Introduction
Performance statistics in football can be illuminating at best, misleading at worst and potentially meaningless in the middle. I’ve previously written about my feelings towards the use of metrics and data analysis in football and FM here.
Spoiler alert – I find them fascinating and incredibly useful, but only when discussed and applied in ways that make logical sense.
Fellow blogger Steinkelsson recently published an excellent article called ‘Building Baselines,’ a method of analysis from his Villarreal save. This, in combination with FM Eadster‘s tweet embedded above, gave me the idea for this article. Please read Steinkelsson‘s article and check out Eadie’s new ‘Journey Person’ series before reading any further. They are both incredibly worthy of your time.
What is this all about?
Baselines and benchmarks are different things, but effectively what we are interested in here is finding a starting point for comparing statistical performances in Football Manager. Or in other words, defining what good looks like. Generally speaking, we would all agree that 15 for the Finishing attribute is good, but what is objectively a “good” number for Shots on Target per 90 for a striker? Two a game? 12 a game? A Tackling attribute of 17 for a central defender we would all agree is excellent, but is 2.17 Tackles Completed per 90 any good? How could you possibly know? This is what I’m aiming to answer.
Now, I must stress that FM calculates performance statistics differently from real life football. This is because although Football Manager is a wonderful football simulator, it’s a computer game, believe it or not. Various designers and senior figures within Sports Interactive over the years have commented on this multiple times, not just on the SI forums but on various podcasts and in numerous interviews.
In this article I aim to answer Eadie’s question “Ooo 4 dribbles per 90, is that good?” and many others like it by providing tables suggesting what good looks like across all of the key performance metrics in FM23 for each position on the pitch.
I have done this by taking pages and pages of performance data exported from multiple end of season save files in FM23 and calculated Low, Medium and High performance parameters for each metric related to each segment of the pitch, like central defence, wide midfield etc.
Disclaimer – I understand that sometimes you want different things from the norm for your players in their chosen positions depending on your tactical system. i.e. Where I’ve made the assumption that a “good” full-back will have a high number of crosses completed per 90, I appreciate that your personal tactic may play with Inverted Wing Backs who drift inside or No-Nonsense Full Backs who sit back and don’t contribute offensively. Similarly, you may want your Goalkeeper to push out and sweep, making interceptions and then delivering accurate passes upfield. Other FMers may want their goalkeeper to shot-stop and play risk-free passes. I can’t account for every tactical choice and role preference, but the broad strokes of this should be useful.
There are some metrics, like Save Ratio % and xG per shot that I’d have liked to have included, but they unfortunately don’t currently calculate as intended in FM23. Also, there’s the consideration that a high number of tackles per 90 (for example) doesn’t always denote a “better” defender than one with a low number of tackles per 90, perhaps that he plays for a lesser team who have to defend more. Until we include possession adjusted figures for defensive metrics (a future project?), this analysis will always be subjective. The numbers can tell any story you want them to. Use them as you desire. I’m just doing the hours of initial mathematics, so you don’t have to.
For each position/role I’ve created a table. I haven’t covered every role in the game (as that would be bonkers) but I’ve split the roles out as logically as I can.
This is not an exact science, but instead meant to take a screen like the one below which is just a sea of numbers, and turn it into something which gives you a better idea of what those numbers actually mean when comparing player performances.
What the hell does “good” look like?
Show me the money metrics
After many, many hours of collating numbers, here is the result.
The most practical applications of this are when scouting targets, evaluating your own players and comparing the two.
Example – I want a ball-playing centre-back in the summer window and I’ve found one who has 1.8 tackles per 90, 4.3 clearances per 90 and 6.4 progressive passes per 90, is that good? A quick glance at the sheet and it tells you that compared to elite players, this player performs medium to high for tackles, high for clearances and very high for progressive passing. The numbers are backing up your gut feeling that the player is a good find.
Example 2 – I’d like a goalscoring striker but the target I have in mind only has 24% for shot conversion rate and only 0.6 non-penalty xG per 90, despite the fact I can see he has a high Finishing attribute. Will he perform as well as I hope he will? While there are no guarantees, a look at the sheet shows that 24% shot conversion is actually very high, and 0.6 nPxG per 90 for a goalscoring striker makes him elite at getting into high opportunity positions before taking a shot. The numbers again back you up!
I could go on, but I’m sure you get the idea of the kinds of question these numbers can help answer for you.
How did I do it?
I have three end of season save files from different save universes. In each one, I ran a Player Search for players who had played a minimum of 2,000 minutes and played for sides in the “big five” European leagues – the top tier in England, France, Germany, Italy and Spain.
To capture the right kind of players in the results, I ensured that players were natural in the required position. To fulfil the DNA required to match descriptions like “midfielder – destroyer” or “wide attacker – provider” I included a filter for a couple of key attributes to ensure the results held the closest match to the profiles of players who fit the description. Like a midfield destroyer had to be high in Aggression and Tackling. A wide attacking provider high in Dribbling, Vision and Crossing, for example.
I then took a cut of five players from the top, middle and bottom of the pool of results for each of the statistics in question, then calculated an average for each. Once I had an average Low, Medium and High number for every field, I then repeated the same work for the other two save files, and averaged those results to ensure it was a fair and even spread, mitigating anything unique to a specific save universe.
For those of you who like data analytics or dare I say it, the “moneyball” approach in Football Manager, hopefully the sheet will help guide your decision making when faced with endless columns of seemingly baseless performance statistics.
This work analysed elite players and while you can and should benchmark your players against these numbers while playing at any level, there may be value in me repeating this analysis but with lower-league clubs and lesser players, to give you a different sheet of comparison points for your analysis. Would that be worthwhile? Let me know on Twitter.
Thanks to FM Tahiti for being the sounding board on my selection of key metrics for each role before I got stuck into the calculations.
Thanks for reading.
FM Stag