if the data exists). Is there a pattern forming?
Then to the eye test. You need to watch a lot of football. What can you not see with basic data?
Perhaps for a centre back you might want to see how often they are beaten by pace/movement. This is where you need to both use your judgment and have "calibrated" your eyes to see what is good or not. It may only be a small percentage of difference between a top player and an average. Look at how they use the ball, are they good at coping with all different types of opponent? A large enough sample will take a lot of viewing. Remember strikers A, B and C from earlier, you want to see all of those against the defender. Look at the 7-0 losses as well as the 3-0 wins. For the amateur scout, this isn't worth it, you aren't going to spend 8 hours on a tweet, but in the professional game clubs will want to build up a lot of knowledge before committing to £30m in wages/fees.
For players with no data, such as youth players you really need to have both calibrated your eyes to what is good and be aware of the development changes in players. That 6ft bearded 17 year old with 50 goals might be nowhere near a good as the 5ft 2 player on for the last 10 minutes but who can spot a pass blindfolded. This is the hardest type of scouting, anyone can spot a Rooney, few can say for sure about a Lampard type who kept improving and improving up until about 25.
Proper due diligence is vital with huge money transfers. On and off pitch scouting work is really important when you consider £50k a week 5-year contracts.
Nobody would use data alone, nobody would ignore data altogether, if you want to go into that world think about improving on the offering available from the data companies with your own eyes. My DM's are always open if you have anything interesting you are working on and want any input.
Comments
Post a Comment