Author: admin

Loyalty, recruitment, brain types and the ABC-model

Player agents often complain about the lack of loyalty of the players they have signed. They assume that loyalty is an inherent trait some players have and others don’t. Of course, it is painful to see one of your biggest talents sign with a different agency just before their big breakthrough. In most cases leaving the agency has little to do with loyalty and more to do with the player’s brain type and the ABC-model. In this article I will describe what an agent can do to breed loyalty into his players.

First of all, the whole idea of people having traits is a backward idea. In reality people acquire knowledge through associative learning and skills through instrumental learning. In terms of football: associative learning gives you game intelligence and instrumental learning gives you technique. How do we know whether a player has game intelligence or technique? We see that in how the player behaves. For we cannot look into the soul of the player.

The behavioral patterns of a player are, for the most part, acquired through instrumental learning. Through instrumental learning the brain creates probabilistic relationships between the behavior and what this behavior gets you. The brain of the star player has learned in extreme detail how to shoot the ball in order to get the result the player wants: a goal. Instrumental learning works according to the ABC-model. In this model A stands for Antecedent which is everything that happens before the behavior or is necessary to make the behavior possible. B stands for Behavior, the desired or undesired behavior you are targeting. C stands for Consequence which is everything that happens after the behavior. There is overwhelming evidence that Consequences have a much, much bigger influence on our future behavior than Antecedents. Nevertheless, in most cases we continue to try to influence people through Antecedents rather than through Consequences.

‘So when it comes to loyalty, there isn’t an inborn trait that some players have and others don’t. Instead, there is the history of all the Consequences that the agent has given in response to the behavior of the player. To understand this you first have to specify the desired behavior. To do this we have to take MARCO into account. Behavior is only behavior if it is:

  • Measurable. If you can’t measure it, it ain’t behavior.
  • Active. If a dead person can do it, it ain’t behavior.
  • Reliable. If you can’t measure it reliably, i.e. different people come up with completely different measurements, it ain’t behavior.
  • Controlled. If it is not under the control of the actor, it ain’t behavior.
  • Observed. If it is impossible to observe, it ain’t behavior.

As you can see: loyalty ain’t behavior. Loyalty can’t be measured, a dead person can be loyal, if you can’t measure it, you can’t measure it reliably, loyalty is not under the control of the actor and you can’t observe it directly. So we have to specify the behavior that makes us think that a player is loyal. Most agents would specify that behavior as the player not signing up with another agent. But here, the first mistake is made. Dead persons never sign contracts with other agents. So not signing a contract, ain’t behavior either. Instead the right specification is to honor the contract the player signed. Dead persons can’t honor contracts. You can measure how long the player is honoring the contract and you can measure this reliably. Honoring the contract is completely under the control of the player. And we can easily observe the player honoring the contract.

So the desired behavior is honoring the contract and the undesired behavior is signing with another agent. The ABC-model teaches us that players do more of what has been rewarded with positive consequences in the past; in the same way players do less of what has been punished or penalized in the past. A player breaking his contract and signing with another agency, doesn’t do so out of disloyalty, but because honoring the contract has not given him enough positive consequences. On top of that honoring the contract with the agency always has at least one negative consequence. For the fee that the player pays the agent, is experienced in the brain as a penalty. Players want money, so spending money is a negative consequence. It is the task of the agent to compensate for this negative consequence, by more positive consequences. At first the agent does this by promising the player more positive consequences. But these promises are Antecedents and have little impact on the future behavior of the player.

Only when the player really does get what he wants as a result of him honoring the contract, only then the player gets a positive consequence. So the ability to get the player signed with a big club for a high salary, is the most important job of the agent. Yet, this happens only every few years. That means that after the first signing the agent made possible, it will take a long time before the next big positive consequence will be there to reinforce the player’s brain to honor the contract. Furthermore, this future positive consequence is also uncertain. The player might get an injury that ends his career. Or it may turn out that he is less talented than thought before. Or just a case of bad luck. Research clearly shows that long term uncertain positive consequences have way less impact on the behavior of a player than short term certain consequences. Therefore the agent has to make sure that the player is rewarded short term with a high degree of certainty for honoring his contract. If an agent does this then the player will continue to honor his contract and everybody will think that he is a very loyal player. Whereas in fact it is the behavior of the agent rather than the player that makes the player appear to be loyal.

What kind of short term positive consequences are there for the agent to give to the player? In short the agent can choose between the following categories:

  1. Material rewards:
    1. Direct material rewards: food or drinks.
    2. Indirect material rewards: money or valuables.
  2. Social rewards:
    1. Attention. It is important that the agent regularly checks in with the player to ask how he is doing.
    2. Compliments. If the player achieves something on the pitch during a match, make sure he is complimented for it as soon as possible after the match.
    3. Status. An agent can create different classes of players within the agency so a player feels he is promoted within the agency as he develops. Just make sure that you set-up the program in such a way that there are only winners.
    4. Information. Many players love to have access to the statistics of how they did or video’s of their best actions.
    5. Opportunities to develop one self. Players not only want to become better at football, they also want to develop themselves mentally.
    6. Keep their social media up to date. Keeping their social media up to date has negative consequences for players as it takes time and energy. So often they love it if the agent takes care of it. Updating their social media accounts as soon as possible so the player sees his fans rewarded as soon as he comes off the pitch, is a positive consequence for most players. Also because this enhances their status.

As players most of the time get plenty of material rewards, the best choice for agents is to go for social rewards. The easiest way to discover what kind of rewards the player is looking for is by asking the player himself. This may seem obvious, yet it is the second mistake most people who use the ABC-model make. They fall into the pit called the Perception Error and assume they know what is a positive reward for the player. So ask your players, how they can be rewarded on top of everything they already get from the club. 

Brain types

The third mistake is disregarding brain types. In the same way that there are different body types, we also have different brain types. Your brain type determines your evolutionary behavioral patterns. These behavioral patterns determine:

  1. How you are motivated.
  2. How you deal with your emotions.
  3. How you learn.

Brain types determine in a large part how the Dopamine reward system in your brain works. Therefore, if you know someone’s brain type you can predict with a high probability how you can reward him with positive consequences. Here is the list of positive consequences for each brain type:

Type #1, the Perfectionist can be rewarded with control.

Type #2, the Helper can be rewarded with love and attention.

Type #3, the Successful Worker can be rewarded with material rewards and hopeless projects where he has a small chance of becoming the project’s hero.

Type #4, the Romantic can be rewarded with justice served.

Type #5, the Analyst can be rewarded with autonomy, personal freedom and being left alone.

Type #6, the Loyalist can be rewarded with safety.

Type #7, the Hedonist can be rewarded with new things to do and variation.

Type #8, the Boss can be rewarded with power.

Type #9, the Mediator can be rewarded with harmony.

As loyalty also is an evolutionary behavioral pattern, some brain types have special issues concerning loyalty as can be seen from this list:

Type #1, the Perfectionist has no special issues with loyalty. Yet, as Perfectionists feel that they must act in accord to the morals of the group, it helps if you make honoring your contract one of high principles endorsed by the whole group.

Type #2, the Helper has no special issues with loyalty. Yet their craving for love and attention is so high that if the agent fails to make compliments, give little presents and keep in touch with them, the agent risks being put in the so-called out group and that will lead to a parting of the ways.

Type #3, the Successful Worker has an issue with loyalty. Successful Workers are very loyal when they are relaxed. If they are stressed, they seek social stability. In both cases it is unlikely that they would break their contract. Unfortunately, when neither stressed, nor relaxed, they become reckless, antisocial and highly sensitive to material rewards and promises of material rewards. In that state, they can be easily poached by other agents.

Type #4, the Romantic has no issues with loyalty. In fact, if breaking the contract is seen as an injustice it is unlikely that the Romantic will break the contract. On the other hand, if the agent’s actions appear to be unjust toward the player, other players, clubs or people in general, they might very well break their contract even if it means a worse outcome for themselves.

Type #5, the Analyst has no issues with loyalty. If it is clear for the Analyst that he has lots of autonomy, personal freedom and is left alone, he will not risk losing this by signing with another agent.

Type #6, the Loyalist has issues with loyalty as the name implies. It will take quite some time and thorough research by the Loyalist before the Loyalist signs with an agent. Nevertheless, once they sign, they honor their contract. Not so much out of loyalty, but because they see too much risk in breaking the contract. Unfortunately, Loyalists are probably underrepresented in football as the game and the culture are not their thing.

Type #7, the Hedonist has no issues with loyalty. The one thing to watch out for is that if the Hedonist stresses he becomes quite sensitive to material rewards. Furthermore, there is the risk that he lacks a clear sense of morality. Meaning that if stressed, the Hedonist can easily be bribed, even illegally, to sign with another agent.

Type #8, the Boss has no issues with loyalty. In fact, the Boss likes to receive a clear manual from a higher power he respects. He then blindly follows the rules in the manual and will in fact enforce these rules with other players. So if there is a rule in the manual that states that you always will honor your contract he will do so and he will try to forcefully make other players within the agency comply with that rule as well.

Type #9, the Mediator has no issues with loyalty. In fact, the Mediator is likely to become dependent on the agent and would find it emotionally difficult to leave the agency. As most players have a type #9 brain, this is the common experience of agents. They mistake these dependency issues for loyalty and then complain that the other players lack these issues. 

So besides using the ABC-model to positively reinforce honoring the contract, it is also smart to take into consideration the brain type of each player.

Davy Klaassen to Ajax

Based on the Wyscout data for the 43 matches Klaassen played for Werder Bremen in season 19/20 & 20/21, he has:

  • 86% probability that he is able to contribute to the Werder Bremen overall,
  • 5% probability that he is able to contribute to the attack of Werder Bremen,
  • 97% probability that he is able to contribute to the defense of Werder Bremen,
  • 83% probability that he is able to contribute to the build up and transitioning of Werder Bremen.

Based on the Wyscout data for the 19/20 season Werder Bremen as a team had:

  • 31% probability that the team will win or draw a match,
  • 30% probability that the attack will score,
  • 43% probability that the defense will concede a goal (lower is better),
  • 44% probability that the build up and transitioning will create an opportunity.

Based on the Wyscout data for the 19/20 season the Bundesliga has a FBM League Strength score of 123 points. (91% correlation)

Based on the Wyscout data for the 19/20 season the Eredivisie has a FBM League Strength score of 114 points. (91% correlation)

Based on the Wyscout data for the 19/20 season Ajax has:

  • 87% probability that the team will win or draw a match,
  • 69% probability that the attack will score,
  • 28% probability that the defense will concede a goal (lower is better),
  • 53% probability that the build up and transitioning will create an opportunity.

Given that the League Strength of the Eredivisie is lower and that the club probabilities of Ajax are higher, it is a realistic idea to see Klaassen play for Ajax.

Based on the above data including minutes played, difference in league strength and difference in team strength, we calculate the following probabilities for Klaassen playing for Ajax.

  • 94% probability that he is able to contribute to Ajax overall,
  • 12% probability that he is able to contribute to the attack of Ajax,
  • 99% probability that he is able to contribute to the defense of Ajax,
  • 92% probability that he is able to contribute to the build up and transitioning of Ajax.

As you can see the performance of Klaassen will be very similar for Ajax as it was in the 19/20 season for Werder Bremen.

If we were to substitute Van de Beek for Klaassen Ajax would get the following probabilities:

  • 92% probability that the team will win a match (+6%),
  • 70% probability that the attack will score (+1%),
  • 26% probability that the defense will concede a goal (lower is better) (-2%),
  • 53% probability that the build up and transitioning will create an opportunity (+0%).

This would result in 5 additional points in the table.

To conclude: Ajax is slightly better off with Klaassen.

Shadow team born this century anonymized

To track how we are doing in finding talent at a relative early age (20 years or younger), we publish our shadow team anonymized and keep track of how these players are doing. As soon as any of these players transfer to another club or become a household name, we update this list and reveal the name. Or if they turn 25 in case the did not break through. Valuation at date is the price where would virtually buy the player. That way we can see how much profit would make virtually.

We paper trade the players as if we bought them for the valuation at data. The we sell the players when they reach the age of 25 or when the make a major transfer. That way you can see how well we do.

So far we have spent 65.875.000 euro and earned 56.000.000 euro for a loss of –9.875.000 euro. From 2021 on,players need to be born 2001 or later.

Of the 117 players on the list 77 has increased in valuation (73%), 6 have decreased in valuation (4%) and 34 have no change in valuation mainly due to still being too young. (23%)

PlayedIDDate first in shadow teamValuation at dateTransfer/Currently valuedPositionFBM scoreBorn/playerContract/SoldSold
64612-4-2018126 million transfer to RennesRW8.66Jérémy Doku26M26
804422-4-20190.53CM6.8420022022
87398-6-20190.20.5CB6.7720022023
1261415-7-202000.7CB6.1620022023
1125817-9-20190.050.1CB6.052002unknown
1228918-5-202011.2CF5.9920022021
1112127-1-202000.25CM6.1320012021
1241422-6-20200.81RW5.620012021
1206325-1-20200.10.25CM7.7620002021
810727-1-20200.750.7CF7.2920002021
109571-8-20190.40.3AM7.1720002022
115137-3-20201.11.3RB7.0620002022
110465-3-20200.40.6DM6.8820002022
1165720-12-20190.10.25CB6.4720002023
74062-3-20200.70.635GK6.2420002021
823021-10-201934LB5.9720002024
1173230-10-20192.7530 million transferAM7.22Odilon Kossounou30M30
118414-12-201905LB7.4220012024
122305-6-20203.53.6CF8.220022023
126652-7-20200.43DM7.520012024
1272121-7-20200.10.1LW6.52000unknown
1274922-7-20200.40.5RW62001unknown
1279023-7-20202.54LW5.720012022
1279325-7-20200.30.35CF720012023
1279425-7-20200.50.5CF6.520022023
1279525-7-20200.150.5CB620002022
1279625-7-20200.81CB620012022
1234326-7-20202.32.3AM620012025
1280426-7-20200.10.2CB5.820032022
1281026-7-20200.050.4RW5.720042023
1009328-5-20192.54.5CB7.420002024
128551-8-20200.20.2CF5.920022022
128562-8-202002CM5.620032021
128695-8-202001AM720002022
128745-8-20200.20.6LW6.320012023
128899-8-20200.92CB620002024
1290210-8-20200.10.1LW6.62003unknown
1290310-8-20200.150.1LW7.52003unknown
1291011-8-20200.250.5CM8.220012022
1292712-8-20200.20.8CF5.920042022
1292212-8-20200.151.5CF7.120012021
1292913-8-20200.50.8RW620012023
1034012-12-201800.5RW820002021
1035219-8-202004.5LB6.2520012024
1297519-8-202003CM5.520012021
1298322-8-20200.73CM6.1220022024
1302526-8-202000.1AM6.320052023
121033-3-20200.0252.2CB7.3120002022
814517-8-20190.50.55RB6.8520012022
73835-9-20200.50.575LW7.8220002021
130755-9-20200.30.4DM7.1320002021
563929-5-20190.20.3AM5.9520002022
130968-9-20200.20.4CB5.6820002023
81303-5-20190.10.7DM7.8920002022
804619-4-20190.250.6CM7.3920002022
1209412-9-20200.050.6CM7.1120002022
1316215-9-20200.42CB7.5620002023
1315915-9-20200.0511CB6.4320022025
132816-10-20200.050.075CB6.2220002022
1329914-10-20201.25DM7.520012024
1330018-10-202032.5CB5.720032023
1330122-10-20200.151.2CB5.8620002023
1334428-10-202034.5LW5.9620012023
133563-11-20200.10.8CM7.520022025
133573-11-20200.10.4CF720022025
134718-12-202013CF5.520022022
134729-12-2020011CF6.620012023
1349110-12-20200.30.3CM620012022
1344210-12-20200.20.4CM820002023
986310-12-20200.30.7RB7.620002024
985010-12-20200.10.4RW6.820002022
1312613-12-20200.51.2LB620012023
1127024-12-202023CB6.320022021
135493-1-20211.81.8AM620022024
135536-1-20210.7250.8CF620012023
135346-1-202100CM62004Unknown
135356-1-202100CM820052022
135366-1-202100AM82005Unknown
135376-1-202100CF720052023
135396-1-202100CF62004Unknown
135606-1-202100CF62004Unknown
135616-1-202100CF720042022
135626-1-202100CF720042022
135636-1-202100AM72005Unknown
1365211-1-202100GK6.220022022
1367714-2-202100RW7.520022022
1367914-2-202100LW6.92002Unknown
1368417-2-202100CF6.32002Unknown
1369419-2-202100CF6.420042021
1369519-2-202100AM6.720042022
1369619-2-202100RW72004Unknown
1399719-2-202100CF62004Unknown
1399819-2-202100CF5.720042021
1369919-2-202100LW5.820042022
1370019-2-202100AM6.92004Unknown
1370119-2-202100AM6.62004Unknown
1370220-2-202100CF6.820032021
1370324-2-202100.3RB7.920022026
1370625-2-20210.50.5GK620012024
1370827-2-202111CF5.720032023
137224-3-2021010CM620022025
1373415-3-20210.83RW720022023
1373516-3-20210.70.8CF620012024
1373616-3-202100CM7.520032021
1374822-3-202122DM6.520022022
1376924-3-202133CB5.920022021
1377125-3-202136CB7.120012024
1380630-3-20210.80.8CB72002Unknown
1381130-3-20210.10.2CB6.820012021
1383331-3-20210.50.5CM720032022
125256-4-202111.1CF620022025
138448-4-202101CM6.720022024
138428-4-202100.3LB5.820022024
1387016-5-20213.54CM6.320032024
1387117-5-20210.0750.4CB8.520012021
1387217-5-2021214RW720032022
1425913-7-202100CM72006Unknown
144543-8-20210.10.1RW720022025
144583-8-20210.10.1LW720022025

Current value is public valuation of the player by TransferMarkt in million euro. FBM score is our propietary score to rank players. Only players who score 5.5 or higher make it to the list. PlayerID is the ID of the player in our FBM database.

Relative strength MLS

Here is the list of the relative strength of teams in the MLS based on their performance in the previous season according to Wyscout data. If it were a simple competition than this would also be our prediction for the league table. Previously predictions like these have had a 80% correlation with the actual league table a year later. So will see how it goes for the MLS:

RankFBM Wyscout score
1Atlanta73
2Los Angeles72
3Salt Lake71
4Toronto70
5Dallas67
6New York City65
7Columbus63
8Seattle62
9Portland61
10Chigago57
11Minnesota56
12San Jose56
13Colorado54
14New England54
15LA Galaxy52
16Montreal51
17Philadelphia50
18Kansas City48
19DC45
20Orlando42
21New York RB40
22Houston37
23Vancouver36
24Nashville35
25Cincinnati30
26Miami19

Showcase: Niklas Dorsch

Niklas Dorsch is a defensive midfielder of Heidenheim, playing in 2. Bundesliga. Dorsch has been on our radar for the last couple of years and with only one year left at Heidenheim, he is an interesting player to follow. He plays for Germany U21. We think he would do well in 1. Bundesliga. For that reason we show here that he would be a good addition to Eintracht Frankfurt.

Dorsch at Heidenheim

Here is Dorsch’ most recent FBM contribution chart:

Although Heidenheim lost and Dorsch did not play his best match, especially in the first half as can be seen from his contribution chart, he is still an exceptional player according to his FBM stats:

Yet, these are his stats for playing in the 2. Bundesliga. How would he do at Eintracht Frankfurt? We think that Dorsch is a good replacement for Hasebe at Eintracht Frankfurt. Hasebe’s most recent contribution chart shows he is not playing well at the moment:

Also his FBM stats are less than those of Dorsch:

Dorsch is slightly better than Hasebe at their highest performance, but Dorsch beats Hasebe on average performance, current performance and worst performance. Yet, Hasebe plays on a higher level. So we have to take that into account.

Dorsch playing for Eintracht Frankfurt

Taking into account minutes played, difference between both clubs and both competitions, we get the following results for Dorsch playing at Eintracht Frankfurt:

What you see in the first row, is the performance level of Eintracht Frankfurt in the 1. Bundesliga. In the second row we subtract Hasebe’s contribution to the performance of Eintracht Frankfurt. That is only a small difference because Hasebe is not contributing that much on average. In the third row we add Dorsch to the expected performance of Eintracht Frankfurt. Finally, we can see how Eintracht Frankfurt’s performance would increase or decrease in row 4. Overall performance of Eintracht Frankfurt would rise as would attack and transitioning. Defensive performance would suffer slightly.

Eintracht Frankfurt’s FBM Team Score would increase to 115 points up from 102 points. There is an 80% correlation between FBM Team Score and future ranking in the league. If other teams would not improve Eintracht Frankfurt would rise to rank 10 in the league table if they played with Dorsch rather than Hasebe.

What is Dorsch worth?

Our model takes into account, position, highest transfer fee in the current season, record transfer fee, difference in competition, club, player age and length, international status and FBM stats. Due to the Corona crisis, it is much more uncertain how future transfer fees will develop. Our model is still based on the pre-Corona circumstances. 

When we calculate what we call the replacement fee for players. This is the amount of money the current employer of the player can expect to spend on a replacement who is as good as their current player. In short: clubs should not transfer players for less than the replacement fee, nor buy players for more than the replacement fee. As the replacement fee differs from club to club, there is room for negotiations. We also calculate what the player would be worth one year later if he is able to transfer to an even better club. All assuming his FBM stats remain the same.

Here are the replacement values for Dorsch:

Replacement value for Heidenheim£2,592,419
Replacement value for Eintracht Frankfurt£3,861,435
Replacement value for Schalke 04£6,672,353

TransferMarkt currently values Dorsch at £4.05m. We think that Dorsch is slightly overvalued on TransferMarkt. If Eintracht Frankfurt were able to buy Dorsch for less than £3,861,435, they would have a good deal. A deal that might make them almost £2m a year later, if they would transfer Dorsch to the next club and Dorsch would perform at the level we expect him to do. As a reminder, we predicted that Dalmau would be worth 1.75m euro to Heracles and they transferred him a year later for 1.7m euro.

For us the most important thing about FBM stats, is that we calculate the probabilities that a player is able to contribute to a specific team. Here are the probabilities that Dorsch is able to contribute to Eintracht Frankfurt:

Probability that Dorsch contributes to Frankfurt63
Probability that Dorsch contributes to the attack of Frankfurt72
Probability that Dorsch contributes to the defense of Frankfurt25
Probability that Dorsch contributes to the transitioning & build up of Frankfurt46

Case study: Watford

Many scouts wonder why their advice is being ignored by the higher ups. The reason is that whatever scouting report they have drawn up, their report fails to answer the most important question:

What is the probability that player X is able to contribute to the team?

The answer is a number between 0% and 100%. This answer is never given in any of the reports or presentation scouts give. That means that the decision makers have to calculate this answer based on the report the scout has provided. Of course, they never do this consciously. Yet, our brain makes these kinds of estimations unconsciously all the time. If a scout does NOT explicitly answer this question, the brain of the decision maker is going to make the probability estimation all by himself. In almost all cases, this estimation will be lower than the players the decision maker prefers himself. That is the reason why even the most successful scouts only have contributed to a handful of transfers. Most transfers happen for other reasons than provided by the scouting team.

It really doesn’t matter whether we are talking about data, video or live scouting. If the final report fails to answer the question about the probability that a player is able to contribute to the team, the decision maker is going to answer that question and probably in a less favorable way.

So let’s look at an example. If you are using Wyscout data as a data scout, how can you then answer this most important question: 

What is the probability that player X is able to contribute to the team?

First you need to build a model that transforms Wyscout data into probabilities. Bayesian networks are most suitable for this job, but there are other ways. We prefer to use Bayesian networks. Second step is to validate your model. For validation we have created a Bayesian network to transform Wyscout team data into team probabilities. We calculate the following four probabilities:

  1. What is the probability that a team is going to perform well?
  2. What is the probability that the attack of the team is going to perform well?
  3. What is the probability that the defense of the team is going to perform well?
  4. What is the probability that the passing game of the team is going to perform well?

Here are the results for the Premier League and Watford:

Validation comes from the 89% correlation (R2=80%) between the probability to perform well and the rank of the team. This is in line with this correlation in other competitions. So to be clear: 

  1. The probability of Watford to perform well is 38%
  2. The probability of Watford to attack well is 37%
  3. The probability of Watford to defend well is 54%
  4. The probability of Watford to pass well is 47%

The next step is to look at the individual players of Watford. Normally we would look at all the players (except the keeper), but for this exercise we only look at the most recent starting XI:

Again, these stats answer the following four questions:

  1. What is the probability that a player is able to contribute to the team?
  2. What is the probability that a player is able to the attack of the team?
  3. What is the probability that a player is able to the defense of the team?
  4. What is the probability that a player is able to contribute to the passing game of the team?

As long as a player has at least one of these four probabilities quite high, he is an asset to the team. Of course, if it is only one category, he is a specialist rather than a generalist, unless that category is the overall category.

Taking into account minutes played we can then calculate the contribution each player has made to the team probabilities of Watford:

The contribution of these ten players is:

Here one can see that although Sarr has quite weak data in Wyscout, his contribution to the attack of Watford is on par to what is expected of him.

One can also immediately see that Pereya is the weakest link. So let’s look at a replacement for Pereya. As this is an example only, I am going to use a replacement who obviously would be better suited than Pereya. The player I am going to use is Liverpool’s Mané.

Here we use our transfer model. This gives the following results:

Let me explain this. First we start with the probabilities of Watford and Pereya we have already seen. Taking into account minutes played, we subtract Pereya from the probabilities of Watford. What this means for Watford is that the probability to perform well remains unchanged, but the probabilities to attack, defend and pass well drop a bit. 

Then we look at the probabilities of Mané playing at Liverpool. As you can see, for all but defense, these probabilities are much higher than Pereya’s probabilities. But in part, Mané is playing well at Liverpool because he is playing together with other great players. That won’t be the case if he transfers to Watford. So we have to take into account that his performance will drop a bit. But how much? Fortunately, we have a Bayesian model to calculate precisely that by taking into account the relative strength of both teams and minutes played. To make it explicit:

  1. The probability that Mané is able to contribute to Watford is 87%.
  2. The probability that Mané is able to contribute to the attack of Watford is 98%.
  3. The probability that Mané is able to contribute to the defense Watford is 5%.
  4. The probability that Mané is able to contribute to the passing game of Watford is 52%.

What this would mean for Watford is that their probabilities also go up when we add Mané with his Watford probabilities to Watford as is shown in the final row. With Mané playing for Watford the new probabilities for Watford are:

  1. The probability of Watford to perform well is 45%
  2. The probability of Watford to attack well is 43%
  3. The probability of Watford to defend well is 46%
  4. The probability of Watford to pass well is 55%

GIven the correlation between overall team performance probability and rank, we can also see that Watford would rise to somewhere between rank 10 and rank 15 in the competition once Mané is playing for Watford. 
Rational decision makers use these kinds of models to calculate for every player they are seriously considering hiring what the probability is that the player is able to contribute to the team and what this means for the team. Once you have ranked all players according to their probability to be able to contribute to the team, you try to hire the best player available. That is how we were able to transfer Dalmau to Heracles for instance.

This is the kind of work that we are going to teach at the Football Behavior Management summer school at the VU-university in Amsterdam in juli 2020. Due to the current circumstances this will be an online course.

Wyscout data to Bayesian team ranking

Without live matches I found time to work on my third iteration of my Bayesian model to turn Wyscout data into Football Behavior Management (FBM) data. To be clear: we only accept correlation above 80% and R2 above 60%. So far all 7 competitions checked have a correlation of at least 80% and sometimes it goes up till an astonishing 95%!

The Wyscout team data we use are:

  • Average goals scored
  • Average goals conceded
  • Shots off Target
  • Shots on Target
  • Passes inaccurate
  • Passes accurate
  • Recoveries (low, medium, high)
  • Losses (low, medium, high)
  • Challenges failed
  • Challenges won

Als please note that we use team data for these correlations that can NOT be traced back to individual players. Unlike the correlations we get with FBM probabilities that are based on stats of individual players and that can be traced back to these players.

Premier League

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Liverpool19984707454
ManCity28193728353
ManUnited36677636651
Chelsea46676547451
Leicester56275606652
Tottenham65964546249
Wolves75961545850
Arsenal85664527947
Sheffield95452535450
Burnley105441444850
Southampton115244445149
Everton124945425747
Newcastle134433365147
Crystal144336365447
Brighton154149406248
West Ham163944435448
Aston Villa173533365246
Bournemouth183433355346
Watford193431335147
Norwich202125235944
Correlation with rank 93% (R2=88%) and points 90% (R2=81%)

Bundesliga

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Bayern16193738054
Dortmund25789697852
Leipzig35484706852
Leverkusen45380607350
Gladbach55269606149
Wolfsburg63951485450
Freiburg73750455749
Schalke83745415847
Hoffenheim93651408047
Koln103444435448
Hertha113443435546
Augsburg123033414646
Berlin133033384848
Frankfurt142846435450
Mainz152730345145
Dusseldorf162431315547
Bremen172131305744
Paderborn181831345444
Correlation rank = 91% (R2=82%) and points = 96% (R2=93%)

Bundesliga 2

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Bielefeld15380656751
HSV24675606849
Stuttgart34575547350
Heidenheim44456515649
Darmstadt53952495549
Aue63846495147
Kiel73855486048
Greuther83751505447
Hannover93545387650
Regensburg103434444447
St. Pauli113348465547
Bochum123246445647
Osnabruck133042435347
Sandhausen143042415250
Nurnberg153040385448
Karlsruher162833394648
Wiesbaden172833394748
Dresden182430404447
Correlation rank = 88% (R2=78%) and points = 93% (R2=87%)

Bundesliga 3

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Duisburg15164546150
Mannheim25065566150
Bayern II34851436445
Braunschweig44850525049
Unterhaching54757545450
Wurzburger Kickers64751515249
Ingolstadt74654535151
1860 Munchen84652515349
Hansa Rostock94555517351
Uerdingen104445445548
Meppen114351525248
Kaiserslautern124149485349
Viktoria Koln133852446346
Magdeburg143756515452
Chemnitzer153747485248
Zwickau163636484247
Hallescher173349445749
Munster183338435046
Sonnenhof192524284847
Jena201825295046
Correlation rank 79% (R2=62%) and points 85% (R2=73%)

Eredivisie

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Ajax15687697253
AZ25688707651
Feyenoord35075606850
PSV44972576650
Willem II54450505448
Urecht64172606449
Vitesse74160535949
Heracles83656535648
Groningen93554477452
Heerenveen103354466148
Sparta113343465147
Emmen123247396247
VVV132826305046
Twente142747406246
Zwolle152643386046
Sittard162636385446
ADO171926315046
RKC181532296045
Correlation with rank 91% (R2=83%) and correlation with points 92% (R2=85%)

Jupiler Pro League

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Club17074616353
Gent25574606651
Charleroi35463645051
Antwerp45356555350
Standard54959535849
Mechelen64448485448
Genk74452446147
Anderlecht84374596849
Zulte93647447546
Mouscron103349455748
Kortrijk113345455448
STVV123343376245
Eupen133033335447
Cercle142331315446
Oostende152224324645
Waasland162019235144
89% correlation with rank (R2=80%), 89% correlation with points (R2=79%)

Dutch Eerste Divisie

RankPointsFBM Wyscout scoreFBM Wyscout AttackFBM Wyscout PassingFBM Wyscout Defense
Cambuur16681716252
Graafschap26274645953
Volendam35567556449
Jong Ajax45473596847
NAC55064555951
Go Ahead64851535247
Excelsior74749525247
NEC84561565849
Almere94450497151
Telstar104448455846
Den Bosch113860526247
Jong Utrecht123845505046
Eindhoven133440415446
Jong AZ142848406544
MVV152730384646
Top Oss162533325646
Roda JC172234395046
Jong PSV182245376544
Dordrecht192031325445
Helmond Sport201720254845
91% correlation with rank (R2=83%) and 88% with points (R2=78%)

Showcase Sheffield: Sander Berge or John Lundstram

What we do with FBM contribution statistics is calculate what the probability is that a player is able to contribute to a specific team. For a player can strengthen team A, but weaken team B. We do this for four scenarios: 

  1. The best case.
  2. The most likely case.
  3. The worst case.
  4. The current form case, based on the current form of the players involved.

To calculate the probability that a player is able to contribute to a new team we look at:

  1. Difference between competitions.
  2. Difference between teams.
  3. Minutes played.
  4. FBM contribution stats.

We always look at how team A would do if they replace player X with player Y. In this showcase we look at how Sheffield United would do when they replace John Lundstram with Sander Berge.

The FBM contribution stats for John Lundstram are:

John Lundstram

The FBM contribution stats for Sander Berge are:

Sander Berge

At first sight one can see that Berge scores higher in almost every category. Only the probability that Berge is able to contribute to the attack in the most likely scenario (average) is significantly lower than that of Lundstram. Yet, it is important to take into account that the probabilities for Berge are with Genk playing in the Belgium competition and Lundstram’s probabilities are with Sheffield playing in the Premier League. So we have to adapt these values.

Calculating Sander Berge’s expected contribution to Sheffield

Because probabilities don’t point to a single event happening for sure, but for multiple possible events happening with a certain probability, we always look at the four different scenarios mentioned earlier. What we do in each scenario is subtract the actual contribution of Lundstram from the FBM stats for Sheffield and then add the expected contribution of Sander Berge to Sheffield. Sander Berge’s expected contribution is based on his actual contribution to Genk deflated or inflated to compensate for the differences between the competitions, teams and minutes played.

The most likely scenario

In the most likely scenario Sheffield will have a very similar overall performance. The attack of Sheffield will be a bit less strong with Berge rather than Lundstram. Defense will be the same. Transitioning & build up will be much better with Berge. The reliability of the team will remain the same. A 3 points higher FBM team score translates into one additional point in the competition for Sheffield.

The best scenario

In the best scenario Sheffield will not improve much. Sheffield will trade a slightly better performance overall and in attack to an even less functioning transitioning & build up. As you can see the best scenario is worse than the most likely scenario … worse for Sheffield. But with the best scenario, we look at the best performance of both players and compare those.

The worst scenario

As you can see in the worst case, Sheffield is better off in almost every category except defense. The reason is that Sander Berge, even after compensation for a different team and a different competition, still has a higher floor in his performance than John Lundstram.

The current form scenario

The current form scenario shows that impact that Berge can make on the play of Sheffield if he is able to keep his current form at Sheffield. Our FBM contribution stats predict that he probably performs a bit less than this as per the most likely scenario. But there is a decent chance that Berge will perform like this at Sheffield, netting Sheffield 2 extra points at the end of the season when they start with Berge instead of Lundstram.

Conclusion

In all scenarios Sheffield is better off with Berge than with Lundstram except in the best case scenario. When both players play at their best, it makes little difference to Sheffield who is in the starting XI. Yet, Berge is currently performing close to his top performance, while Lundstram is currently performing below his median performance. This means that in the short term, Sheffield’s performance is most likely to improve when they start with Sander Berge.

Probability that Sander Berge contributes to Sheffield84%
Probability that Sander Berge contributes to the attack of Sheffield75%
Probability that Sander Berge contributes to the defense of Sheffield84%
Probability that Sander Berge contributes to the transitioning & build up of Sheffield22%