The Most Important Attributes for Lower Leagues.

This image taken below is from an excel spreadsheet i made just now. It is the top 20 players from each position, (sorted by avg rating) in the leagues skrill north and south.

I fished out all the attributes from the game using the print screen and imported it into excel. It doesnt take that long in all honesty, although it may appear that i did.

The data shows the most common attributes for these top players from each position.

It is interesting that at this lower level, specific attributes across all positions are very common and almost a necessity. Determination, Natural Fitness being the most surprising. It is also no surprise to see defenders with good jumping, heading, positioning and strong mental attributes.

I plan to build a team based on these key attributes by position. I will do this by making filters and searching for players of the exact quality required. (Hopefully they wont all be expensive!)

If there is interest then i can make these filters available. Or i can do this for more leagues to see the contrast in attributes as you go up playing levels.

Click to see the full breakdown of attributes.

Thank you for this!!

Thank you for this!!

Really solid work mate. It'd be interesting to compare this with the top flight leagues. Especially if you normalized it as well so that you compared the 'shape' of the attribute distributions rather than the 'scale'. I would really love to see whether fitness/teamwork/determination is as useful in the top flight.

That is interesting. As a lower league (Skrill North) manager myself it is interesting to see how much my squad concurs with this. Natural Fitness makes sense actually - with a very small squad including strong first-choice XI and weak backup, NF is important for players to recover between games, especially when there are two in a week.

Interesting stuff. I would have thought physical attributes would have been by far the most important at lower levels... My reading of the data is that physical and mental attributes are keys, rather than technical. Whereas at higher levels I would have thought technical would increase in importance due to nearly all top players having good physical stats.

More for different leagues please

When the patch is out I'm going to start a save with Cambridge United. I shall use this as a reference point and see where it takes me.

Would be handy to have each league up also, although I can assume it would take a bit, as would help with each step the club takes.

So which attributes are the most important, the greens or the reds?

Interesting stuff. I would have thought physical attributes would have been by far the most important at lower levels... My reading of the data is that physical and mental attributes are keys, rather than technical. Whereas at higher levels I would have thought technical would increase in importance due to nearly all top players having good physical stats.

That was the case with FM until recent editions. Time was, if you had a speedy striker you were guaranteed 30 goals per season regardless of other attributes, and a bit of muscle at the back was the only requirement. For Lower League management, the physical attributes were far too overweighted. Since then, SI have improved the balance so that whilst Physical atts are perhaps still prioritised over the others, mental and even technical attributes matter too.

I personally have created a club for Skrill North, an Academy consisting almost entirely of 16 year olds. They are bottom of the table in terms of strength, stamina and jumping, but on average are near the top on many mental and technical attributes. Whilst they do get barged off the ball all the time and conceded many late goals due to tiredness, they're holding their own in the play-off spots, so they're doing okay.

Really good work mate, always wanted to do something like this but never got round to it. Will definitely be using this tonight!

I post this information in the hope that it still holds good and I believe that it does. I refer to the final Championship Manager season 03/04 official strategy guide, the last guide to be released.

Players Attributes. Crucial.

Premiership, 16+.

Championship and League One, 14+.

League Two, 12+.

Conference [skrill], 10+.

There is also a similar guide for staff. It is thus.

Premiership, crucial, 16+, important, 12+.

Championship and League One, 14+ and 10+.

League Two and Conference [skrill], 12+ and 8+.

This has been my method for finding staff and players ever since and I have found that it works out rather well. However, if the aforementioned parameters have been changed or improved upon since, perhaps someone from SI might enlighten us all. I hope that this is of some assistance to you all.

Determination is no surprise for me. No matter what level I'm playing in, it's the single attribute I focus all my squad building on and I've been doing that in every CM/FM iteration.

The difference between high Determination and low Determination player is stark; their behaviour on the pitch, the eagerness to come back after being a goal or two down even without team talk, their focus in "easy" matches, i.e. avoiding complacency etc. It's much easier to manage a determined group of players so it's worth the patience needed to assemble such a squad.

So which attributes are the most important, the greens or the reds?

Green is more positive for the player position (the score 13 would be green in this example as it is a commonly high attribute among the highest average ratings for the position)

When the patch is out I'm going to start a save with Cambridge United. I shall use this as a reference point and see where it takes me.

Would be handy to have each league up also, although I can assume it would take a bit, as would help with each step the club takes.

I think there has been mostly positive feedback from this, and i hope it can continue.

I will post other leagues up from the conference north/south at some point and set up a landing page so you guys can see the stats in comparison.

I did make a filter for a DC and D L/R, however i dont know how useful it is because it returns 0 results!

It may be a better option to create a filter from the 5 most important attributes for the league/level you are in.

DC for example should have

13 Jump

12 Natural Fit

12 Head

12 Brav

12 Det

If any Defenders come up under this filter then you can pretty much guarantee as long as other attributes arent extremely poor (like 3 pace or 3 marking etc.) Then you will get a good performer. At this low level, these attributes are very common among the top performers so the importance of these is high.

I think there has been mostly positive feedback from this, and i hope it can continue.

I will post other leagues up from the conference north/south at some point and set up a landing page so you guys can see the stats in comparison.

I did make a filter for a DC and D L/R, however i dont know how useful it is because it returns 0 results!

It may be a better option to create a filter from the 5 most important attributes for the league/level you are in.

DC for example should have

13 Jump

12 Natural Fit

12 Head

12 Brav

12 Det

If any Defenders come up under this filter then you can pretty much guarantee as long as other attributes arent extremely poor (like 3 pace or 3 marking etc.) Then you will get a good performer. At this low level, these attributes are very common among the top performers so the importance of these is high.

The problem is that you'd need to scout the entire division before this would work, as most players have masked attributes (i.e they show up as a '-' until you scout them).

An unscouted player has 12 attributes hidden out of 32, then the probability of an attribute NOT being masked is = 20/32.

The probability of a player having all five of those attributes NOT masked = (20/32)^5 = 9.5%

Therefore if an ideal DC exists, but you haven't scouted him yet, there is only a 9.5% chance that he will show up in your filter.

-------

My solution to this is to filter by position (as you don't need to scout players to know their potential positions) and then filter by 1 (maybe 2) attributes at a time and look at all the players that come up manually to see if they fit the mould.

If you search for only 1 attribute at a time, the chance of the ideal unscouted player NOT showing up in a particular search = 12/32 = 37.5%, which is terrible.

BUT, if you search for each of these five attributes individually, the chance of the ideal unscouted player NOT showing up = (12/32)*(11/31)*(10/30)*(9/29)*(8/28) = 0.4%, which is great.

Therefore if you filter by 1 attribute at a time and then do a small bit of manual searching, the chance of your ideal DC slipping through the search is only 0.7%.

-----------

To make the comparison clear:

Chance of an ideal, unscouted player not showing up in a 5 attribute search = 90.5%

Chance of an ideal, unscouted player not showing up in five, 1 attribute searches = 0.4%

Obviously this doesn't matter if you have the facilities/time to scout everyone available, but more often than not your scouts will not be able to take a look at everyone. Of course, your filter is still useful as you can just turn each of the filters on/off easily

Low league players need to be physically strong and 10+ determination. That's all you need really for low leagues.

Low league players need to be physically strong and 10+ determination. That's all you need really for low leagues.

What.

The OP pretty much proved what stats do well in Skrill N/S using some solid research, and you come in here making these unsupported statements. In fact he pretty much showed that strength isn't an important attribute, so the first part of what you've said is just plain wrong.

You'd do well in journalism mate!

I am not into journalism. I am just a middle aged guy who enjoys playing FM.

OP showed concrete proof. I simply stated my experience. I know which is better but for low level leagues, I know what i am looking for from experience.

I am not into journalism. I am just a middle aged guy who enjoys playing FM.

OP showed concrete proof. I simply stated my experience. I know which is better but for low level leagues, I know what i am looking for from experience.

Well it looks like your experience is bang on regarding determination, but you should replace the 'strength' thought in your head with a 'work rate/teamwork/stamina/fitness' thought.

As long as they are a thug and have good determination, they've got the foundation to build upon. The rest can be done in training for years in my 2nd team.

The problem is that you'd need to scout the entire division before this would work, as most players have masked attributes (i.e they show up as a '-' until you scout them).

An unscouted player has 12 attributes hidden out of 32, then the probability of an attribute NOT being masked is = 20/32.

The probability of a player having all five of those attributes NOT masked = (20/32)^5 = 9.5%

Therefore if an ideal DC exists, but you haven't scouted him yet, there is only a 9.5% chance that he will show up in your filter.

-------

My solution to this is to filter by position (as you don't need to scout players to know their potential positions) and then filter by 1 (maybe 2) attributes at a time and look at all the players that come up manually to see if they fit the mould.

If you search for only 1 attribute at a time, the chance of the ideal unscouted player NOT showing up in a particular search = 12/32 = 37.5%, which is terrible.

BUT, if you search for each of these five attributes individually, the chance of the ideal unscouted player NOT showing up = (12/32)*(11/31)*(10/30)*(9/29)*(8/28) = 0.4%, which is great.

Therefore if you filter by 1 attribute at a time and then do a small bit of manual searching, the chance of your ideal DC slipping through the search is only 0.7%.

-----------

To make the comparison clear:

Chance of an ideal, unscouted player not showing up in a 5 attribute search = 90.5%

Chance of an ideal, unscouted player not showing up in five, 1 attribute searches = 0.4%

Obviously this doesn't matter if you have the facilities/time to scout everyone available, but more often than not your scouts will not be able to take a look at everyone. Of course, your filter is still useful as you can just turn each of the filters on/off easily

One easy answer to this, although it does mean you have to start a new game. "disable attribute masking" check this. I dont like the attribute masking but thats just me!

Quote Originally Posted by alucasa View Post

Low league players need to be physically strong and 10+ determination. That's all you need really for low leagues.

What.

The OP pretty much proved what stats do well in Skrill N/S using some solid research, and you come in here making these unsupported statements. In fact he pretty much showed that strength isn't an important attribute, so the first part of what you've said is just plain wrong.

You'd do well in journalism mate!

Lol great reply there. Couldn't have said it better. Sorry alucasa, but Foot is spot on.

(Apart from calling 'attributes' 'stats'. They are not stats. Its a pet hate of mine, sorry foot)

I am not saying OP or Foot is wrong.

But FM is a game and a game should not be played by only one mathematical method.

If you re-ran the same season about 250 times and got the data, you could easily get a good statistical model from Regression Analysis

(Apart from calling 'attributes' 'stats'. They are not stats. Its a pet hate of mine, sorry foot)

Yeah sorry mate, calling them 'stats' makes no sense at all. I must've had statistics on my mind.

If you re-ran the same season about 250 times and got the data, you could easily get a good statistical model from Regression Analysis

You'd probably end up with the exact same pool of 40 players making the top 20. I think you'd be better off taking data from 250 consecutive seasons, or taking 250 samples of a season in 2030, so that the game world has enough time to reach an equilibrium away from the start point.

Great work, look forward to seeing more if you end up doing so

Great work, look forward to seeing more if you end up doing so

thanks, will post some more stuff when i get a free moment.

Yeah sorry mate, calling them 'stats' makes no sense at all. I must've had statistics on my mind.

You'd probably end up with the exact same pool of 40 players making the top 20. I think you'd be better off taking data from 250 consecutive seasons, or taking 250 samples of a season in 2030, so that the game world has enough time to reach an equilibrium away from the start point.

On the contrary - 250 seasons consecutively would give poorer data - the initial conditions are roughly the same in the same season, so you'd easily control for variables like team-mate and manager quality. Of course given the ME it's likely to be a dynamic thing - dependent on the players existing in the division

This is a great thread, so many little gems.

It would be interesting to compare resualts in championship league and at either premier league or champions league. I would hesitate to say that the main stats will be there abouts, but it would be interesting.

This is a great thread, so many little gems.

It would be interesting to compare resualts in championship league and at either premier league or champions league. I would hesitate to say that the main stats will be there abouts, but it would be interesting.

Wednesday evening i will do Premier League.

Just to see if there is a big contrast. Then if there is i will fill in the gaps ;-)

What would be good to have additionally is something like the standard deviation for the average of each of the stats for each position. It will give you an idea of just how key a particular stat really is.

For example, if you have Heading 15 st dev 1 you know that this is a really key attribute to have high. If it was st dev 3, it is less important because players have a bigger spread in this attribute. I mean you could even go all out and find the standard error, which is not that much more effort if you are using excel or some similar software. The problem is you have such a small sample size that just a few really high attributes can skew your data.

Finally, I wonder about some positions, for example midfield, where you have many different roles which require different skill sets (DLP and BWM) for example. This is a good example because DLP do not have to be able to tackle well, BWM do. Thus, if you had an even mix of both roles, you would get a lower value for tackling than you may expect. Same for other attributes that are role specific. Some, like passing, I expect would be unaffected. I doubt you would want to break this down by role (I don't know how you would go about it, you would need to know in what role a player was played). But it is another thing to consider when analysing data.

Man, I never thought I would be talking data analysis on here.

On the contrary - 250 seasons consecutively would give poorer data - the initial conditions are roughly the same in the same season, so you'd easily control for variables like team-mate and manager quality. Of course given the ME it's likely to be a dynamic thing - dependent on the players existing in the division

I see what you mean. As in, it makes more sense to compare 250 iterations of the same season than to compare 250 iterations of 250 different seasons, as the comparison is more direct/controlled.

My thoughts were that the distribution of players to leagues/clubs might not be at an 'equilibrium' at the start of the game, and may therefore not adequately describe the way FM itself works. I thought that if you ran the game over 250 consecutive seasons that eventually the game engine would reach an equilibrium, as the influence of the 'human' starting point of 2013 would be long lost, and the dominant attributes would be dependent only upon the way the AI selects players, and the way the game generates new players. Of course, this is less useful if your plan is to dominate the game in 2013/2014. Instead it might just give some insights into how the game world progresses naturally, and whether the game has a natural tendency towards favoring certain attributes over others.

To be honest, I'd like to see both simulations done

Finally, I wonder about some positions, for example midfield, where you have many different roles which require different skill sets (DLP and BWM) for example. This is a good example because DLP do not have to be able to tackle well, BWM do. Thus, if you had an even mix of both roles, you would get a lower value for tackling than you may expect. Same for other attributes that are role specific. Some, like passing, I expect would be unaffected. I doubt you would want to break this down by role (I don't know how you would go about it, you would need to know in what role a player was played). But it is another thing to consider when analysing data.

You're right, the role itself is quite important. However the game doesn't give you the data on what role a player has been employed in for the entire season. The top-goalscorer in a league is only shown as a 'ST' rather than a 'TM', 'AF' or a 'F9', so I think it is impossible to break it down by role.

SkillfulSpence, how are you transferring the player stats from the gameworld into your excel spreadsheet. Are you using FMRTE, or are you typing them in manually?

I agree that the SD would be really useful.

Perhaps the OP could make his data public so others could take a look at it?

SkillfulSpence, how are you transferring the player stats from the gameworld into your excel spreadsheet. Are you using FMRTE, or are you typing them in manually?

I am using the print screen method. Print screen to txt file, then opening it up in excel as a .csv

you have to change the | character to a new row or column then it comes through nice and clean(ish)

doesnt take too long. I can do a league in about an hour, its just mind numbingly boring process with lots of the same mouse clicks etc.

I am using the print screen method. Print screen to txt file, then opening it up in excel as a .csv

you have to change the | character to a new row or column then it comes through nice and clean(ish)

doesnt take too long. I can do a league in about an hour, its just mind numbingly boring process with lots of the same mouse clicks etc.

I might be able to make a python script that can do it pretty fast. Will look into it tomorrow if I get some time after Uni.

I am using the print screen method. Print screen to txt file, then opening it up in excel as a .csv

you have to change the | character to a new row or column then it comes through nice and clean(ish)

doesnt take too long. I can do a league in about an hour, its just mind numbingly boring process with lots of the same mouse clicks etc.

How do you print screen to a txt file rather than to a jpg? Do you use 3rd party software, or is there a really simple hot-key combo built into windows that I don't know about? The fact that this feature could exist is blowing my mind.

Some good analysis there SkillfulSpence, I do love a good insight into FM with solid research to back up the theories. Shame there isn't a whole lot more going on.

Have you given some thought as to how the evolution of the match engine will impact on the 'Average rating' of players? I would imagine that as new updates are released, the match engine will undergo changes, thus altering the actions of players and thereby affecting their average ratings.

It's plausible that analysis done on an ME build several months old may no longer be relevant

How do you print screen to a txt file rather than to a jpg? Do you use 3rd party software, or is there a really simple hot-key combo built into windows that I don't know about? The fact that this feature could exist is blowing my mind.

Its one of the options in the game! You goto options menu in game and print screen!

See this video to show you how to do it. (in this video he doesnt really have a good reason to do it)

Some good analysis there SkillfulSpence, I do love a good insight into FM with solid research to back up the theories. Shame there isn't a whole lot more going on.

Have you given some thought as to how the evolution of the match engine will impact on the 'Average rating' of players? I would imagine that as new updates are released, the match engine will undergo changes, thus altering the actions of players and thereby affecting their average ratings.

It's plausible that analysis done on an ME build several months old may no longer be relevant

i did the same thing actually last season, and i must say the results are fairly similar to this years.

I am now about to dig it up and see if i have found it.

It is in a different format to this years one, as i have improved it since last year.

Here it is i found it!

Thanks SkillfulSpence. It might be interesting to run the same report out for the worst rated players in each position as well. Then take the bottom figures away from the top to see which attributes are making the biggest difference. Or maybe this is the same or similar to the standard deviation?

What would be good to have additionally is something like the standard deviation for the average of each of the stats for each position. It will give you an idea of just how key a particular stat really is.

For example, if you have Heading 15 st dev 1 you know that this is a really key attribute to have high. If it was st dev 3, it is less important because players have a bigger spread in this attribute. I mean you could even go all out and find the standard error, which is not that much more effort if you are using excel or some similar software. The problem is you have such a small sample size that just a few really high attributes can skew your data.

Wow what a great idea. This is a better way of looking at it for sure.

i have found this article and it looks like you can work out standard deviation relatively easily using the same range. I hope i kept all the data, i have a feeling i didnt so i may have to redo the table again. But no biggy.

Also, i think you mean Attributes and not stats right?

Its one of the options in the game! You goto options menu in game and print screen!

See this video to show you how to do it. (in this video he doesnt really have a good reason to do it)

You just blew my mind and made my day at the same time. I didn't know that this feature existed. Thanks a bunch!

How about one for the goalkeepers aswell :o great stuff anyway just made my three signings for York City based on this advice

How about one for the goalkeepers aswell :o great stuff anyway just made my three signings for York City based on this advice

glad to see this information being applied to someones game! Let us know how these players perform. York City are Conf National right?

Goalkeepers would be a good one to do also, however i think this may be weighted heavily to which team is capable of keeping more clean sheets, or are average ratings done by who makes the most saves in a game and therefore the bottom of the league GK has a higher chance of making clean sheets? Would be good to get someone elses thoughts on this?

To be fair, most SN/SS managers would know that when usually playing twice a week with very limited players you need the spine of your team to have great determination, fitness, stamina & work rate attributes. That's the same as you usually succeed if you can find a striker with high pace and decent finishing abilities, particularly alongside a strong target man.

York are league Two

If you re-ran the same season about 250 times and got the data, you could easily get a good statistical model from Regression Analysis

What do you suggest would be the dependent variable for the regression? I guess you could use average rating, but hmm... not sure if that would really yield much. Any suggestions?

And to the OP: Great work! It coincides with what I use as a "rough guideline" when looking for players. But you took it out of "unwritten obscurity" and into hard facts - I like it!

What do you suggest would be the dependent variable for the regression? I guess you could use average rating, but hmm... not sure if that would really yield much. Any suggestions?

And to the OP: Great work! It coincides with what I use as a "rough guideline" when looking for players. But you took it out of "unwritten obscurity" and into hard facts - I like it!

cheers Xinxin, im glad you found it useful. I am keen on doing other projects that can help us managers uncover secrets of the game that would be overlooked without an analysis of data.

I am also keen on the strategies in game that are played and finding the best counter strategies to these. See this thread and the excel doc, its a work in progress that i am keen on getting feedback on.

What do you suggest would be the dependent variable for the regression? I guess you could use average rating, but hmm... not sure if that would really yield much. Any suggestions?

And to the OP: Great work! It coincides with what I use as a "rough guideline" when looking for players. But you took it out of "unwritten obscurity" and into hard facts - I like it!

Avg rating is the most likely candidate, but you could do a Tobit model on Man of the Matches (although that wouldn't be much good I think), or if you had stupid amounts of data then possibly on the rating achieved in individual matches. Doubt that could work given the number of players on each side.

You could actually do panel data if you ran it for 10 years and followed a number of players, and then do a Monte Carlo-esk estimation/calculation from multiple 10-year runnings, which would give you the highest accuracy overall

Avg rating is the most likely candidate, but you could do a Tobit model on Man of the Matches (although that wouldn't be much good I think), or if you had stupid amounts of data then possibly on the rating achieved in individual matches. Doubt that could work given the number of players on each side.

You could actually do panel data if you ran it for 10 years and followed a number of players, and then do a Monte Carlo-esk estimation/calculation from multiple 10-year runnings, which would give you the highest accuracy overall

I think the standard deviation method would be the most useful set of info you could possibly have on this. As it would tell you exactly how important the attribute is in the average rating score. Taking into account the range of the attribute for each players position.

I dont see how useful a monte carlo estimation with this data would be. But i have never done such a calculation to be fair.

I understand that it can be used for predictions using simulations. And similary with regression you can use it to make predictions.. but we dont need to simulate anything as we have our simulator and we dont need to predict anything.

I may be wrong as i said i have not done regression or monte carlo analysis before, if you can give me more insight into how this would be useful then i would happily listen and implement if it is indeed useful.

The next step for me at this time is to do the same thing as linked above for the Prem league to see the contrast in attribute scores.

I will then do the same thing but using the standard deviation, which i think will be closer to what i am trying to present.

As i said though i am open to new ideas, and if i become to understand it would be more efficient to do it using a different method then i will happily listen

p.s. i have read about tobit model, but i dont understand this and i dont understand how it can be useful info for what we are trying to achieve!