clock menu more-arrow no yes

Filed under:

Who would be the best replacement for Willian and Pedro at Chelsea?

New, comments

Developing a full-fledged player ratings model to help compare various options

Getty Images

So, I’ve rewritten this article three or four times now. And the main reason is that I keep changing my mind on how to best utilize the data. Sometime early last week, I decided that I’m simply going to build a full-fledged model, in Excel, and make it available to everyone to use/tweak/debate/etc. My goal is to find the best right wing for Chelsea, to replace our current combination of Willian and Pedro, who are simply not getting it done in my not so humble opinion.

I don’t really care in the model whether a player is an inverted winger or traditional winger, there are pros and cons for each option. But what I do care about is winning. And players that win. And do good things. So I compiled a bunch of data, of the publicly available kind, as that’s what’s available to me, and I applied some factors, and some preferences, and gave things certain scores, to help build a little black box to find my ideal right side target. And I think for a first version it’s turned out pretty darned well.


The Scoring

Here’s how my scoring breaks down. First, I’ll show the high level group scoring, and then break each group down, with a bit of formulae for each. The Excel file is available via link HERE, so anything complicated will be left to the reader to dissect and either approve of or improve on.

There are seven major categories:

  • Shooting (and Scoring) - Up to 17 points
  • Creating - Up to 17 points
  • Dribbling - Up to 10 points
  • Passing - Up to 10 points
  • Defense - Up to 8 points
  • Fitness - Up to 8 points
  • Intangibles - Up to 30 points

These total 100 possible points, which makes things nice and easy to score. I included a total of 43 players in the model data, a bunch of players I thought were realistic targets, along with a few top level players (Messi, Neymar, Hazard, Sterling, Dembele, Mbappe, Bale) to help provide some sanity checking and comparisons.

Do keep in mind as even as great as someone like Messi is, and yes, he’s the top scorer (SPOILERS!) with the model, he still only got a 75. It’s a tough scale, and there are things some players are better at than others, but I think the scoring is fair.

Some of the data came from Transfermarkt.com, some of the data came from whoscored.com, some came from understat.com, and the FM attributes came from their new Scouting mobile app.

Let’s break down the individual sections next:

Shooting (And Scoring)

So this section is all about players being good at both scoring goals and shooting the ball well. Kind of important. As mentioned it’s worth 17 total points. And those 17 points are reached by the following individual scores:

  • Goals Per 90 - 3 points
  • xG Per 90 - 3.5 points
  • Shots Per Goal - 2.5 points
  • Shots Per xG - 3 points
  • Shots Off Target % - 2.5 points
  • FM Attributes (Finishing, Long Shots, Attack) - 2.5 points

For almost all of these scoring metrics, I’ve defined a minimum (floor) and a maximum (ceiling) to then derive the amount of points for. Most floors are zero, and most ceilings are just slightly above the highest performer for that metric. Some, such as the Shots Per Goal or the Shots Off Target, are scored with the lower score being better. I’ve given them weights based on what I think each metric is worth both in the section and in the overall score. But the beauty of this model is that all of these values are defined on a separate tab, so those wanting to tweak, can tweak to their heart’s content, without needing complex formula knowledge in Excel. Handy!

Creating

For me this is as important as Scoring, even though I know we direly need someone to score goals on that right side. This section is also worth 17 points, to show the obvious value. Here’s how it breaks down:

  • Assists Per 90 - 3 points
  • xA Per 90 - 3.5 points
  • xGBuildup Per 90 - 3 points
  • Key Passes Per 90 - 3.5 points
  • Crosses Per 90 - 1.5 points
  • Crossing Success % - 2.5 points

I felt these made up a good balance of chance creation skills, and should rate a player fairly on their ability to produce scoring opportunities.

Dribbling

We’ve been spoiled for a long time having Eden Hazard in our side, as he’s literally one of the best dribblers on the planet. Good enough that he’s often fouled, kicked, and bloodied every match. Yay, English Football. I do understand and realize, though, that there are good wing players that aren’t gifted dribblers. And this is okay. Not ideal, but okay. Here’s how the scoring breaks down for this 10 point section:

  • Dribbles Won Per 90 - 2.5 points
  • Dribble Success % - 2 points
  • Times Fouled Per 90 - 1.5 points
  • UT (Unsuccessful Touches) & D (Dispossessed) Per 90 - 2 points
  • FM Attributes (Technique, First Touch, Dribbling) - 2 points

I like to include UT&D because the few times you lose the ball, the better, obviously. Now, I do realize that some players have more touches, and more possessions, than other players, and you know who you are WAGNH users that will ask about this (is it possession normalized?), and unfortunately I don’t have the time to easily normalize this data based on possession. Not with what’s publicly available, at least. I know the cool kids, like Squawka and Statsbomb do this sort of thing, but I’ve got four kids. And no cool secret data streams. So it is what it is.

Passing

Any good system, be it one belonging to Sarri, or Conte, or Pep, or even Mourinho, is going to need wide players that pass the ball well. Interacting with fullbacks/wingbacks, recycling possession, linking up with forwards, it all involves being good at passing the ball. The true tragedy, though, is that finding good passing metrics at the season level is difficult. This is definitely a section that will improve with future versions. Here’s how the scoring for this 10 point section breaks down:

  • Long Balls Per 90 - 2.5 points
  • Long Ball Success % - 2.5 points
  • FM Passing Attribute - 2.5 points
  • FM Attributes (Vision, Decisions, Teamwork) - 2.5 points

Defense

Now, obviously this isn’t something critical, as no one scores really well here for the most part, but when players are good at this part it is definitely noticed. Here’s how the 8 points are scored:

  • Interceptions Per 90 - 1.5 points
  • Fouls Per 90 - 1.5 points
  • FM Defense Attribute - 1.5 points
  • FM Attributes (Tackling, Determination, Anticipation) - 3 points

Fitness

I’ve struggled at coming up with the right name for this section, at first calling it Endurance & Speed, but the reality is, for me it’s the fitness level. Speed is such a hard thing to measure, and I care as much about whether a player can actually last the season and play the minutes as I do whether he’s a speed burner (not everyone can be Mbappe, after all). Here’s how I break down the scorers in this 8 point section:

  • Minutes Played - 2 points
  • Appearances - 1 point
  • Starts - 3 points
  • FM Attributes (Acceleration, Pace, Stamina) - 2 points

In future versions, I want to try to go back 2-3 seasons on this, scaling back for each year, but taking into account long term longevity. But for now it’s just for the previous season.

Intangibles

Here’s my big one. It’s still ‘stat’ based, but it’s those things that aren’t solely on the pitch, but rather about the player, who he players for, where he plays, etc. All the stuff that’s tantamount to success but isn’t seen during the 90 minutes each weekend. Here’s how I break these down, in this whopper of a 30 point section:

  • League Strength - 7 points - This is based on my personal opinion of each league. I’m still tweaking these values in my head, as I write this.
  • Start Probability - 5 points - Is this player more likely to start for us? This is derived from a combination of starts the previous year plus a modification based on years played in Top Flight leagues. The more experience the better.
  • Winning Experience - 8 points - Another section where I’d like to go back a few years in future versions, but for now it’s solely based on how successful the club they were on last year was. Better table result, higher points.
  • Readiness - 5 points - I use the FM Current and Potential attributes to derive this.
  • Age - 5 points - This one is simple. My ideal target age is 24 for a player we sign. Score gets penalized in each direction the number of years away from 24, and more than 5 years off, you get a zero.

Results and Scores

Now, I’m sure most of you skipped the above, to read after you see the scores, disagree with them, and then need to identify exactly where I went wrong. As I mentioned above, Lionel Messi was our number one scorer with the model, followed by Neymar, and Eden Hazard. SHOCKING! Here’s how each player scored, with the sub-section scores as well:

Obviously, from that top list we aren’t going to sign Messi, or Neymar, or even Raheem Sterling, even though he’d be an amazing get for us. But for the rest of them....

  • Thauvin would be an amazing get in my opinion, and I’ve wanted him all summer. Seems to be zero rumor on his leaving Marseille, though, for some reason.
  • Draxler gets a bit of benefit by his play not only for a top club but also by playing CM a fair bit, which skews the metrics some. Good player, but the model is overvaluing him I think.
  • Depay is another guy that there hasn’t been much chatter about out of Ligue 1. He prefers the left side anyway, so wouldn’t be an ideal get even if we were interested.
  • Insigne is another good player but left side preferred.
  • Dembele is going nowhere, even though I’d love him at Chelsea.
  • Mahrez is going to City, so that’s done and dusted.
  • Bale is likely going to stay at Real Madrid, especially if CR7 leaves.
  • Asensio is likely not leaving Real Madrid either.
  • Bernardeschi isn’t going to leave Juventus, he’s the future there.
  • ADM I’d love to get, even if he’s old, but only if we also got a younger player as well. He’s a Willian upgrade IMO, and my model shows it.
  • Martial we probably can’t pry away from Mourinho, and he’s left side preferred.
  • Suso would be a fantastic steal from AC Milan. Underrated IMO.
  • Hey, look! It’s Willian! Good to know he is at least in the top half of this table. We know he doesn’t suck, we just need better.
  • Malcom is a player everyone loves, and even though he’s slightly underscored compared to Willian, he’s also YOUNG, and even though it scores low, it could be seen as a big benefit.
  • Thorgan, baby Hazard, is also an interesting consideration. Solid player, good scorer.
  • Correa, Lemar, Shaqiri, Anderson, Castillejo, Guedes, Brandt, Sisto are IMO all in a similar bucket depending on exactly what you are looking for. Likely none should start for us, though.
  • Mbappe, Double You Tee Eff, Over! He’s this low because he’s stupid-young, and because he didn’t play a full season last year. That’s all. If you’ve watched the World Cup at all, just skip him and how he breaks my model.
  • Philipp, Zaha, Lozano, Traore, Verdi, Bailey, Politano, Gnabry, Williams are again, in a similar ‘bargain bin’ type bucket, depending on what kind of depth you are looking for. None would be starters for us though, unless we’ve given up.
  • Pulisic. Oh, sweet young American prince, your time will come. Much like Mbappe, he’s heaping with potential, and that’s about it right now. Better to let him sit in some nice German casks for a little bit longer and age gracefully.
  • The rest? Meh. Neres, Berardi, Oyarzabal, Chiesa, Richarlison. Fool me once, shame on you.

The model

I know you all want to see what it looks like, while you wait for that whopping 1MB download over your dialup connections. Oh, what’s that? You have fiber-optic connections? And laugh at a 1MB download? Okay. I hear you. Here’s what it looks like anyway.

As always, please feel free to contribute feedback, suggestions, and further ideas in the comments section!