Search
On FM Scout you can chat about Football Manager in real time since 2011. Here are 10 reasons to join!
fm23 early access - download steam key

Moneyball Guide - The art of numbers, not attributes

By on Aug 26, 2022   4238 views   1 comments
Football Manager Guides - Moneyball Guide - The art of numbers, not attributes
This guide is written by Daljit / Rashidi, a tactics expert and moderator on SI forums. He also creates video content on YouTube, so make sure to check out his channel BusttheNet Gaming.

THE ART OF NUMBERS


As some of you are probably aware my livestreams with Palermo, just became a bit more interesting, at least for me. We have done away with displaying attributes in the game entirely. No attributes shown, I will be relying on coach reports.

I want my coaches to assess the players and give me a rating of how they stack against players in the squad. I want to know if a coach thinks one player is considerably better than another. And I like to play it that way.

I think the "proper way" is the way you enjoy it. And this is how I enjoy the game. To each their own.

A lot of people may not understand how to play completely clean even without star ratings, this is a way for me to show them how they can incorporate the understanding of statistics into the game.

Now it's not 100% possible for me to escape attributes unless I get a custom skin made. So I will just remove the attributes and avoid the training page.

It does however make it interesting when it comes to scouting and player assessments. Out goes attributes and in come statistics. Why have I done this?

I am not on an ego trip, the goal was to show that sometimes paying too much attention to attributes can be misleading. 2 players with slightly different attributes can still have big swings in performances during games.

Sometimes it's hard to isolate whether an attribute is causing this or something else. The range of "something else" is pretty big in Football Manager. There are a whole host of hidden stuff that is under the hood that can play a part in causing swings in performance.

If there wasn't anything like that in the game then you could give every player 15 for attributes and they would perform identically in every game. And we all know that never happens.

And I am not the first to play with attributes, many others have done that. But why?


UNDERSTANDING YOUR PLAYERS


Whether you play with masked attributes or unmasked as a setting honestly makes little difference. That's because hidden attributes play a big part in the game. A player's mental state can affect him if the team goes down by a goal. No two players will deal with pressure the same way.

Playing without ANY attributes forces me to drill into each players performances, it makes me pay more attention to slight nuances in the 3D representation of what happens in game. I spend more time thinking of how the player is reacting to my shouts.

And I look at statistics to tell me if a player is working hard enough, or whether he has the football smarts to do what I need him to do.

It's challenging playing like this, because playing with attributes makes things easy and misleading at the same time. It's easy to think you may have found the right player for a role, but it's also misleading because there is more information out there that can paint a different picture.

I am sure many of you have signed a striker and gone on to question the signing when he's failed to do the job. There are numbers in the game that will tell you if a player is a great finisher, whether he spends more time trying create chances for others.

Numbers exist within the game that can split the difference between two players and tell you one is better than another simply in terms of how he can read the player before others.

Statistics can help you determine which of your players has been performing better over a period of time and numbers don't lie unless of course you have been doing something crazy in the background to make them unreliable and naturally we need to assume that the numbers are accurate.

I am not trying to convert people to this statistical approach, I still think people should play the game the way they want to play it. And if people feel attributes are the way to go, I am all for it. I am just doing this so people don't forget that stats can be just as important. And I think people who can incorporate both an understanding of attributes and statistical discovery will have a lot more fun playing the game.

A taste of what's to come

So here's a primer. I was out looking for a great striker, we had one in the team but he simply wasn't scoring as many as I would have liked. We went out to search for another. So I created a view and a filter.

The custom view gave me access to information about a team, which I now use across multiple saves.



I wanted a striker who had a good XG/Goals ratio. I spotted someone called Morten Hansen who had played 12 matches for the senior team and had a 9/4 ratio which is a lot. In other words he had an XG of 4 and goals of 9.

We sent a scout out to reveal his hidden qualities. He was spirited, pacey and consistent with good versatility. Without even looking at his attributes we signed him.



He is only 18 and he has transformed our attack.




PLAYING BY STATISTICS ALONE


I am not recommending that everyone go out and play without attributes. In fact playing without attributes requires you to do several things.



Playable Leagues vs Non Playable Leagues

In playable leagues and depending on your detail level you have access to stats and match reports. This makes playing without attributes a possibility, you don't have access to stats from non playable leagues. If you are playing view only then you will have limited access to stats.

From what I have gathered it's best to play with a combination of playable and view only leagues. My database has 170k players in it and I can't remember how many leagues are loaded. I think there are a lot.

A COLD START

When you begin a save, invariably your appointment happens in July, by then no statistics are available for the preceding season, which makes playing without attributes almost impossible for someone trying in the lower leagues where we sign players by trialing or picking up free transfers. Most free agents hardly ever play a game. So statistics will be in short supply. So you need to do your scouting by June 23rd.

You can go attribute-less on any skin all you need to do is go to Preferences > Skin Colours > Advanced settings > and set the "A" value for RGBA settings to "0".



With a cold start, after taking over a club, well you end up depending on your assistant managers recommendations or just using the role ability algorithm to make judgements on what roles your players can play and once the statistics start to show you make adjustments. This is naturally harder and something I wouldn't recommend to anyone unless they were Cleon.

An alternative and probably easier method is to use the first window in pre-season to leave attributes on until preseason is over without signing ANY players. Then once you are comfortable with your team you turn attributes off.

Funny thing, you do have access to attributes if you go to training and progress. For me this is usually more than enough.




TRAINING AND YOUTH DEVELOPMENT


Min maxers of training will probably not enjoy this style very much because they are the sort to move a player from one role to another because they want to develop players in a targeted way. And it's understandable, because it's slower than having all the attributes available at one glance.

When training players I spend more time looking at 3d and stats to make my evaluation. I also typically follow a simple plan when it comes to youth players.

If I want them to get a crack at the senior team early I use role training that is simple. Young strikers are put on poacher training for example. Or I just assign them into roles that they are expected to play in the tactic.

So far I have not seen too much downside to training, since my training is generally balanced anyway.


TACTICAL SHENNANIGANS


I usually look at stats after a few games to determine who should do what. The stats I am looking at really depends on the system I plan to play, and, it's not going to be the same for every team.

Some teams may see a bigger focus on defensive and positional awareness. Maybe I am an underdog and I need to make sure my defenders are positionally and physically solid. So I may zoom in on height and interceptions as a requirement for the majority of my players.

So a one size fits all is something I want to avoid sharing. This is what I am doing with Palermo in my current livestream. Now please be aware my journey playing attribute-less with them only happened this season. Every new player now being added and every match plan I make for my opponents is now being decided without attributes.

With Palermo for example Support players need good interceptions, passing completion %, tackling ratios. Some may need good key passes/90.

When I go into streams you will notice my match preparation before the Arsenal and Bayern games which we won and you would have seen me checking the stats of the opposing teams and sometimes even watching highlights or their losses. I would do this to prepare my choice of starting 11.

You can't access previous season chalkboard stats from within the game. If you want to keep them you need to export them before Jun 23 out to an excel spreadsheet.
Personally I am > 50 years and my days of using excel sheets for playing FM died in the days of CM. I don't want to play FM with an excel spreadsheet, but it's actually pretty easy to export and keep them in an excel spreadsheet.


THE GOALKEEPERS DILEMMA


Traditionally assessing a keeper through stats has been difficult. We have usually just gone with how many has he conceded. Today we have the xG model which doesn't make things easier.

What numbers are we supposed to look at, there can be a lot. Do we look at xG against, conceded, shots on target against? That was my recent issue.

So our goalkeeper Matteo Pisseri has opted to retire, previously in my games I would just search by attributes it made life easy. Now with this different approach, I have found that keepers can be a lot more annoying to find.

In the game from custom views we can only generate these statistics.
1. Clean Sheets
2. Conceded
3. Conceded/90
4. Save Ratio (which doesn't split between shots from outside the box and those from inside the box)
5. Expected Save Percentage (which I assume is the inverse of xG in other words these is the percentage of shots he was expected to save, more on this later)
6. Saves Held, Tipped and Parried
7. Mistakes Leading to goal

When you go to player performance reports from coaches you get under Advanced Goalkeeping a scatter group that plots expected goals prevented vs save percentage which is incredibly helpful if we had expected goals prevented as a searchable option under custom view. Why?

Save Ratio by itself can and is usually misleading, it needs other metrics.

A goal area is roughly 244cmx732cm and an average goalkeeper is expected to cover 24% of the goals area (gonna avoid the math here and keep things simple), but if we assume keepers are 1.90 on average and you use those dimensions you will arrive at a similar number using a Vitruvian circle to assume the goalkeepers area.

On average most keepers save around 69% of shots on target, these ratios will depend on whether the shot is inside the box or outside the box.

Therefore since a keeper already covers 24% of the area of a goal, it implies that 45% of his saves (which is basically 69%-24%) will come from..... his attributes :-) I don't think we are too far away since, on average the penalty conversion rate is around 76%.

Ideally to evaluate any keeper it would be easier if we had...

Clean sheets per 90 vs Shots on target against per 90 as a statistic by teams which we currently don't. Why would this be helpful, If we plotted that as scatter group then any keeper who faces a high shots on target against per 90 and has a high number of clean sheets would be extremely valuable. Such keepers are keeping a high number of clean sheets in spite of facing many shots

THE SAVE RATIO

Save ratios are helpful, but they don't account for how sides defend.

Let's say you have a very organised defence and you are giving strikers poor quality chances, then a keepers save ratio could be higher than a keeper who has a poor defence that allows them more shots from within the area.

What we need is Expected goals per shot on target against. We also need to add Expected Goals Conceded/90 to the filters as well.

The best thing we have at the moment is expected goals prevented and I would hope that is added as an option we can filter.

So I have ended up at a crossroads on my save as I decide how to find my keeper with the current tools without having to use an excel spreadsheet, cos I made a promise to myself that I would NEVER use spreadsheets for a game ever again :-)

INFORMATION ACCESSIBLE BY LEAGUES

Within the league data we can find under Player Detailed several bits of information, and I think I found a workable ratio.

Expected Save Percentage - I am assuming this stat calculates the difficulty of the shots a goalkeeper faces, based on variables such as shot placement, velocity, distance, etc, to determine the chance that an average keeper will save the shot. Basically in a nutshell, the chance to save a shot.

Expected Goals Prevented - Essentially how many high chance goals a keeper prevented

Saves per 90

Expected saves per 90

So this is how I am gonna find my next keeper:

A keepers Expected Save Percentage cannot be lower than his save ratio.



In our own league we have these keepers who have a decent expected goals prevented. Next.



My eyes immediately land on Andrei Vlad.

Now there is a scatter group for advanced goalkeeping stats which is why i recommend you always SAVE before Jun 23, otherwise you could lose the preceding seasons stats. When you go to check Reports > Player Performance, you will be able to access the scatter groups.



We want our goalkeepers to be in the Quality Shot stopping / saves most shots quadrant. And it clearly shows that I am heading in the right path when I search for a keeper, Andre Vlad is in the same quadrant, his expected goals prevented is slightly better, and if I use this method, who needs attributes! So this keeper is a firm B, in my books. A would be the keepers higher than 80% and >6 prevented.

OUR SEARCH FOR A KEEPER

The search for a keeper was focused on looking at several metrics.

1. Conceded / 90 v Expected goals conceded per 90

2. Expected goals Prevented vs Save ratio

The first metric depends on your team and you need to do a deeper dive. The second metric is faster but less accurate but at least you get a picture of whether a keeper is reliant. We whittled it down to a couple of keepers.

There were several contenders, one was Russo who happened to have some crazy numbers in Serie B and we decided to bring him on as a backup. For less than 1m that was a no brainer.

When you scout players depending on your skin you can get detailed scatter groups and polygons. Using the Tato skin these are found within a tab.

A young Brazilian keeper caught our eye. By 18 he had had two full seasons of football with a top flight Brazilian club.

His expected goals prevented vs save ratio indicated he was good.



The media description of him was "Commanding Goalkeeper" we signed him for 17 million which is the most I have ever paid for a player (I never scrooge on keepers by the way), and then once he joined the club his description changed to Wonderkid and his value went from 17 million to around 70 million.

Really happy with our keeper signing. Stands tall when needed, Loads of solid reflex saves and is already just behind Mike Maignan and Onana as the most effective keepers in Serie A.


THE METRICS


Time to list out the metrics that I usually pay attention to, these are to assess players.

There are plenty of people who play the game using metrics. Some use them to develop a moneyball strategy, others just use metrics to determine the best players for the system. I typically land in the latter group.

However when it comes to signing players I usually think in moneyball terms by attributing their values. To do so I will usually use a simple way of doing this by evaluating a metric and then dividing it by the salary per annum of the player. I prefer using salary per annum as my budgeting is usually done annually.

There are some notable content creators, but the best one I am aware of is FMStag who even has a calculator you can use to do moneyball strategies.

I myself don't want to use excel spreadsheets as I want a simple way of doing it. Using spreadsheets is definitely a superior way, but I just don't want to play the export/import game on between FM and Excel.

I will list out all the metrics explain why they are important for me and then within this post or maybe within another I will use case studies to explain my decisions.

Caveat: For any player evaluation to be accurate, that player should be within a playable league. In other words, if you want your data to be accurate, playable leagues are the way to go.
When you evaluate players you typically need a good data range. You ideally need players to have made at least 5-10 appearances. If you out to sign wonderkids, I have done a video on this which explains how I use filters to find the best wonderkids. The principles of both are almost the same.

All examples I mention below with data, come after one season of simulation.

GOALKEEPERS


Save Ratio - Measures how many shots a keeper saves against those taken.
Generally most average keepers will have an average ratio of 69%. Shots outside the area usually have a higher save ratio. The game however does not distinguish between the two, and a good yardstick for a decent keeper is around 75%.

So if you want to search for a good keeper you start with a 75% save ratio.

Expected Goals Prevented - This piece of information is only available via the league detail stats for goalkeepers. And you cant really filter it. What I recommend doing is to have as many leagues set to playable. This is where you will need to shortlist all keepers with a high save ratio, then physically check each keepers advanced goalkeeper statistics. Some skins may help you with that information or you may need to go to the league stats. Within a leagues player detailed stats you will find information on the keepers

Expected goals prevented basically informs you on the quality of his shot stopping, the more he prevented these high xG chances the better a shot stopper he is.

Saves Tipped / Parried / Held - While this information is useful, it only serves to paint a picture of the type of keeper you have. If you want someone with good handling chances are he will be generating more “held” numbers than “tipped” or “parried”

Outfield Stats


ROLES / POSITIONS THAT REQUIRE DEFENSIVE CONTRIBUTIONS

Clearances - When a defender has high clearances, it means that he brings the ball out of defence to clear it with a pass. So double check that against pass completion or key passes. Clearances can be a very good indicator for a ball playing defender

Headers Won Ratio - Central defenders need to win headers. I expect most central defenders to win at least 85% of all headers. If you find anyone who does higher numbers than that, it's worth investigating.

Height - In simple terms a player who is at least 1.9 is likely to have a high jumping reach, and if he has good headers won ratios, then he could have good heading.

Interceptions per 90 or Possession won per 90 - Either one of these tells us that a player is great at reading the game. He most likely has good anticipation and concentration. I rate this very highly for all players.

Tackles won ratio - This tells us how many successful a player is at winning the tackle. Some strikers could have good tackles ratio, which is indicative of bravery and aggression.

Tackles Per Game - Aside from the ratio, tackles per game give us an idea of how many tackles a player makes. Be wary of this, there might be times when some players may have more tackles per game, this doesn't necessarily mean they are good, just that their side tends to be more defensive. While I do value tackling metrics, I place interceptions per 90 at a higher level because that metric indicates the footballing intelligence of a player.

Key Tackles - How many times a tackle stopped a goal scoring chance. When I signed Tobias Sagusten Andersen for Palermo at the age of 17 for free, he was already making a few senior appearances for his club in Norway. His data indicated that even at 17 he was winning 87% of his headers and more importantly he had a high number of key tackles. With his interceptions per 90 at almost 4, I didn't hesitate to sign him without even looking at his attributes. He was signed by Palermo in Serie C, he is now the captain and one of the leading defenders in the Serie A.

These group of metrics serve to outline a players defensive contribution to a game. Ball playing defenders typically have good clearances. Any player who brings the ball out of defence will have higher clearances than a player that does not. A player who generally has a good spread of outfield stats that cover these general defensive contributions is a player that will be good in most roles that require him to support midfield and defence.

ROLES / POSITIONS THAT REQUIRE CREATIVE OUTPUT

Key Passes / 90 - Key passes are basically a cross or pass that led to a goal scoring chance taken. If a goal is scored then that key pass is registered as an assist. A key pass can come from set pieces or open play. Players who can produce a high number of key passes could either be skillful at playing a killer pass or great at set pieces. It's another important metric to check.

If I am looking for players in midfield to unlock defenses I am normally looking for players who do around 1-2 key passes a game. When it comes to fullbacks who are expected to cross I am looking at around 3-4 key passes a game. If my game is based on a heavy crossing style then I expect even more, but I rarely go above 4. Generally when a player is doing more than 4 I always check to see if he is a set piece taker.

A player with high key passes per 90 is worth checking out because either his set pieces are usually generating goal scoring chances or he is great at unlocking defenses with killer passes. This ties in very closely to assists and clear cut chances.

The issue with key passes / 90 is that completed crosses that don't lead to a goal or a freekick that connects but is saved by a keeper are also counted. So I tend to add

Crosses Completed - If a player has high key passes / 90 and high crosses, I will hesitate, I may even go into his detailed stats, watch a game or two to find out if he is taking set pieces. The combination of key passes / 90 and high crosses completed could indicate a player who delivers set pieces very well.

A good example would be Bernado Silva. In a simulated season of data I discovered that he was doing 1.72 key passes per game with 13 assists but he also had 70 crosses completed. This told me that he was probably having a large number of his goal scoring chances coming from crosses.

Assists and CCCs - A completed cross that ends up in a goal is not counted as a key pass but as an assist. A ccc is a goal scoring chance in the game. I noticed that since FM22 when you mouseover CCCs, it now says overall number of chances created instead of clear cut chances. This also correlates to the data that I have been studying.

Dribbles per 90 - This is a useful metric to identify two kinds of players. Strong wingbacks and outfield players who like to break tiers by dribbling through them. A player who can dribble can carry the ball through the tiers forcing sides to collapse on themselves. Since the player is skillfully bringing the ball up, defending sides tend to fall back.

Sometimes more than one defending player can be drawn to a skilled dribbler. Eden Hazard in his prime was one such player, who used that skill to drive at defenses drawing opposition players to him thereby creating space for others.

Dribbles per 90 is useful to identify mezzalas too. You don’t need a high number you just need someone who is doing between 1-1.6 a game. Players like Carlos Soler can do around 1.67 dribbles per game. Another good example, Christian Eriksen who does around 1.45, James Maddison, 1.37 and Luca Paqueta at around 1.2.

Fouled Against - Is also useful, a player who has a high FA number next to him is usually one that the AI targets with hard tackling. Every time your assistant manager recommends hard tackling as an OI it's usually against a player who needs to be stopped. And players with high dribbling, agility, first touch are usually the targets.

Bernardo Silva typically receives the most fouls against in a season as a midfielder. So if I were looking for a winger, inside forward, inverted winger or a mezzala on attack. I would zoom in on dribbles/90, fouls against, key passes/90 and pass completion rate.

Distance covered per 90 - Is a useful measure and sometimes can help you determine if a player is pacey, but it's a very inaccurate measure. While a high distance covered/90 might indicate good work rate, you still need to send your scouts out to find out if a player has pace. That's really the only way.

Passes completed per 90 High key passes alone isn't usually enough to identify great playmakers, you also need to know if they playing a lot of passes during a game, generating ccc and of course getting assists.

Pass Completion Ratio - Indicates how many passes a player completed

A player with high passes / 90 with low key passes/90 could simply be playing as a central midfielder on defend duty within a system that is generally played on balanced mentality, suggesting that the player makes a lot of short passes. A playmaker would generally be seeing around 1.76 key passes a game with high passes per 90.

In a simulated season, Casemiro had 87 passes per 90 but only 0.87 key passes with an assists/90 number of 0.10, which suggests that his playing as a CM(D). Frankie de Jong on the other hand had 80 passes per 90 with 1.56 key passes a game which suggested he was playing a bigger part.

When scouting the player I found that he also had the trait plays short simple passes, this suggests that he could have been played as an advanced playmaker on support or a mezzala on attack. I did find someone even better; Jude Bellingham who tore through the creative numbers and had a good xG / goals ratio to boot.

SCORING STATS


Shots / 90 vs Shots %

I could be alone on this but I like to compare these two metrics. I could end up with a striker with a high xG but could be playing for a side that is so good that it's generating a high number of shots for him per game. Personally I get worried when players are taking a LOT of shots to score.

xG and Goals - At the very basic level this tells you how good a player is at scoring. Typically you want a ratio of at least 1.3 or higher, which means that if you find a striker and his xG is 10 he needs to be scoring 13 goals to be considered as a very good finisher of the ball. A player who has a ratio of less than 1 is not finishing his chances. I do not use xG to assess if a player is good at placing himself in good goal scoring positions although I do understand the merit of that.

The issue with xG numbers in the game is that they cover penalty and non penalty goals. We want to isolate penalties to determine if a player is finishing well in open play. To do that we need to account for penalties

Penalty Scored - xG value of 0.76

Penalty Taken



Mbappe has an xG of 18.71 for 39 goals which gives him a ratio of 2.1 which indicate that he has fantastic skill, his shot on target is 51%. This alone might suggest one should sign him. However we need to dig a bit deeper.

Now while Mbappe is good, Moussa Dembele is interesting, his ratio is 1.6 with a shot target ratio of 59% which suggests that he usually works the keeper.

We should not stop there. Some players also take penalties. And penalties are good xG chances valued at 0.76. We want to know how good a player is at converting non penalty chances. So we need to do some calculations to get a better picture.

Kylian Mbappe has an xG of 18.71 and has scored 39 goals. However 7 came from penalties. If each penalty produces an xG of 0.76. Then 7 penalties have to be discounted from his overall xG. This then gives us a value of 13.39 (18.71 - 5.32). And we need to knock 7 goals from the total giving us 32 goals.

His non penalty xG is then 13.39 and he has scored 32 goals. Now if we were to take his annual wage of 21,467,000 each goal has cost the club 670,843. There are other players who have good xG ratios and could come at a third of his value. That in essence is Moneyball.

If I wanted a striker who had a good xG ratio but could also be a nuisance for defenders I would check their tackles ratio and the tackles per 90 number. Here a standout player would be Dominic Calvert Lewin who has 0.76 tackles per game with a ratio of 70% which means he wins 70% of his tackles plus he has a good xG ratio of 1.8. On top of that he wins 8.2 headers a game with a header won ratio of 67% .I would sign him in a heartbeat. Why?

When I evaluate players I like to get a big picture view. I like to evaluate a players TOTAL contribution during a game. Naturally there are ways in which we can extract this information and then apply them to an excel spreadsheet and finally give scores to players. You could even use excel to create your own scatter groups if you so desired to look for specific kind of players.

What I did in the last example was to evaluate the attacking and defensive contributions of my striker to make my assessment.

What I plan to do is explain how using my formation as a starting point I go about determining the stats I will be tracking for each position.



When it comes to Scoring stats there are a range of numbers I will evaluate that finally gives me an idea of what a players non penalty goal xG is. Basically how good is he at finishing chances that are not penalties. This can also be a cool way to see if your system is creating chances.

I will also evaluate the creative output of my players, there are a couple of creative types in any team: some cross, some pass, some do set pieces. A basic way is to just whittle them down to chances created per 90 over passes played per 90. That gives us a ratio that is indicative of the proportion of creative output from a player

Finally we have defensive actions that can take several forms, I take all defensive actions ranging from headers won, tackles won and clearances. Interceptions per 90 are also included. I wanted to find a way of splitting key tackles per 90 as well, but i think I will ignore it for now in the calculation. However Key tackles per 90 are usually a decent indicator of how alert a player is at stopping a goal scoring action with a tackle.

And yes I have done a simple spreadsheet, so out of practice.



That's my team Palermo from this season.

I am still working on this... so there is more to come including a view I use to evaluate my starting 11's contributions during a season. That spreadsheet above can also be used to import data from player search windows.

We added Ramiro Hernandez to our team this season. We signed him because he had a good interceptions/90 ratio and stood 1.88m tall. Previously our best defensive pairing was Lorenzo Pirola and Saguesten Anderson. That partnership has lasted from serie C all the way through to Serie A.

After a season of play I typically review how we have done. There are several metrics I like to track:

GWIN, TCON / 90 and TGLS / 90

These are broad metrics that can help. Tobias and Pirola as a defensive pairing we win 51% of the time. When Hernandez plays he is in winning matches 59% of the time. I also created a scoring system for myself that measures striking, creative and defensive actions. I am not the only one to do this, FMStag has also done something like this as well, and I think most people are going to end up along similar paths.

We are lacking another striking option up front, overplaying Morten Hansen is having its side effects. So we went out to look for a winger / inside forward. The metrics I used were dribbles / 90, goals, xG, CCC created by 90, fouls against, distance / 90. The goal here is to find someone who takes on defenders, and either draws fouls or gets inside the area to score.

Time for me to do some analysis, as the defensive actions formula was amended. I realized I was averaging it wrong. And don't forget you can also use the data hub.



Let me explain the spreadsheet really quickly. I have basically split Defensive Actions, Creative Actions and Scoring Actions separately. Far right I have calculated the non penalty xGs for goals. Essentially trying too find out what the goal output is not counting pens.

Defensive Actions = Clearances per 90, Interceptions / 90, Headers / 90 and Tackles / 90
Creative Actions = Passes / 90 divided by Chances Created per 90
Scoring Actions = Basically the ratio of non pen goals to xG

A player with good defensive actions only is generally playing a defend duty and / or a role that is not expected to create.
A player with good scoring actions and good creative actions could be a scoring threat like a mezzala, inside forward etc. It just says that the player can create chances and score
A player with good creative actions could be a playmaker, fullback that gets up or a winger that creates chances with.

I basically haven't finished this because I also track fouls against with crosses attempted or shots attempted to track players that break tiers, these could be inside forwards or wingers on attack duty or even wingbacks on attack duty, that I normally consider separately to identify different kinds of options.

For example: I could be looking for a inside forward, so I will look for good scoring actions, plus take-ons, e.g. dribbles / 90 and indications that he is drawing plenty of fouls. Currently I am evaluating strikers who can do a job as inside forwards so I am using this metric within my filters.

LAST SEASON
We were in the top 3 for quite a while last season but fell away in the final stretch because I became experimental again. I decided to mess around with my set pieces and it tossed me for a curve. Plus I lost sight of some fundamental basics.

Lorenzo Pirola and Saguesten were my standout defenders, however this season even after signing a new defender, I was hesitant to break up my defensive pairing because it was still good. I was too focused on how good Pirola was at winning key tackles. Now when I extracted the full numbers for the season, it painted a different picture.

In terms of defensive actions Sagstuen Andersen was 6.1, the new defender Hernandez was close in numbers to him but Pirola should have never been starting. He was winning fewer headers than the other 2 defenders. And like I mentioned in the last post there is a potential that the Hernandez/Sagstuen partnership could be the strongest partnership we have ever seen. Don't forget the data hub has great tools too.



Pirola's defensive output was decent but nowhere near them. Now returning to my spreadsheet, I am now going to evaluate midfield.



In midfield our holding mid Tarcisio is easily the best choice in terms of defensive performance, flanked by Casadei and Faticanti, who are both good all rounders. Casadei and Faticanti play as Mezzala and BBM in my system, but both need to work hard and win the ball back. Faticanti has defensive actions of around 2.35 and Casadei has even higher numbers which has surprised me. At 2.79, this now poses a problem for us heading into next season.

In terms of training now I need to focus, Raimondi and Dinho along training plans that work on their anticipation and concentration. We only have Christian Kouan coming off the bench to cover for them. Casadei, Rodriguez and Sagstuen are the oldest players in the starting 11(27,28,26) the rest of the squad is between 18-23

Raimondi has a bit of bad form in the last few games of the season and I plan to give him more game time and use him more consistently, he appears to have a good range of passing. As a RPM he will also be able to train his mental and physical attributes eg. concentration.

Dinho is by far our worst player, while he has great defensive actions, and has played well. 4 games are just too few. He needs to go out on loan.

Amerigo Bova is 23 playing as a left back, Holwerda is 18 and is also a left back. Holwerda has a 60 million transfer value plastered on his head because he is that good, but in terms of performances, Bova is putting better numbers in defensively. The choice is simple. Bova is first choice next season and Holwerda will continue learning from him.

When I look back at my season we were imperious at one point and brittle for almost 2 months. Identifying Pirola as the weak link via stats was eye opening.


GOING ATTRIBUTE-LESS


I do play without attributes, though I am not the only one who does this in the community. FMStag is a great source of information for a statistical deep dive into the moneyball side of the game. You can find stuff from him here Player Search Tool - Moneyball. I strongly recommend following him and his updates.

And of course Cult0fFM who has a Moneyball Series on his channel. He covers a lot on his series and it should be easy enough for anyone to follow.

I don't claim to be the authority on how to apply a moneyball approach to Football Manager, and believe me I will make mistakes, but so far I am enjoying this approach so much it really is the first time I am playing an edition of FM from one cycle to the next.

What FMstag offers is a search tool you can use, there are slight differences in our approach, but the essence is the same. My goal here is to help you understand the process of how one would apply data analytics to evaluating your squad and performances.

I hope I can make this easy to follow, so if you want more information you can check these sources out. The Cult of FM actually explains why he uses Team Conceded/90 vs Team Goals Scored/90 as well. And it's not hard to follow.



Let's get cracking.

For any numbers driven approach you will need benchmarks, you will need to set expectations. And you absolutely need to know the context in which you are analyzing that data.

When you are evaluating players you want to evaluate players playing at similar levels of competition. When you are looking for additions to your squad, you actually need to have a clear idea of what you want to achieve with your tactic.

I will walk you through the process of how I do that, starting with the tactic, then the sources of information I need in the game and finally how I present that information to myself so I can make informed decisions. Numbers don't lie, but it's easy to use numbers the wrong way and that can mislead you. So having a realistic approach is important.



This is the tactic and these are the metrics I want to pay attention to when analyzing my squad. In red are the scoring metrics for each group. I covered the metrics earlier and will cover that again in a video that I will link for those of you who prefer listening to my boring voice.

HOW WILL THE TACTIC PLAY

The tactic is meant to be played on normal tempo, but I want to be able to switch between low and fast tempo too. I want to draw teams into the middle where I will press them and tackle them hard through personal instructions I have given the 3 centre mids.

I expect to win the ball in transition, which basically means that I want the opposition attacks breaking down there, the mezzala and the box to box midfielder are expected to feed either the flanks via the fullbacks or find a way to play the ball into dangerous areas.

Occasionally I expect the mezzala to play the ball with the outside of his foot to find the Advanced forward on the right with dangerous passes. The box to box midfielder in some games is expected to arrive late into the box and in other games I expect him to be the sort who will provide extra protection down the flanks.

Naturally the left side of formation could be weaker, so I expect the left central defender to do help that area out with good reading of the game via a solid set of defensive actions.

In attack I expect to see the fullbacks, the mezzala and the attacking playmaker creating the most chances. The AP(A) will be the focal point of most of my attacks so he is expected to take players on with good dribbling.

I do expect him to break the opposition’s defensive midfield tier and sometimes arrive into the box either with or without the ball. This necessitates having good intelligence, on and off the ball. He will also be pitching in with goals as I expect him to be a free striker against certain formations.

The strikeforce could alternate in terms of roles, in some games I may want to change one role to a pressing forward so that we keep one defender occupied, this could lead to him pulling the defender away making space for the AP(A) to score or create even better goal scoring chances.

I do not expect to be getting most of my assists from crosses, in fact most of them should be coming from through balls.

When I make any tactic I have expectations and with those expectations I will have be doing several things. I will play the game on a highlights mode that allows me to see enough action to determine if things are going to plan. And I will be using statistics to measure the output of my players.




SETTING UP THE SAVE


Before jumping in on how you can assess your squad we need to first set up our save properly.

The game uses two match engines, the quick match engine and the full match engine. The QME is used in view only leagues and the FME is used in playable leagues. So it's important for consistency that you extract data from playable leagues only.

That's really the only requirement, if you want a stats driven game.


ASSESSING YOUR SQUAD


When you are playing without attributes there are only a few pieces of information you can use to assess your squad

  1. Height
  2. Ass Man assessment of current ability
  3. Ass Man assessment of current ability in selected role
  4. Ass Man opinion of this players style
  5. Scouts opinion of this players main strength
  6. Scouts opinion of this players main weakness
  7. Average Rating
  8. Last 5 games
  9. Playing Time
  10. Appearances
  11. Game Win Ratio
  12. Team Conceded per 90
  13. Team Scored per 90
  14. Player of the Match
  15. Assists
  16. Goals
  17. xG

I use these in a view like this to give me an instant understanding of my team. On my streams you probably have seen more information, but this is the very basic information I go for

Each player also has detailed coach reports that are very helpful. The coach report also contains pros and con and an assessment of his CA/PA.



When you start the game and if you play without attributes, your only source of information will be the data that comes from your staff, and you will need to depend on their assessments before match data starts coming in. There is plenty of information there already that can be a suitable starting point for anyone seeking to play without attributes.

Once the season is underway you will start collating the information within your views and you will start making assessments. To do that you will also need to look at the data hub because not all the data can be displayed in a view.

One of the most important I feel is possession lost and gained / 90minutes.



This is the possession gained and lost per game, and is accessed through the scatter groups for team data. You can also access these for players via the players search options. Unfortunately it doesn’t come as a filter in views, which is a real bummer.

Another scatter group which I like to refer to is the advanced goalkeeping numbers that measure how good your keeper is vs the rest of the league at expected goals prevented.



Why have I listed these sources of information? The moneyball approach usually works as a great starting point before you dive deeper into the numbers. These are all going to be my sources of information before I launch any further.

So I’ve set my expectations with the tactic, discussed why I will let the assistant manager and coaches play such a big part in influencing my decisions on who should play. Because I am playing without attributes, ultimately my decisions will rest on the little information I can gather from their coach and scout reports. At least until they have played around 5 games.

That's when the fun starts. But before then I need to depend on the Assessment view I use game to game.



Let me explain why I have chosen to display it like that.

Height - Who doesn't want to know height?
Scouts assessment of style - Quick way to have some kind of indication of where the attributes fall
Ass Man assessment of CA - This acts as a way for me to tell what his overall attribute level is.
Ass Man assessment of this players CA in the selected role - this helps to guide me when I have opted to play someone in a role completely unsuited to him. (SI's algorithm isn't half bad)
Scouts opinions of pros and cons - Frankly speaking there are NO attributes, I do not some quick warning on the players weaknesses. I usually go to the coach reports to get more detail and then map out an appropriate training schedule. And yes I have a specific way of training and no I don't use anything but my One Training Schedule to Rule Them All. It's balanced and it works.
Average Rating - Nuff said
Last 5 games - Haven't you heard of streaks, players hitting a vein of form. If a players last 5 games is > than Av Rating, he is on fire!
Playing Time - To prove to the whole world -- Look I ignored every players unhappiness levels, some I dumped into the reserves, we did it "live" see no unhappy people, well there is one, but he doesn't count. He failed to deliver consistently, his name is Tripledelli, only the viewers of my stream get the running joke.
Appearances

And the rest is what I pay attention to, GWIN, TCON/90,TGLS/90, POM, Assists Goals. This is the data I pay attention to when I am playing games. I am looking to ascertain which players add more to my attacks. Some players have notoriously bad team game winning ratios, like Alessio Holwerda, when compared to Raffaele Celia. This is stuff I should be paying attention to, but I wasn't doing that. This lesson I learnt came from hindsight. In future I promise to pay attention to this page when I am playing the game.



Now it's time for the moneyball data view.

This is the view I use to export all my data to spreadsheet. I hate FM, it's made me use an excel spreadsheet for a game again! This view probably has too much information and is still very much a work in progress. I have plans to use it to identify high performing wingers and inside forwards as well, but until I come up with a metric for them, I won't be sharing it just yet.

You might be wondering why do I have total crosses showing? It's not necessary right? Well the game does not differentiate between a cross from a corner and some set pieces which could be classed as corners. So whenever I see high key passes / 90 there is a chance that he could be taking a lot of set pieces




EXAMPLE: Search for an INSIDE FORWARD / STRIKER



Now we're on the hunt for more options upfront. So far this season we have done ok. I've had seasons where my striker banged in 100 goals a season, which frankly is nothing on Football Manager. Playing the game via stats makes it a bit more challenging and I do appreciate the effort it takes.

My spreadsheets are evolving, and the calculations keep changing. My metric for the IF includes me tracking like how many times they get fouled per 90 mins. It's usually an indication that he has some skill on the ball, and is willing to take on players. Other metrics included in the IF metric include Dribbles / 90, Non Pen Goals vs Non XG Goals.

Naturally I like my teams defending from the front, so I compiled all the defensive actions into its own metric. Once again mine could be different from others, but so far my conclusions from last season have borne themselves out.

As a team we concede far too many goals, we score a lot too, but it's a rock and roll kind of football. Entertaining, but I want to control more games and have the option to dial it up. To do that we sometimes need to attack teams from wider positions and drag them out.

The 4312 has been fantastic and so to has the Box system we are using, but I do so like Inside forwards attacking the centre with a blind run running through the middle. We have the players to pull it off. Last season we had an IF called Facchin who gave us that magic. He was on loan.

This season we need to strengthen our side so I filtered out any player with an average rating of less than 7, he must have played at least 5 games. He needed to have an XG/Goals ratio of 1, which basically means if he scored 1 goal he had an xG of 1. At least he meets expectations. We also wanted someone who could take on defences and find the cross/pass.

In my scoring system I have a
Striking Score which is basically the Non Pen Goal vs Non Pen XG
Creative Score which measures things like CHC C/90 vs Pass Completed /90
Defensive Score Defensive actions per 90

I also created a metric for IF and Wingers, different people may want different things so I doubt sharing this would be of any use. I am still testing this but I reckon I have most of these spot on.

I have filtered everyone by my IF Metric. I really want someone who can give me options either out wide as an IF or play as another striker.

The top 4 targets will be scouted
Zuluga appears to play more a supporting PF which explains his low scoring output
Zapata looks decent, but 20 appearances with that goal output probably another support player
Neither of them appear to be tackled as much as I'd like

Navarro and Leones look interesting, more appear to be better options for getting goals and they have a bit of creativity in them. Along with the two below them all 5 of the options after the first to create between 0.38-0.5 chances a game. Leones and Konta are interesting. Konta's scout report says that he sometimes doesn't go into challenge, which indicates low bravery. There is no mention of pace. He has consistency, strength, stamina and agility. His personality is sporting. There is a player further down the list my scouts are keen on Gabriel Pignataro. They have given him an A rating, I have made an offer for him. However I could cancel the offer.

We need someone to start, not someone who could be ready in 2 seasons. Pignataro has the advantage of having Italian as a 2nd nationality. At 4 million he could be a second signing.

We have already signed a central defender, unfortunately than was based on the previous spreadsheet where there was an error in one of the defensive actions metrics. It shouldn't be too big of a problem, since we really only signed him so we could send him out on loan again. Imagine his shock. Nevermind we have some Sicilians who can convince him to take my offer.



Once the scout reports come in I take a look at them where there are videos I watch them too. Konta is the first. While he is physically good, his lack of bravery may be an issue, he does use his left foot. In terms of CA he is some way behind my own players




The scouts seem to love Navarro, they must see a big price tag and go for it. Personally his CA might be good, his lack of jumping reach could be an issue. I do want my IF to bully the opposition fullbacks, he doesn't really fit my needs.




The numbers for Leones were good, he had 1.799 striking score, which is impressive for me when I usually think 1.3 is a sign of a good striker. He has jumping reach, but he doesn't really hound defences a lot. His tackling ratio is 0.38 which is probably the lowest amongst the group of players, this could come down to either Konta or Leone, but I need to wait for the rest of the scout reports to decide.

Both players will come in on their left foot, but Leone already has cuts inside as a trait and he can operate on both flanks, which does give me options. Leone will also adapt quickly as he will fit into the core group of players while Konta could take some time adjusting.





Exercise: Pick the best defender

If you had to choose the player most likely to finish as Best defender in the Serie A, who would you pick?

Would it be the first 2?


or


Well the first player is Alessandro Bastoni, Inter’s defender, the second defender is Merih Demiral from Juventus and the last defender was Tobias Anderson from Palermo.

The player who finished as MVP was Tobias Anderson the player with the worst attributes. I am revealing the attributes now, as I always wanted this to be a learning process for those who wanted to use statistics in the game. And without revealing at least one player, it'd be hard to get some perspective.

Don't worry, the rest of the players will remain a secret, till they find a new club or retire.



702751200_Awards2930.png.7b00fdf335ffb71e60b6b5f39b3ff386.png


ATTRIBUTES VS PERFORMANCE


This is the game of Football Manager. It isn’t a game entirely of attributes, and I am going to explain how he became the best defender. We applied statistics to our understanding of the game and we improved on our performances.

In doing so we improved how the team played across the board and thereby put us in the position to win the title. Yes we won the title with some distinctly below average players.

Tobias has been with the club for 10 years. He joined us as a wee lad at the age of 17. He made his first senior team appearance at 17 for Raufass in Norway and we immediately pounced on him because of the numbers he was generating.

So how did we go from a Europa finish to a title win, with largely the same players.

Last season wasn’t a bad one, we finished in Europa cup slots, if we had managed a better defensive performance we could have been in the top 4.



Our defence had done pretty well, our goals conceded was 1.16 vs and expected goals against per game of 1.33. That number told me that either our keeper was outstanding or our defence as a whole was. We set out to improve on those numbers in our 2nd season.



I have a spreadsheet that tracks data from the league. I group performances into 3 broad categories. A defensive score which tracks all defensive actions, a creative score which tracks the chances created by players and a scoring metric that measures how well players are finishing open play shots.

The defensive score showed that our top two defenders were scoring 4.9 which wasn’t bad. It was a good number that season. However in that season I was usually pairing Lorenzo Pirola with Toby Anderson. I had fallen victim to player feedback. Each player had told me that they liked the way Pirola organised his defence. However as a pair of defenders they were an inferior partnership to Ramiro Hernandez and Toby Anderson.

We needed changes so in the summer we secured the signature of Leones We needed tactical options if we wanted to beat the best sides or at least make it harder for us to lose. We had been using the 4312 and the 433DM sporadically. Next season we would need to play the 433DM against the top 5 sides if were to make it harder for them and we could also use that in Europe.

Since the 433DM was a new system we ended up playing with it more to gain tactical familiarity, plus we had no issues switching between that and the other formations because we were essentially using the same players in different positions.



Our second season was phenomenal by any standard. We won the title despite losing a creative stalwart to an eight month injury. Next season we need to sign another Guti because he wont’ be available till March.

Our tactical approach not only saw us lift the title but we put out some staggering numbers defensively.



We had improved on our defensive numbers. Our expected goals per game vs conceded had improved. Our goals per game had also improved.

I compared how our defenders had done against the other defenders in the league and I was surprised.



Only 3 defenders in the Serie A who had an average rating of 7 or higher put out a defensive score in excess of 5. Demiral was outside the top 20.

Ultimately attributes alone don’t make for a good defence, it's a collective effort. In goal our keeper was the second best in terms of expected goals prevented for the season.

He saved 3 penalties, had a save ratio of 83% and expected goals prevented of 11.72. He had an outstanding season and was instrumental in us winning the league title.

We used a statistical approach to improve our game. We did it by:

Identifying our weakness the season before - it was defensive
Coming up with a solution - We decided to strengthen our defensive play by using a formation that was also harder to beat.

There were some games where we had to become creative.

Towards the end of the season we lost our creative engine to a long term injury, and in the match against Bologna we also lost Tarcisio to a red card.



Even with the sending off or Tarcisio we ended up taking the game to Bologna with 5 attacking duties, but that wasn’t enough.

Fiorentina was one of the must win games of the season, with the title at stake, only a win would do, we actually shifted to a Houdini-variant which has been tested on the Total Tactics Tester as one of the strongest tactics in the game and pulled out a 1-0 win.



Tactically some small little changes were also made to my systems. We had to take the game to a lot of teams that wanted to sit back and defend. Our defenders have some of the best clearances in the league. In other words they are good at bringing the ball out of defence. So I opted to take a lot more risk with them by asking them to dribble more.

This would allow us to break tiers with the defenders, they would dribble up the pitch encouraging our midfielders to push higher, this would draw players into pressing them thereby giving us more space. Naturally it was a risky move, if our defenders failed we would be giving up easy goals. It was a risky move, but we needed to break defensive teams down even more. It was not a strategy I used against top teams.

Without knowing their statistics I would have never known to take this risk. In fact if I had played the game entirely by attributes, Tobias Sagsusten Anderson would have been transfer listed 3 seasons ago. Instead he has now captain and legend of the club. He has never given us less than the best and he has consistently been a top defender who gets better with age.



A video explaining everything including highlights of the changes I made can be watched below.



I hope you enjoyed the series.


Ending Thoughts


I think Football manager can improve the kind of statistics it provides. At present those who want to play the game via stats need to understand what the stats actually mean. For example, the difference between a key header and a header. Or a key pass and a pass.

For me it's not that big of an issue, but to go into the game expecting a range of statistics for a specific role may be challenging for some people. If someone were to ask me the difference in stats for a BBM and a Mezzala, I would probably be scratching my head.

It will depend on how you want to use them I have a simple method. I divide stats into 3 broad groups, and then I just apply the stats I need for a specific position. I am finding it quite easy at the moment. Naturally without attributes showing you raise the level of difficulty.

It's important to pick the right leagues, for example when you expect to be looking for a starter in a premiership side. There will be variances between the FME and the QME (view only vs playable), so you need to scout the right leagues.

Then you need to figure out if the player is playing at the desired level of competition. Currently I am looking for an IF and I have narrowed it down to 3 players, but one of the players is only 18, he's fantastic, the scouts love him and he is cheap, but I don't think he measures up to other players at top flight level.

This is the reason why for most people I would recommend using attributes, at least then the knowledge you develop by studying statistics only helps to cement your knowledge of the player.

Stam's avatar
About Stam

I started FM Scout for fun in the distant 2004. I'm proud of how this place has grown into a vibrant community and I try my best to improve it every year. Husband and father of two.

Your content on FM Scout

We are always looking for quality content creators, capable of producing insightful articles. Being published here means more exposure and recognition for you.

Do YOU have what it takes?
Like & Subscribe

Like our Facebook page & Subscribe to our YouTube channel!

Discussion: Moneyball Guide - The art of numbers, not attributes

1 comments have been posted so far.

  • JR's avatar
    Genuis as always from "the Master" that is Daljit / Rashidi aka BTM.
    I have been trying to work out how to use the Data since inception.
    Even before Data Hub.
    The light bulb in my brain would only flicker moments of inspiration but this article has now put it into focus & on full beam.
    This article will be game changing for all those who take the time to read it
FMS Chat

Stam
hey, just wanted to let you know that we have a fb style chat for our members. login or sign up to start chatting.