Searching the Stats

He’s made a name for himself as a Football Manager fiction writer, but here on PNIG, Chris Darwen turns his hand to how stats can be used in the game. 

Friday, 8 May 2015
20:28
This is what happens when a story-teller starts to try and understand the game a little more. This is what happens when I start to read a book and try and apply it to the real world of FM. This is what happens when all the little things that I have thought about for the last 20 years of football management gaming all falls into place.

I’ll say it loud and I’ll say it proud – “I DON”T CARE ABOUT ATTRIBUTE STATS!”

I mean it, I don’t care. I try to recruit scouts I trust, but more importantly I now look at the way a player has performed and try and use performance stats to find value.

Don’t misunderstand me, I am not saying that attribute stats do not matter at all – but look at it this way. If you only have a small transfer and wage budget you are unlikely to be able to entice the young striker who has 20 for finishing, 18 for pace and 20 for flair. He is likely to cost a hell of a lot of money. My system will definitely help you find some bargains, I am sure of it.

To give my explanation of a system I need to explain a little bit more about how I have played the game over the years.

Back on the first ever CM’s I never once used to look at the attributes. Never. I mean it, never. What I did was sign players that had high average ratings in the league I was in or below that seemed to be good value for money – basically that I could afford them. It did not fail me and I built amazing sides of young players with potential and performance at other sides and turned them into European Champions time and time again. I think I have always preferred to play this way and now, on FM15, I am still yet to actually pay a transfer fee. Admittedly, I exclusively start saves with teams at a lower level so I’m rarely a professional club let alone have a transfer budget but I consistently win promotion after promotion and find absolute diamonds in the rough.

Next, I think you will need to understand my system of play a little more so you can understand why particular performance stats are important to me.

formation (1)

mentality (1)

team instructions (1)

We play a 3-3-1-3 formation, no width. The GK is a SW/D, the back three are all CD/D, the midfield three line up from right to left as a RCM/A, CM/D, LCM/A, an Enganche between the lines and then two F9’s flanking a Trequartista.

The style is control, very fluid so I am looking for possession with a purpose.

Why this particular way of playing? In about mid-January I decided I wanted to start playing the game like I used to play Champ Man, which meant this style of formation which won me everything and this system of recruiting players, though nowadays with FM you have loads more data to play with.

Each time I start a save my aims are simple:

  • Only sign players on free transfers
  • These players must be U23 on the day of signing
  • If anyone makes an offer that exceeds the value of a player, sell them
  • Always have a replacement lined up for players that are wanted by other clubs – this replacement, ideally, will come from the U18’s
  • Always replace a player with a better player if he becomes available and fits the criteriaThis means play recruitment and identification is crucial
  • On my current Guildford City save I feel have tweaked my recruitment system enough to finally share it with people. It is not the finished article and I am sure many of you will have questions and see potential flaws, which using the wisdom of the crowds should help me refine it and improve it in time for FM16.

So, what do I look for when recruiting?

I’ll try and break it down into what I don’t look for and what I do look for – and very little of this is about their attributes, it is all on their performance.

Goalkeepers

Firstly look at my two goalkeepers attribute stats – I am sure they don’t impress you that much.

Rob Brown (1)

Aaron Moody (1)

But remember, to me their attributes are not that important.

A lot of people when thinking about what makes a good goalkeeper performance wise would probably point to shots saved and clean sheets. I don’t look at shots saved, clean sheets in an isolated fashion. If the ‘keeper has made 50 saves or kept 72 clean sheets it means nothing to me, as he might be playing behind a top defence. Equally, if he has let a shed load of goals in it might be because he is playing behind a terrible defence. I do look at these stats, but then I look at where they are in the league. If they are low in the league and he has kept clean sheets, this is good. If they are high in the league and he has saved many shots, this is also good.

I am interested mainly in their average rating, because in a good or bad team a ‘keepers average rating can always be high, I look at their “mistakes leading to goals” because the best ‘keepers make fewest mistakes and I look at their pass complete % as I want good distribution from the back. The best ‘keepers make the fewest mistakes, and if the ‘keeper has a high % completion rate then he is probably used to playing out from the back. Both my ‘keepers do well in these key areas. Between them they have made one mistake leading to a goal this season, average a shade over 7.00, complete 18.75 passes per game at a % of 86.5%.

How does this compare to a real life goalkeeper?

We are top of the table, so when I flick into the real world to compare I go to Courtois at Chelsea. I cannot find a seasonal comparison on Squawka but it does tell me that he has made zero mistakes leading to a goal. Courtouis has completed 48% of his passes, but to be fair Chelsea prefer to be a little more direct from the back which is something I will need to take into account when recruiting.

In summary – when recruiting (on a sample of 15+ games) I am looking for a ‘keeper that has played a high number of games, made very, very few mistakes and has a passing completion % as high as possible. All ‘keepers should be able to save shots, I just don’t want mine making any mistakes! Finally, I would prefer them to have an average rating of as close to 7.00 as possible.

So let’s look at my two goalkeepers and the two that I am considering signing for next season.

goalkeeper squad (1)

As you can see, Moody and Browne have shared the duties this season but their overall performances have been very similar. It is unlikely that I will find goalkeepers that match their passing stats as we are one of the few teams that look to play it out from the back, but I am interested in mistakes leading to goals and average ratings in the two I am considering signing.

goalkeeper shortlist (1)

So the two lads on the shortlist were the only two goalkeepers known to me that had played that many games, made no mistakes and had an average rating of 6.7+. You can also see that my scouts feel they both have potential, so could well be worthwhile signings. If I felt the stats matched what I want then I would over-rule my scout reports, but it feels better if the scouts agree. I will approach these guys and the one who wants the better deal from my point of view will end up being signed this summer, without even checking their attributes.

Defenders

I’ll show you two of my defenders at random.

Schroeder (1)

Pringle (1)

In fairness, I am really impressed at how we have developed Pringle’s attributes – he did not look like that when I signed him.

With the defenders, I don’t look at tackles won. If the player has had to make a tackle, he was probably in the wrong place to begin with. If he has made lots of tackles it might be that the opposition have decided to attack him as they see him as a weak link in the side, therefore meaning he is under pressure to make more tackles. I am more interested in the tackles completed ratio, as this gives a fair reflection of how they have done.

I am interested in aerial duels won, as I want my back three to dominate in the air allowing us to recycle the ball a lot more, and I am interested in interceptions. Because my back three are quite often left exposed, they need to be able to read the game pretty well. I am also interested in their mistakes leading to goals, total completed passes per game and their pass completion %.

Let’s take Nicky Schroeder, who plays in the middle of the back three, and I will compare him to John Terry, again using Squawka. Nicky has completed 90% of his tackles, wins 6.87 headers a game, has got 208 interceptions in 15 games (average of 14 per game) and completes 22.47 passes a game at a % success rate of 60% – compare this to JT who completes 87% of his tackles, wins 2.63 headers a game, averages 0.83 interceptions a game, completes 41.26 passes a game at a completion rate of 89%. What does this tell me? That Nicky is excellent defensively, but his passing game needs to improve.

In summary – when recruiting (on a sample of 15+ games) I am looking for a defender that has the most interceptions, over 5 completed aerial duels per match and very few mistakes (ideally less than 1 per 20 matches). The passing part has to be taken in the context of the leagues I am playing in – very, very few teams play from the back at this level according to the AI. All my defenders are averaging 20+ completed passes and around 55-60% completion and we like to pass the ball. Therefore when recruiting I am looking for players that are around those stats, ideally. As close to possible, but the interceptions, aerial duels and mistakes are non-negotiable.

So let’s look first at my current defenders then the ones I have shortlisted for next season.

defender squad (1)

The first stat that stands out for me with my current players is mistakes leading to goals. We seem to make a lot of mistakes. I am using my players here to help me form some benchmarks for recruitment, which fits into my summary above.

Now, I will probably only be looking to sign one defender from the group below.

defender shortlist (1)

Welch would probably be ruled out due to low average rating and passing completion %. He got on the shortlist due to interceptions, aerial wins and passes completed being some of the higher stats out there.

Ben Carter has a lower passing % figure, but I’d still be interested in him due to his aerial duels and incredibly high amount of interceptions considering he has only played 18 games.

My favourite on here though, after a lot of thought, is Courtney Fern. Second highest amount of completed passes, decent completion %, good interception number and over 5 for aerial duels won. His average rating is also decent. In second place would be George Artemi, purely because he has a smaller sample in just 16 appearances. I will move for those two first of all.

Midfield

Again, I’ll show you two random midfielder’s attributes.

Blake (1)

Whyte (1)

I’m sure you’d agree, nothing too impressive there.

In midfield it is all about possession with purpose, not just keeping the ball but what do we do with it. So I am interested in total passes per 90, pass completion, key passes. If I am looking for a player that is playing the defensive role then I am also looking at interceptions and aerial duels. To cover the full range, I will look at my CM/D, Blake, and compare him to Matic.

Matic, who many consider the best defensive midfielder this season, passes the ball 62 times per game, at a completion rate of 86%, with 0.62 key passes per game. He has 2.16 interceptions and 2.26 headers. Take a look at my midfielders below, in particular Blake who plays the same role as Matic.

midfielder squad (1)

In summary – when recruiting (on a sample of 15+ games) I am looking for a CM/D or enganche that passes the ball over 65 times per game at a completion rate of 80%+ with 3 key passes per game. For my CM/A I will accept 45+ passes per game at a completion rate of 80% with 3 key passes per game IF they are getting 0.8 shots on target per game at around 40% accuracy.

Let’s have a look at the shortlist.

midfielder shortlist (1)

Both the lads there are close to the passes per match level, near the % required but are way down on the key passes per game and shooting %’s. One is getting 0.8 shots on target per game. If I decide I need to bring in a midfielder, I may decide to make a sacrifice here. Neither of them could provide competition for Blakey, they would be CM/A’s.

AM shortlist (1)

The extra lad here could do me a job as an Enganche. He misses the total passes but beats the % and key passes.

Attack

I have to show you Celal Osman. When I signed him he was going to be my third choice AM, maybe play him midfield. Then we had an injury up top and as I was running a tiny squad he was my only option so I gave him a chance.

Ossie (1)

Ossie history (1)

He has possibly the worst striker attributes I have ever seen, nothing about him works. But he has scored all those goals. Wait and see how.

For the lads up front I am interested in how often they hit the target and how well they pass the ball. To create a goal it will need a good pass, and then once they are in position, how well do they do? So that stats I look at are shots per 90, shots on target per 90, shots on target ratio, then all the passing ones from above.

strikers squad (1)

Let’s look at Ossie. Ossie takes 3.42 shots per 90, 1.88 of those are on target which is 55% – his form has dipped a bit recently to be fair. Passing wise, 27 completed per game, 2.93 key per game, 84% accuracy, which feels good. But let us compare him to Harry Kane. Kane shoots 3.2 times per game, 1.88 of which are on target with an accuracy of 59%. Ossie matches up pretty favourably to this. Kane passes the ball 25 times per game, 0.74 of which are key and has a completion of 75%. Ossie owns him on this.

In summary – when recruiting (on a sample of 15+ games) I am looking for strikers that get 2 shots on target per game at an accuracy of not less than 50% ideally. Pass completion should be in the mid 80%’s and, ideally, 28+ passes per game.

Who is on the shortlist?

striker shortlist (1)

The lad at the top, Mendy, has a very small sample so I would probably have to rule him out but he is getting 3 shots on target per 90 currently in the substitute appearances he is making.

Realistically, the top two contenders here are Harmsworth to play as a Treq, due to the fact he gets 1.55 shots on target a game at a 50% rate. His passing concerns me though. The second choice would be would be Ben Pope, perfect on the passing stats and just under on the shooting so could be an F9 option.

So in summary. They are the stats I feel are important to how we play, this may not the same for you. I look at my players performance vs real life excellence to see how we compare. If we are over the real life excellence then we are good and I use that as a bench mark for recruiting my next players. If we are below that real life bench mark then I need to be recruiting. When comparing, I try and look at my team’s style, league position and performance and compare like for like in the Premier League – normally not just with one player, with a few, just to make sure I am getting a decent average. (I have only shown one real life comparison here as a quick example).

Now, when it comes to recruiting – players with the best attributes etc etc will always cost more. If I can find players that may not have the best stats but have consistently hit the

numbers I am looking for in their performance, then there is a chance they maybe cheaper AND better than bringing in someone more obvious. Now, at my level I am rarely paying a transfer fee – in fact on FM15 I am yet to pay a transfer fee, but using this system I am often able to find real value in the wages I pay as well.

This has worked for me, and in the words of Graeme Souness “show me your medals”

  • Clapton FC (promoted from Essex Senior League to Conference in successive seasons)
  • FC Carl Zeiss Jena (promoted from 4th tier to Bundesliga in successive seasons, thenrelegated in first season – total wage spent per week in Bundesliga £68,000)
  • Padova (reformed club in bottom tier of Italian football – 4 successive promotions upto Serie D)
  • HFC Haarlem (reformed club in 4th tier of Dutch football, promoted to Eredivisie in 4seasons, database corrupted)
  • Guildford City (started in 9th tier, back to back promotions see me in the RymanPremier next season)The eagle eyed will notice that I have only played one season in the top flight and got relegated, so if anybody wants to try this method of recruitment at a higher level I would be fascinated in their feedback.Any questions, feel free to ask! Chris

You can find Chris at his blog as well 

Advertisements

25 comments

  1. Jeromie · · Reply

    How do you find the players to shortlist and when do you go out and find them

    1. afhstewart · · Reply

      Chris will be scouting in the normal way and using attribute searches as well.

  2. I change the player view to the filter I’ve created (above). Then I choose the position that I am looking for, see who is top of the charts and send my scouts to watch them. I also have my scouts searching more generally for talented U23’s, and I then look at their performance %’s as above before deciding whether to move for them.

    1. Jeromie · · Reply

      Thank you so much

  3. Nice. Very nice read.

  4. In order to achieve the remarkable levels of success outlined here- is it essential to use this very narrow, European style formation or could such success be emulated with a formation that uses more width? My concern is going against a side that has good players on the wings that would take advantage of the narrow formation in the midfield and defense.

    1. No, the tactic is just my choice. You can apply this recruitment system with any tactic, but might change your % benchmarks depending on how you want to play – for example you might not care as much about passes completed if you play more direct.

  5. […] creating role specific shortlists for first team players and youth players using ideas gleaned from this excellent article by Chris Darwen (@comeontheoviedo) and I strongly suggest that you give it a read […]

  6. Do you find limitations in leagues you don’t have loaded with regards to getting enough stats?

    1. Chris · · Reply

      To be honest, no – but I can see how this could be a problem.

      Most of my saves have been starting at a lower level so I am really only looking at the players in and around me at the same kind of level. My biggest limitation so far has been in the first six months, where there is no data to work with. Mind you, I have never been a make drastic changes in the first six months kind of guy anyway.

      1. I can see how it wouldn’t be much of an issue for you. Thanks for the reply.

  7. The system bombed for me, played 5 matches and got hammered in each by 3+ goals each time? maybe it’s the team I have

    1. By system do you mean the tactic, or the recruitment method?

  8. […] “He’s made a name for himself as a Football Manager fiction writer, but here on PNIG, Chris Darwen turns his hand to how stats can be used in the game. Friday, 8 May 2015. 20:28. This is what happens when a story-teller starts to try and understand the game a little more. This is what happens when I start to read a book and try and apply it to the real world of FM. This is what happens when all the little things that I have thought about for the last 20 years of football management gaming all falls into place.” Put Niels In Goal […]

  9. Would you be able to give some hints into key statistical targets required for recruiting wide players?

    1. That is a fine question, and something I am dabbling with in a new save. My advice would be based around how you play and what the most important thing to you is in that position. For me and a full back, it is pass completion % and interceptions so I would filter my scout reports and work from the top down on both of those. In a tactic I’ve used playing defensive wingers it was the same two, but I was a little more interested in how many passes per 90 minutes. For out and out wingers it could be anything, depending on what you want from them.

      1. Well I will be rotating between an adaption of your (Bielsa’s) 3-3-1-3, a Barca inspired 4-3-3 where cutting in IF’s are vital to punish space created by a F9 and a high pressure 4-2-3-1. So in both formations I require width the roles would be IF for both sides and probably wing backs to sit just behind them. I just think these three formations and the players required for each mesh together rather well and can be used in rotation to overcome different opposition systems- where I fall short, and always have is player recruitment and boggling my mind looking for players that fit – so much so I actually don’t enjoy saves long as a result. Just hoped you would have some kind of efficient moneyball system for finding decent wide players.

  10. Hi, do you have to play 10-15 games before you can start the search for players?

    1. They need to have played 10-15 games for them to have enough stats to provide a wide enough sample

  11. This system make sense on a lower league (and in the real life since you can’t see the player’s attribute stats) where you have many good option for a free transfer, because the players have short term contracts in these leagues.
    I don’t like the idea of looking at passing stats, it’s more related to the team’s strategy than to the player’s quality, you search for players with a high pass complete %, because you play with a possession based tactic, but this way you can lose some good players that are below your requirement just because they play in teams that use different tactics or maybe because they play out of his natural position.
    I do like the idea of looking at aerial dueling stats, it’s a easier way to see if the player it’s a good header (I used to take the average between strength, jumping, heading besides his height) and isn’t related to the way he played if you take the percentage of success.

    1. Chris Darwen · · Reply

      The system is also working at the top level now, in a series I will be publishing at the end of the month. The point of this is a guide – the benchmarks I set for my players may not be the benchmarks you set for yours as we might set our teams up in different ways. The points remains the same and works though – you can find better value in the market by looking at performance stats as opposed to attribute stats. If your team does not play a possession based game, what stats could you look at instead? My new save involves wide players, so I have had to consider different factors, and the system still comes good at finding the right players at the right price.

  12. […] right central defender search was inspired by this article from Chris Darwen. Putting it into practice, I found several candidates whom I felt would fit the […]

  13. […] blew me away. I was lucky enough to host pieces by Chris Turner (Welling United) and Chris Darwen (stats-based recruitment) on my own blog, putnielsingoal.com, and there was also a great series on Moneyball in CM01/02 by […]

  14. […] of my nominator Alex Stewart, but I have tried to take it to the next level by incorporating a real Moneyball/Soccernomics angle on how to find […]

  15. […] Darwen: „Searching the Stats“ […]

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: