Football Manager Sabermetrics Principles

So, Comolli went out the window at LFC, but that doesn't necessarily mean the Moneyball and Sabermetric principles have gone the same route. In fact, it's important to remember that the club is still owned by a person who brought an old baseball giant back to the top by applying Moneyball principles. What really interests me, however, is whether playing Football Manager with Sabermetrics principles can work. The problem, as many have pointed out, is that Football seems far less quantifiable than baseball. Is that really true? I don't believe it is. It's a matter of identifying which metrics are important to the success of a football club. For now, I fear more astute minds than mine must take charge of sorting that out.

How Do You Use Sabermetrics With Football Manager?

Without identifiable metrics it's pretty hard to use sabermetrics with football manager. It's also really important to remember that while sabermetrics might be about stats Moneyball isn't! That's right, Moneyball is not about stats. In Moneyball the goal is to get value for money by taking advantage of market deficiencies. That is something we can work with in Football Manager. We can device a Moneyball transfer strategy pretty easily. We can also start analysing which stats affect the outcome of a match the most. I think goals for and against might be important, but I'm just guessing.

A Football Manager Sabermetrics/Moneyball/Soccernomics Strategy

The best guide for a transfer strategy comes from the book Soccernomics. It's a great book and you should be able to see an ad for it on this site. If you're interested in the economy of football it's more than worth a read. Some principles regarding transfers are outlined that could be applicable to Football Manager.

Never Buy Starts of Recent World Cups or Continental Championships

How many times have we seen a player purchased directly after a continental championship only flop exceptionally at his new club? You can't judge a player only based on a handful of games in a tournament that is likely to be one of the highlights of his career. Moreover, stars from championships tend to be grossly overvalued. Don't ever purchase a player based solely on his most recent performances. Check the players history, injury record etc.

Don't go Dutch on a Brazilian!

Frankly, I have no clue if nationality affects transfer value in Football Manager. It might be worth researching. In real life Dutch, Brazilian and players of other fashionable nationalities tend to go for a premium. See if you can avoid paying a premium on players of those nationalities. Consider the value you might get from players of more obscure nationalities.

Don't Keep Players Past Their Prime

Just as with the stock market you should sell when you can get the most money for your players. Each season you should look at the players seasonal performance. If it seems obvious the player is starting to decline; sell him! Kill your darlings.

How Lyon Did It Before They Lost The Plot

Lyon dominated French football the past decade by being smarter than the rest in terms of dealing with the transfer market. Here are the simple rules they used.

• Running a football club isn't a one-man show. Transfers will be decided by a group of people.
• In Football Manager that means you need to get several opinions on a player before signing him. Send out your best scouts, your assistant manager and anybody else you think can offer a valuable opinion. I'd say at least 2-3 others should approve the signing before you go ahead with it.
• Buy players in their early twenties.
• The reasoning behind not going for 17 or 18 year old is that you have a far better idea of how good the player will be when he is aged between 20 and 22. The danger, I have to admit, is that you may have to pay a hefty price for the players. It will be very interesting to observe.
• Help your foreign players move to the new country.
• This can be difficult in FM, but it might be worth securing at least one other player speaks the language of the player. It will probably also be a good idea to get someone to tutor him to build  a relationship.
• Don't buy strikers. Develop them.
• Ooh, it's tempting to disregard this rule. Strikers are after all some of the most exiting players in the game. If you decide to follow the rule and not buy strikers it is probably important to strengthen you attacking and shooting training capabilities to develop your own strikers.
• Sell players if you're offered more than they are worth.
• Sounds obvious, but it can be hard to sell your best players. If you're offered twice the value of the player you should definitely consider selling.
• Replace your players before you sell them. If you notice a rising interest in one of your players that is the time to look for a replacement.

Sabermetrics is a key aspect of Moneyball. Implementing it in Football Manager is a process. We have started with the part of Football Manager that is the easiest to apply these principles to. We won't stop there though.

Positional Analysis In Football Manager (Re)Defined

After the frantic final days of June I have slowed down the posting a bit (actually I haven't made any), but that doesn't mean I'm having a summer meltdown. Right now I'm working on collecting data for all the positions I have yet to analyze. During that process I'm thinking about a couple of things. Things that may lead me to revisit my initial posts on the matter.

The fact is that I failed to set specific criteria for what would make an attribute count as an essential attribute. That means I'll have to revisit my initial posts about the positions and my initial posts about the players in the Lota Schwager squad. None of this, however, will happen until I have collected data on all the positions. This post will define, as clearly as possible, the criteria defining the analysis.

The Player
Any attribute with a value of 16 or more will count as a strong attribute in the player. Most would probably agree that 15 is a pretty decent ranking for an attribute. To ensure that we get only the best attributes I decided to make 16 the treshold used in this analysis.

Players valued at more than £1 million or the top 100 will be analysed. If the value of the 100th player is tied among multiple players all players of that value will be included. That means there can be more than 100 players in the analysis. However, as seen in my Sweeper research teh number can also be lower. If the number of players analyzed is lower than 100 I will define the data as insecure. In truth 100 players probably isn't enough for a proper analysis, but time constraints have caused me to set that limit.

The Data Set
If an attribute has a rating of 16 or higher in more than 30 percent of the players I consider than attribute to be significant. The reasoning is that 16 is a pretty harsh requirement and more than 30 players in a set of 100 is a pretty significant group.

Evaluating The Player
The purpose of this is te enable us to make the right choices in the transfer market. So we want to rank the players according the attributes we find significant. However, and attribute with a rank of 16 or higher with 60 percent of the players should be considered twice as important as an attribute that only has a ranking of 16 or higher in 30 percent of the players. To that end we calculate an attribute facotr for each player. Here's an example.

• Heading - 35%
• Anticipation - 40%
• Composure - 35%
• Off The Ball - 50%
• Jumping - 60%
• Strength - 55%

This is not a real example, but could feasibly be a target man type analysis. As you can see each attribute is assigned a percentage value. This percentage value indicates the ratio of players where this attribute was significant. In this example the total percentage value is 275%. This number is important in calculating each attributes relative significance. We do this using this simple formula:

Attribute Percentage * 100/275

From that we get these values.

• Heading - 13
• Anticipation - 15
• Composure - 13
• Off The Ball - 18
• Jumping - 22
• Strength - 20

The sum of those values is: 101.

So, finally we are ready to calculate a players base ability according to this made up analysis. The player in question R. Andom a target man with unknown potential. His stats are Heading 12, Anticipation 15, Composure 13, Off The Ball 14, Jumping 20, Strength 17. His base skill level, according to my analysis, is:

((12*13)+(15*15)+(13*13)+(14*18)+(20*22)+(17*20))/101 = 15.66

So R. Andom has a base skill value of 15.66 when performing in this made up position. That's actually not a bad rating. Although his headers aren't always accurate he seems able to find space by using his strength and jumping higher than anybody else.

I'd love to hear your thoughts.

Problems With Football Manager And Moneyball

Over at Howard Hamiltons Soccermetrics blog there is a great blog post about Moneyball and Soccer. Howard describes a metric he believes would be awesome in regards to football. I would love to get to work on that metric, in my own amateurical(is that actually a word?) fashion. However, as far as I can tell that's goiing to be a bit difficult. He describes some factors it might be wise to take into account for each of goalkeepers, defenders, midfielders and strikers.

Let's first relate those to what's available in FM. Italicized stats are those you can feasibly derive from other stats while bolded stats are those you can't easily figure out.

Goalkeepers
Suggested stats: saves, goals conceded, penalty kicks conceded, corner kicks conceded, passing/distribution %, balls won.

Defenders
Suggested stats: goal kicks forced, balls won/lost, corner kicks conceded, fouls conceded, penalty kicks conceded, passing %.

Midfielders
Suggested stats: passing %, balls won/lost, assists, fouls conceded, fouls won, corner kicks forced, shots, goals

Strikers
Suggested stats: shots, goals, passing %, balls won/lost, corner kicks forced, fouls won.

For the more defensively inclined positions that's a lot of metrics you simply either can't get or will have to work very hard to get. In my view this is a major shortcoming in the game. Another major shortcoming is that you can't see the players stats for previous seasons. Or at least I haven't figured out how to (if you have please be so kind as to tell me in a comment below!) do it.

In my view Howards startiing point is sound, and he definitely know a lot more about statistical analysis than I do (by a lot I mean much, much, much more than I do). Having said that, I can't stop thinking that the categories in his Moneyball and Soccer post are a bit too narrow. I mean would the same stats apply to a central defender and a fullback? What about the not so subtle differences between defensive midfielders, wide midfielders and attacking midfielders. These are things I believe would need to be adressend, but if it all got sorted out you would have an amazing metric to go by.

So, where does that leave us. Well, the game isn't totally useless. We can start to look at how the stats available to us influence the chance of winning matches. Obviously the average rating seems like a good indicator of the players quality, but any random player with a high average rating might not be the piece of the puzzle that makes your team come together.

I suppose it will be a matter of looking at your gameplan and trying to figure out what you're missing to make it complete. Just remember that you might not need a worldbeater to fill in the last piece of the puzzle.

Finally, I just want to plug Howards Moneyball and Soccer blog post one more time. It's really good!

Transfer Strategy - The Moneyball Saga Continues

More chapters of Moneyball has been read and more interesting theories have emerged. Targeting high school players is generally considered a bad idea, whereas targeting college players yields better results. Since football(soccer) outside of the US generally doesn't have that distinction how would that transfer to Football Manager and football in general?

The point is that the very young player is more of a gamble as empirical evidence of his quality is less profound. He is also mentally less prepared to the pressure of first team football. The advantage of targeting youth players is that they are usually fairly cheap.

In Football Manager if a player reaches the early twenties without showing his potential he is more often than not written off completely. Football Manager operates with 2 attributes behind the scenes that is a genuine cause for virtual managers to do this. Those two attributes are current ability and potential ability. These traits are hidden from the manager, as they should be, but it seems logical to assume that if a player is not showing signs of ability when he is close to his prime he never will.

To follow Billy Beanes transfer strategy you would mainly have to target players ranging from 19-23 years of age. Those are players who are no longer considered youth players, have had time to show their potential at their clubs and have yet to reach their prime.

That, however, is not all. You need to determine their ability in the game. Physical stats such as Pace and Jumping have often been highly rated in Football Manager. Now is the time to try to look past that. In a prior article I did some research on the prominent attributes of world beaters, and those are good traits to look for. However, I'm suggesting not using the player attributes as the first metric when considering whether or not you are going to sign a player.

Look at the players performance in the past. How it is developing. And how the player is contributing to the overall team performance.

At the start of the game this is nearly impossible and that's when you probably have to focus mainly on the players attributes. Try to look past the numbers though. Personality should play a huge part. As should a players prefeered moves if he has any.

Right. This is just a bunch of ramblings right now. And honestly I don't really know what you should primarily look for. I'm still investigation and so are a lot of other people. Allow me to suggest looking at the Soccer Analysts blog which is frankly pretty awesome!

Going Moneyball On FM

If you have read some of the previous blog posts you will have noticed that I'm fairly obsessed with Sabermetrics. So, now I've started reading the book that started it all, Moneyball. Although the book is mainly about baseball I have a strong feeling that the principles can be applied to pretty much anything. So far the book revolves around one man, Billy Beane. A guy who was himself touted as one of the best things in baseball since sliced bread. He didn't succeed at the highest level of the game as was expected, but he learnt many valuable lessons.

The main principle I can take from the book so far, and I'm only two chapters in, is never to take what you see as a measure of a players quality. What you need to do is take a good look at first the game, and then the player. You need to find out what is the most important factor in winning the game. Then you have to find the players contributing the most to that.

Since the book has more than a few more chapters in it I'm pretty sure that's oversimplifying it quite a bit. However, it seems like a good starting point.

So, what wins footie games? The obvious answer is goals. It's just not enough, though, looking at who scores the most goals. You need to find the factors that a. creates goals and b. prevents them.

So, how would one go about this? My idea, which of course involves excel, is to take a look at a lot of games played in the game. Looking at the stats and use the data to figure out which are the most important factors to winning.

This should actually make it easier to create tactics as well as sign players. The main problem though would be the lack of stats when just startinig the game. So, the analysis would require a save with at least one season in it. I would have to look at the stats for all games played to determine the most important game winning factors.

Sounds like a lot of work, but also sounds like a very fun challenge.