Bordeaux and the Chase for the Champions League

2017-18 in Ligue 1 was never going to be about a title race in France, because that was sewn up the minute PSG bought Neymar from Barcelona (and just to rub it in, they got maybe the best prospect in world football as well). Rather, where the intrigue in Ligue 1 came from was the cluster of six or so teams below PSG fighting for two Champions League spots. At least for this writer, there was a genuine curiosity about how the standings would shake out in positions 2 to 7 considering the massive changes that had gone on. After seven games where are we? Monaco have been fine despite selling the majority of their title winning squad, Lyon have produced attacking numbers that are quite middling compared to the talent at their disposal, while Marcelo Bielsa and Lille are Ligue 1’s travelling circus act. Bordeaux were one of the teams that were mentioned in that block of six, an intriguing side that could have a high ceiling because of the attacking talent gathered over the past few windows. They are led by a respectable manager in Jocelyn Gourvennec, and on a mandate of sustainability by buying unheralded talents either inside the league or elsewhere. It’s allowed them to get a player like Nicolas De Preville, one of Ligue 1’s most underrated attackers over the past 3-4 years. Younousse Sankhare is a talented midfielder who can press individuals with his athleticism while providing off ball runs in the final third. Alexandre Mendy came in the summer for under €1M despite putting up relatively monster xG numbers at Guingamp (albeit nearly 50% of those minutes came as a substitute). This has been a carefully built squad on a tighter budget than others in Ligue 1. It should be emphasized just how deep this team is in attack for a club that’s not having to play European football. Malcom and Kamano are givens in the 4-3-3 setup, but they have a rotating cast of characters that they could play in between them at striker with De Preville and Mendy. That’s not to mention Gaetan Laborde who was solid last year, and Jonatan Cafu who has been frisky in the limited minutes he’s played this year. That very well might be six competent to great players to work with over a full season. The results so far have been largely impressive. Only PSG has a better total shot and expected goal ratio in the league, and there’s nothing exceedingly alarming with the performances put in. Sure, over the long term they’re probably not going to convert goals to shots on target at just over 37%. At the very least however, they have decent enough shot location to go along with healthy shot volume in a league that isn’t exactly renown for teams shooting over 14 times a match. There are worse foundations to have in attack than what Bordeaux have done so far this year. While perhaps simplistic, one could say that the basis of Bordeaux’s attack is based on transition football. It’s a bit weird considering that this wouldn’t be considered a counter pressing team that uses pressure to create high quality chances the other way, and they don't give you the feeling that they're playing fast football, but it does seem like they’re able to capitalize on individual mistakes https://streamable.com/nazho Bordeaux make this work by always having people commit on the counter, having as many or even more players going forward than the opposition have defending. This is where things like pushing Jeremy Toulalan into a CB position works great, because now you can have more mobile players who can catch up in the play along with the front three. It also takes advantage of Toulalan's passing sometimes being an outlet to get the pace quickened. One thing I'm still wondering with Bordeaux is what is their best lineup within their 4-3-3 setup. On the one hand, De Preville is the best talent that they could put in striker considering his ability to both create for others and himself during his time at Reims and Lille. A slight worry is that considering how much he loves having the ball on his feet, and the same could be said for Malcom and Kamano, it could lead to possessions where it just ends up with them taking really bad shots with little movement to work with. I think having Mendy in there makes for a better fit because then you have a playmaker, an inside shooting forward, and a poacher (to go along with Mendy's ability to hold off the ball and let teammates come to the ball). There's a higher ceiling with De Preville in the lineup, but there's a higher floor with Mendy. It'll be interesting to see where this goes because Mendy has been a killer super sub so far. A problem that could be made with how Bordeaux set up defensively is while they do have the ability to press in certain situations from a medium block, they are still quite susceptible to passes that break their defensive structure. Considering how much they rely on Sankhare to hurry opponents and get play going the other way, sometimes the opposition will drag him out and find openings in the midfield to pass into space and create semi dangerous opportunities. https://streamable.com/pd7q7 Now the good news is that despite those concerns, there’s maybe only three teams in Ligue 1 who can really punish Bordeaux for these problems. And considering the issues Lyon have had so far in attack, it very well might be that only Monaco and PSG could punish a defensive system like the one Bordeaux have implemented. And in the context of merely just trying to finish in the top 3, that might well be good enough. So far this season, the results have been resolute. Only PSG, Montpellier, and Caen concede fewer shots per game and Bordeaux rank in the top three in quality of chances conceded. While the flaws of the system are evident, it might not mean much by season’s end. We highlighted Francois Kamano as a player to watch coming into this season, someone who in a broad sense had a fine season last year. He shot a lot, and produced enough to make the math work in his favor. What I also found interesting about Kamano's game last year was he created high quality chances on average compared to the overall low amount of overall chances he put up. It was a poor man's version of what Ousmane Dembele did last year, and they both share a similar trait in which passing models don't exactly portray them in the greatest light because of their penchant for gambling on passes. The real curiosity with Kamano was with better attacking talent to play with, would he be able to trade maybe 0.5-1 bad shots a game for something better, because while he's shown he is able to create premium chances for others, better shot selection from him could enable him to take the next step as a player. Good news: he's taking around 1 less shot per 90 minutes. Bad news: that hasn't exactly made his shot selection appreciably better Now I get it, he's 21 and there's time for him to figure this out. And it's still early in the season, but we're looking at just over 5000 minutes of game time in his Ligue 1 career and the evidence is showing that on average, the shots he takes are speculative at best. If this aspect of his game doesn't improve, it could very well be that we're already seeing the best of Francois Kamano as a player, and that would be a shame. And then there’s Malcom, who’s arguably the crown jewel on this Bordeaux side. The yin to Kamano’s yang, he’s much more of a creator for others while probably having more functional athleticism than his teammate. He can shift and juke to beat defenders, use his acceleration when needing to get on the end of passes. Whereas Kamano is of the “shoot first and ask questions later” mentality, it’s almost frightening how much Malcom will try and create dangerous opportunities with his dribbling and passing. So far though, the results have been splendid as he’s been able to make something out of nothing with some of his solo runs: https://streamable.com/jbhcp Now you don’t need me to tell you about the problems that come with analyzing seven games without a considerable grain of salt, but it has to be said that so far into the season, Malcom has performed like one of the better creative wide men in Europe. If there ever was a “Leap in Progress” alarm, it would be ringing loud right now with what the kid is doing: Again, it's a small number of games and this could very well be just a nice run of form and not truly representative of his true talent level as a player. But even if he’s somewhere between this current form and what he did previously, we’re still talking about a very talented creative winger who is still be three or so years away from hitting his prime. It could very well be that even a lesser version of Malcom would be good enough to help Bordeaux to make a genuine run at the top three. In a similar manner to Nice last season, teams like Bordeaux need things to go their way if they’re to make a run at the top three in Ligue 1. The difference of budget compared to the likes of Lyon, Monaco and Marseille will always make it tough. In Nice’s case, it was a combination of breakout performances from the likes of Alassane Plea and Jean Micheal Seri and benefiting from the skills of Lucien Favre. With Bordeaux, it might need to be that Malcom in his second full season in France turns into a top 5-10 attacker in the league and incremental growth from players like Kamano and Mendy help bridge the gap for this season. Seven games in doesn’t make a season, and Bordeaux’s numbers will take a hit once they play the likes of PSG and Monaco. Even their first seven games could be nitpicked if needed; while holding Lyon to 8 shots and coming back from 3-1 was impressive, it was done while up a man for the majority of the match. And they were largely unimpressive against Lille when they were once again up a man. I'm also a bit skeptical of whether the front three can click long term, for all that they are talented. Even so, they potentially have six capable attackers, a mobile midfield that can assist in the transition football that they work best in, and a manager who smartly turned a static 4-4-2 setup last year into something greater. Seven games in and so far so good. It might just end with Bordeaux's first trip back to the Champions League since 2010.

Set Pieces and Market Efficiency

At this point, the world transfer market for strikers is fairly efficient. Guys who can get you goals command high fees, are highly paid, and very few good ones get overlooked. They might also get overpaid, but a lot of that has to do with scarcity of elite forwards, and a preponderance of a lot of clubs with a lot of money who need the best. One of the reasons for general efficiency in this area is that the value of forwards is obvious.

  • Goals are valuable.
  • Lots of goals are very valuable.
  • We need to score goals in order to win.

Conclusion: Forwards who score lots of goals are very valuable to us.

It helps that they are also easy to identify.

Why are we talking about forwards and transfers when this piece is about set pieces? Because the market for coaching expertise is laughably inefficient, and better set piece attack and defense comes from coaches.

A couple of weeks ago I was on a panel at the Royal Statistical Society conference in Glasgow when Omar Chaudhuri made exactly this point. “Premier League clubs barely blink at spending £10-20 million on a player, but almost none of them are willing to pay even a few million in compensation for a top coach.”

This is weird when top coaches actually have the most impact of what happens on the pitch of any single person at a football club. Granted, many coaches are average and average coaches seem to have very little impact overall, but coaching impact is a U curve - the good ones and the bad ones cause huge swings in team performance.

Ironaically, one of the few managers where I can remember hearing a compensation fee was paid was Alan Pardew’s move from Newcastle to Crystal Palace. Handsome Pards may be many things, but he’s probably one of the last managers you would want to shell out extra money for. Good process, bad result?

It’s even more ironic because there was a strong rumour that Thomas Tuchel was also part of the extensive search that lead to Pardew being hired. He was on sabbatical then after leaving Mainz, would not have required a fee, but is one of the few coaches in the world you’d be happy to pay for. Ouch.

*Looks at where Palace currently are in the table.*

Ouch again.

I wrote two different pieces on hiring coaches last week, so if you want more information about it, please read these.

What You Really Need to Know About Manager Recruitment

The Right Way To Hire Football Coaches

Talking to Clubs About Set Pieces

As you may or may not know, I own StatsBomb, which has a cutting edge analytics platform called StatsBomb IQ, and offers consulting for clubs across a variety of topics including player recruitment, manager hiring, club valuations for potential buyers, and… set pieces.

Why set pieces? Based on research conducted in summer 2014, I discovered that set piece execution was one of the great overlooked edges in football. As a result, I developed a set piece program that played a tiny part in helping FC Midtjylland win their first Superliga crown (FCM scored 3 goals in every 4 games from set pieces alone before clinching the title). After leaving Brentford and Midtjylland, we continued developing the program, and now offer expertise in this area to clubs.

Now this is the type of edge that people usually don’t talk about once they find it. You don’t want clubs keying on what you are doing in this phase of the game, and certainly don’t want anyone else to find it either. However, owning a consulting business that offers this as a product means you need to tell people about it, which is exactly what I did in detail here.

Changing How the World Thinks About Set Pieces

I understand some general scepticism, and at this point I’m pretty used to initial disbelief [Yo, Alvaro], but justifying investment in this area is incredibly easy.

  • A single goal in the Premier League is worth about £2M
  • Players who can score you goals are very expensive.
  • Players who obviously prevent goals are also very expensive. (Top centre backs command fees of 50M and big wages. Manchester City spent about 150M on fullbacks alone this summer.)

Conclusion: Finding additional ways to score and prevent goals has tremendous value.

Now combine that with my point above

  • The coaching market is currently very inefficient

Cross it with these...

  • Set piece expertise is rare
  • Getting better at set pieces in attack directly leads to more goals without a transfer fee or massive player wages
  • Getting better at defending set pieces directly leads to fewer goals, again without a transfer fee or massive player wages

And you wind up with: Set pieces are both horribly misunderstood and undervalued.

BUT!

Don’t just take my word for it! There’s a strong case to be made that set piece dominance helped power Chelsea to the title last season.

Chelsea were +15 in this phase of the game.

Man City were +2.

They scored nearly identical amounts of goals from open play, but the goal difference in the table between Chelsea and City was +11 in favour of the champions.

There’s also a decent case to be made that set pieces helped save Swansea as well. Swansea were abysmal for years in this phase of the game, but I know for a fact it was a weakness discussed with the owners before the season started. Last season the Swans finished 3rd in the league in set piece goals scored behind only Chelsea and West Brom. That's quite a departure from the 8 they scored in 15-16 and the 4 - yes, FOUR - they scored in 14-15.

Outside England, Monaco beat PSG to the Ligue 1 title last season on the back of 23 set piece goals (PSG scored 13), and they’ve already scored 8 in 7 games this year, despite selling what felt like half their squad in the summer transfer window.

So the first edge comes from knowing you can move the needle significantly in this area if you start to focus on set pieces and give them training time.

Just so we're clear, this has not been common sense or common knowledge in football. A few years ago the analyst community thought performance in this area was largely random, and it’s still treated as an afterthought at the vast majority of football clubs.

A Story

Last year I was invited to talk to a big club outside of England about using data and competitive edges. The discussion rolled around to set pieces, and I gave a presentation similar to what I linked in my intro piece above, highlighting different areas you can focus on with corners, free kicks, etc At the time I was there, the club was doing terribly on attacking set pieces, so I flagged a few quick things they might want to work on to improve.

One of the easy ones was discussing the use of screens on direct free kicks. Allow me to explain…

For goalkeepers, reaction time equates to distance.

The quicker your reactions are, the further you/your hands/your body can travel in order to make saves. Additionally, the further away a goalkeeper sees a shot, the more time they have to react and move to make a save. We even mix up the jargon here when describing what happened in commentary. “He saw it early, so he was able to make the save.” Early is a word associated with time, but what we really care about with GKs is how far they can travel before to intersect that ball before a shot gets past them.

So yeah, time equals distance. Goalkeepers and quantum mechanics for the win.

With direct free kicks, this is quite a big deal. Walls tend to impair goalkeeper sight at least a little bit, and one of the things I learned from Gianni Vio is that attacking teams can make sightlines absolutely miserable for keepers if they want to.

Why would they want to?

Because the later a goalkeeper picks up the ball in flight, the less distance they can move to make a save, and the bigger the target the free kick taker has in order to score a goal.

Translation: it makes scoring a goal from a free kick a lot easier.

I was reminded of this when watching Liverpool face Leicester this weekend. Coutinho’s goal Saturday might have been unsaveable, but he had a goal against Arsenal last season from a free kick that I broke down at the time in a frustrated email to friends at [redacted].

Notice the two-man Liverpool screen in Arsenal's wall. They are standing there to help obstruct Cech's view of the kick as it's taken. There are actually better things they could do to cause problem, but it's a solid start.

Also note how far out this kick is taken. Cech should have ages to see the ball and make the save, even though he's probably a bit too far over in positioning to the far post.

But this is approximately where Cech finally sees the ball. It's hit pretty hard, which creates huge problems (remember, time = distance).

And finally, it's a bit fuzzy, but this is how much Cech missed the ball by. It's a tiny amount.

The reason I bring this up is because of the reaction I got from the coach I was presenting to at the time.

“Well that’s very simple and obvious,” he said.

And he’s right. It’s a gobsmackingly obvious thing to do…

IF YOU KNOW ABOUT IT.

In fact, it's one of the very basic tells Nikos Overheul and I use to see if teams are paying a lot (or even a little) attention to set pieces when we scout them for tactics.

“You’re probably correct,” I agreed. “On the other hand, before travelling here, I watched every direct free kick your team has taken for the last year and your players never screen the keeper.”

*silence*

“I'm with you though, this one is easy to fix once you know about it. Thankfully there are a lot of less obvious things we do to help in this phase of the game as well.”

Sadly, we did not get a chance to help them on set pieces last season.

And despite how obvious my point may have been, I checked their video for this season before writing today - they still don’t screen DFKs in any reasonable way.

For those of you looking to learn a little more about the game, you now have one of our dirty little secrets: always add screens to attacking free kicks.

Wrapping Up

I don’t talk about set pieces much because I don’t want to give away edges to competitors of our clients or ruin the value we offer to future customers. That said, even with more awareness around the value of this phase of the game, most clubs place little or no emphasis on it, despite the fact that it's clearly a cheap way to get extra goals and can both help keep clubs up and win them titles.

We’re currently determining our schedule for the winter break and next spring regarding tutelage for coaches and analysis of set pieces for teams. If you work for a professional club who could use help in this area, please get in touch.

Ted Knutson
@mixedknuts
ted@statsbomb.com

Under Pressure

Models that attempt to measure passing ability have been around for several years, with Devin Pleuler's 2012 study being the first that I recall seeing publicly. More models have sprung up in the past year, including efforts by Paul Riley, Neil Charles and StatsBomb Services. These models aim to calculate the probability of a pass being completed using various inputs about the start and end location of the pass, the length of the pass, the angle of it, as well as whether it is played with the head or foot. Most applications have analysed the outputs from such models from a player passing skill perspective but they can also be applied at the team level to glean insights. Passing is the primary means of constructing attacks, so perhaps examining how a defense disrupts passing could prove enlightening? In the figure below, I've used a pass probability model (see end of post for details and code) to estimate the difficulty in completing a pass and then compared this to the actual passing outcomes at a team-level. This provides a global measure of how much a team disrupts their opponents passing. For 2016-17, we see the Premier League's main pressing teams with the greatest disruption, through to the barely corporeal form represented by Sunderland.

Pass completion disruption for the 2016/17 English Premier League season. Disruption is defined as actual pass completion percentage minus expected pass completion percentage. Negative values means opponent's complete fewer passes than expected. Data via Opta.
The next step is to break this down by pitch location, which is shown in the figure below where the pitch has been broken into five bands with pass completion disruption calculated for each. The teams are ordered from most-to-least disruptive.
Zonal pass completion disruption for 2016/17 English Premier League season. Teams are attacking from left-to-right, so defensive zones are to the left of each plot. Data via Opta.
We see Manchester City and Spurs disrupt their opponents passing across the entire pitch, with Spurs' disruption skewed somewhat higher. Liverpool dominate in the midfield zones but offer little disruption in their deepest-defensive zone, suggesting that once a team breaks through the press, they have time and/or space close to goal; a familiar refrain when discussing Liverpool's defense. Chelsea offer an interesting contrast with the high-pressing teams, with their disruption gradually increasing as their opponents inch closer to their goal. What stands out is their defensive zone sees the greatest disruption (-2.8%), which illustrates that they are highly disruptive where it most counts. The antithesis of Chelsea is Bournemouth who put together an average amount of disruption higher up the pitch but are extremely accommodating in their defensive zones (+4.5% in their deepest-defensive zone). Sunderland place their opponents under limited pressure in all zones aside from their deepest-defensive zone where they are fairly average in terms of disruption. The above offers a glimpse of the defensive processes and outcomes at the team level, which can be used to improve performance or identify weaknesses to exploit. Folding such approaches into pre-game routines could quickly and easily supplement video scouting.   ____________________________________________ Appendix: Pass probability model For this post, I built two different passing models; the first used Logistic Regression and the second used Random Forests. The code for each model is available here and here. Below is a comparison between the two, which compares expected success rates with actual success rates on out-of-sample test data.
Actual versus expected pass success for two different models. Data via Opta.
The Random Forest method performs better than the Logistic Regression model, particularly for low probability passes. This result is confirmed when examining the Receiver Operating Characteristics (ROC) curves in the figure below. The Area Under the Curve (AUC) for the Random Forest model is 87%, while the Logistic Regression AUC is 81%.
Receiver Operating Characteristics (ROC) curves for the two different passing models. Data via Opta.
Given the better performance of the Random Forest model, I used this in the analysis in the main article.

Inferior Tactical Strategies - The Bizarro World of Brentford

Back in 2014, I pitched a document I called “The Blueprint” to Brentford owner Matthew Benham. In that document were four pillars for style of play that when combined created a superior strategy for how to play football. These were based on significant research and analysis across a vast dataset of winning teams and the tactical traits they possessed. My hope was that The Blueprint would form the basis for a future style of play shared by Brentford and Midtjylland, and help guide successful player and coaching recruitment. One pillar of tactical style (obviously) involved Expected Goals. In attack, we want to create the largest volume of high xG shots possible. You can do this in various ways, but it boils down to creating shots from as close a range as possible, and mostly from central areas. In defense, you are faced with a few different choices. You can pick from a couple of varieties of aggressive pressing and attempt to limit shot volume, often at the expense of giving up good shots when opponents do break the press. [Yo, Liverpool.] Or you can concede game control, deep block, and let opponents take mostly bad shots from long range. [Examples: Arsenal. The Fighting Pulises] Or you can pick the Italian/Portugeuse middle block style where you cede the opponent’s half, but blow up everything central in your own half, and try to constrain both volume and quality at the same time. [Conte’s Chelsea, Mourinho’s Manchester United] I have my own preferences with regard to defensive scheme, but whatever you choose, it needs to successfully limit opposing xG somehow. Brentford are currently in 23rd place in the Championship, with 0 wins from 7 matches so far this season. This comes on the back of 5th, 9th, and 10th place finishes the three seasons prior to now. The narrative is that the Bees’ results have largely been caused by “bad luck”, so I thought it might be interesting to dig a bit deeper and see if that was the case. The Attack As I mentioned above, generating high quality shots was something we felt the team should focus on as far back as 2014. The coaching staff were made aware of this, and it was adopted as a core element in Brentford’s style of play moving forward. What was cool is that this actually seemed to work. Brentford’s open play xG per shot went from around 11% under Warbs in 14-15, and up to 12.2% under Dean Smith last season, which is very good for the Championship. Shots were generally getting closer to goal, and the attack was consistently very good. Then this season happened. Through 7 matches, Brentford are generating over 18 shots a game in attack. That’s excellent. But… Average xG per shot? A horrific 6.9%. That’s the second worst figure in the league next to Bolton, who somehow have a 5.5% xG per shot and look doomed already. And on average distance of shots from open play, Brentford are dead last at 20.4m a shot. Now as noted above, Benham owns two clubs. Here are Brentford and Midtjylland’s shots this season plotted side by side. (Note: BFC have played 7 matches, FCM have played 8.) That’s genuinely weird. Midtjylland are churning out amazing quality open play shots. Brentford have tons and tons of shots, but largely from range or wide, while the bulk of Midtjylland's seem to be clustered around and even inside the six yard box. FCM take fewer shots – only 13.25 compared to Brentford’s 18.14 - but the quality difference is so vast that they could take half the shots BFC do and still post slightly better xG numbers. [Note: The numbers in the shot map differ from the numbers stated in the text above because each shot that is part of the maps is granular, whereas the stated numbers are part of calculated values that include possession chains. The chain numbers are more correct, but the maps represent unique shot values and not end-result possession values. These are more likely to differ from one another early in the season. ANYWAY…] The xG distribution charts make the differences even more stark. In basketball terms, this usually gets analysed as per possession efficiency. The Bees play a high-tempo, run-and-gun style but end up shooting almost exclusively long-range twos. The Defense Outside of Lee Carsley’s brief tenure as head coach, Brentford haven’t been a particularly good defensive team at any point since their promotion to The Championship. This year is by far their best performance at limiting opposition shots, solid 10.9 a game, but the quality on those shots is in the bottom 10% of the data set and tied for 4th worst in the league with fellow strugglers Birmingham. Now to be fair, Brentford are mid-table in xG conceded, which is solid. In fact, if Brentford were to match Midtjylland’s attacking numbers and keep their own defensive numbers, they’d be candidates for promotion. But despite the presence of Danish coach Thomas Frank on the coaching staff, it doesn’t seem likely that there will be a Voltron-esque merger of Danish attack and English defense any time soon. The real issue only becomes more apparent when you look at the numbers across both phases of play. The Variance Problem

  • Brentford are currently taking a high volume of low quality shots in attack.
  • They are also conceding a moderate volume of high quality shots in defense.

It’s a mathsy concept, but these two combine to form a toxic variance cocktail that makes poor results more likely to occur. This goes back to a piece Danny Page wrote in 2015 discussing distributions from various quality shots. In the example of Team Coin vs Team Die, two teams create identical xG numbers per game. However, because they differ greatly in quality, they will end up with different point expectations throughout the season. If you had make a choice, you would prefer to create fewer, top quality chances than many low quality chances because the variance in the latter would potentially yield fewer points in the league table throughout a season. In nerd terms, Brentford are practically rolling a 16-sided die right now when they shoot. Their opponents get to roll a D8. That's not a healthy situation. This problem is further exacerbated when you realise that most of the error in xG models is going to be at the extremes. If teams are taking a lot of long range chances against packed boxes, then those chances might actually be worse than the model thinks (and the model already thinks Brentford's are bad). Conversely, if they are regularly giving up high quality chances, those chances might also be quite a bit better than the model’s already beefy assessment. Back to Brentford… Someone on a Brentford forum recently noted that even at Walsall, Dean Smith was considered a “streaky” manager. That’s certainly been true in his time at Brentford, where his teams have combined prolonged losing and winning streaks each season to eventually land mid-table. However, as shown above… there might be a reason for that. The best coaches implement tactics that help control for variance. They can’t eliminate it completely – football is a low-scoring, and inherently variable game. However, coaches that focus on the details and prepare their teams to handle them consistently seem to do better in the league table. For whatever reason, this season Brentford have embraced the opposite.  They are highly variant on both sides of the ball right now, underperforming a lot in attack, and currently paying the price. Equally frustrating is despite a high pace and enormous shot volume, Brentford are near the bottom 5% of all teams in xG from set pieces. In a similar vein to the attacking shot maps above, it’s not as if Brentford don’t know about the value of set pieces. This was another one of the key pillars of playing style in The Blueprint document from 2014 and helped power Midtjylland to their first ever league title in 14-15. Unfortunately, as I’ve explained many times before, having useful knowledge and executing the concepts well are entirely different things, especially when it comes to football. At the end of the day, execution on and off the pitch is what really matters. Will Brentford Get Relegated? Looking at the broader numbers across the league, at this point it seems possible but also highly unlikely the Bees will go back to League One. Unlike Burton and Bolton, who already seem mostly doomed to be flushed from the Championship toilet bowl, Brentford are probably more deserving of a mid-table placement. This doesn’t mean they absolutely won’t be relegated – decent teams have been sent down from the Championship before – but with 39 games left, it’s fairly improbable. Hopefully the explanation above shows that the Bees aren’t entirely undeserving of the current plight either. Brentford’s outputs in attack and defense right now contain of high degree of variance, and are possibly even worse than most xG models expect them to be. Ironically, they also run counter to all the advice we gave them back when various members of StatsBomb were part of the Benham football project. Ted Knutson ted@statsbombservices.com @mixedknuts    

The Right Way to Hire Football Coaches

Yesterday I discussed what you actually get when you hire a new head coach or manager. Today I want to talk about a better way to go about hiring the most important position in the club.

Manager/Head Coach Searches is one of our core offerings at StatsBomb Services, and the feedback we’ve had from clients has been overwhelmingly positive. Because of how messy and haphazard normal coaching searches tend to be, I thought it would be good to shed some light on our process and why that’s a dramatic improvement over the status quo.

On What Criteria Are New Head Coaches Normally Hired?

  •  Reputation (They were a big name player. They worked before at a big club.)
  • Their teams always managed to beat ours when we play them
  • Advice from friendly coaches on who would be good
  • Advice from friendly agents
  • Past Success

The last one is the only factor that actually matters, but it's usually the product of a lot of confounding factors that can be difficult to separate from managerial impact (budget, luck, players, etc). The rest of these are potential disasters in one form or another that don’t do anything to help find what you actually need at the top of your club.

On What Criteria Should New Head Coaches Be Hired?

  • Do they fit our current club needs? (Some managers would be terrible mid-season hires for relegation threatened teams, but great ones if you get them at the start of the year.)
  • Does their tactical style match what we want to do as a club?
  • Have their teams actually been good?
  • Do we have the budget to recruit for them so WE can succeed?
  • Does their personality work within our structure?
  • Do they speak the language? How do we overcome this for candidates who don’t?
  • Can they communicate clearly with players? With management?
  • Can they develop younger players?

There are a host of other criteria a club might care about but these are most of the major ones. Notice how reputation has absolutely nothing to do with any of these.

We care about facts. Data. Objective, actionable information that helps you make decisions. The club itself can decide to care about reputation as well, but we feel strongly that it’s only a minor part of the new coach equation.

How Do You Find A New Head Coach?

When we start this process with clients, one of the first things I ask is, “Do you have an ideal club style you want to play?”

Get rid of concrete examples at the beginning because when you are dealing with the entire world as your potential hiring market, it doesn’t make sense to initially limit yourself. Start with the concept… what would you LIKE to do? What is your academy producing? Why? Should it be aligned with the first team style, or should you keep it more generic in case something changes down the road? How does that change what you are looking for in a head coach now and in the future?

Once you flesh out the conceptual ideal, then you can start the practical discussions.

  • Do you have the squad to play this style currently?
  • If not, will you recruit players that can execute it in the next window?
  • Do you need a bridge coach between what you have currently and where you would like to end up?
  • What does your salary budget look like?

Once you’ve fleshed out what the club wants/needs/is actually looking for, then you start the multi-layered search process.

The first layer data analysis. We have a number of models we use to evaluate manager and team performance over time and compare that to expectation based on budget and squad strength. We can also accurately profile tactical styles to find better candidates that meet club requirements who may be off the beaten path.

Another layer is the usual avenues that potential names come in, via friendly coaches, DoFs, agents, whatever. We’re not discounting these as viable, interesting recommendations – we’re simply applying more scrutiny to them than they might normally get.

Then you throw everyone into a hopper and start to evaluate how closely each name fits the job requirements, quickly coming to a long list of potential decent candidates, and then to a short list of highly qualified individuals.

Why Do You Need Data?

There are a couple of key reasons why we feel the use of data produces a dramatically better process. The first one is that by increasing the supply of potential candidates, you are likely to find more qualified coaches that meet club needs and potentially find cheaper candidates as well.

The second key reason is probably more important though, and only becomes apparent after you’ve conducted a few job searches:

People. Just. Lie.

It’s a natural part of the human condition, and one particularly present during job interviews, head coach or otherwise. If you want the job, you are likely to embellish your qualifications or knowledge on a subject the employer feels is important, even if you don’t have any.

The only way to really combat this problem is to use data. Remember, coaches don’t generally change tactical styles. Therefore, if that style has not been present previously, it is unlikely to magically appear in the future without significant supplemental learning. In general, if a coach hasn’t already proven they can execute what the club wants on the pitch historically, then they probably should not be part of the interview process.

The other thing that happens in interviews is that candidates will agree to execute the team’s required style, but once they have a guaranteed contract, they go back to coaching whatever style is most comfortable for them.

Don’t just take my word for it though… This is another case where the NBA and football overlap in surprising ways. At the Sloan Sports Analytics Conference this year, Daryl Morey of the Houston Rockets and Bob Myers of Golden State were on a panel on negotiation. At one point during the panel, the topic of hiring coaches came up, and Morey pointed out that they had started writing game model + style execution into coach contracts and using it as a trigger for guaranteeing future years. If the coach they hired follows through with the agreement to coach the style the team wants (based on some agreed set of Key Performance Indicators), then everyone would be happy and the contract would continue as normal. However, if the coach was not coaching the required style of play, then the contract would terminate after a single year, thereby limiting the liability on the club side if they have to sack him.

Golden State, who clearly had a superior style of play to most of the league early on, had similar issues with their coach before Steve Kerr (Mark Jackson). This forms a hugely important alignment of analytics, owner fiat, and coaching execution. The personnel on the basketball court are recruited to play the tactical style, which was developed through analytical research. However, if coaches get out there and for whatever reason choose not to coach that style, everything falls apart.

The only way they could protect themselves from coaches saying they would execute the style in interviews and then choosing not to do so was by writing the agreement into the coaches’ contracts. We have seen similar things start to happen with football contracts as well, and expect to see many more of these in the future.

It’s Not Just Data Though… 

Unlike in yesterday’s anecdote about the football club who got down to a short list of new head coaches and no one had watched any of their teams play, we view this as a critical final step in the hiring process. You have all this information… now does it withstand scrutiny from actually analysing the matches?

As part of the viewing process, we also analyse at in-game tactical adjustments, substitution patterns, squad rotation, incorporation of young players, and honestly whatever else the customer thinks is important. We then deliver detailed reports on short-list candidates that highlight strengths and weaknesses, and combine both objective and subjective analysis so clubs have as much information as possible to make this very important decision. Hiring new head coaches is fantastically expensive. Average Premier League salary for the head coach alone is probably close to £3M now. Add in compensation costs for their staff, and sacking a new head coach early could cost between £10-15M just in salary, plus whatever additional impact it has in lost performance, transfer investment, and increased relegation probability.

Conclusion

Hiring a new manager or head coach is one of the most important decisions a club makes, and even successful managers tend to last only three years in a job these days before they move on. Misunderstanding the support needed to make new coaches successful is common, and that can have hugely expensive repercussions, not only from a monetary perspective, but also in the transfer market and league table. Because of this, it is absolutely imperative clubs move to a better, more robust process in their hiring practices like the one outlines above.

If you work in a club and want to talk about this more, please get in touch.

Ted Knutson

ted@statsbomb.com

@mixedknuts

What You REALLY Need to Know About Football Manager Recruitment

Managerial recruitment is possibly the most important thing a football club does on a bi-annual basis. Hiring a poor manager or a bad fit can set off a chain of events that could see a club plummeting through multiple relegations. Hiring a good manager can take an average team and catapult them into title challengers.

However, at the club level managerial hiring is also the activity that might have the single most chaotic, backwards process of anything in football.

Here’s an example a friend of mine relayed to me a couple of years ago:

We got down to the final list of three candidates and something struck me as incredibly, almost impossibly strange…

I looked at the Sporting Director and I asked him, ‘Has anyone actually watched these teams play football?’

*crickets chirping*

No one had. Verifying that what these guys were telling them about their teams and what style they preferred them to play somehow wasn't part of the process.

What’s fascinating is that nowadays you can get video on almost any professional football team in the world. You can even find video down through U18s at a lot of top clubs that’s readily available online. Given the preponderance of potential evidence weighing in either for or against a candidate, not watching their teams actually play football is a baffling choice.

But my point earlier is that recruitment of managers and head coaches is filled with one baffling decision after another. Way more so than modern player recruitment. This is despite the fact that firing a manager and his staff costs millions to tens of millions in compensation costs, and can have serious knock-on effects for the club as a whole.

Because of this, today I want to discuss what teams are getting when they hire a new manager and why that matters. What do you get when you hire a new manager?

  • The Person

This seems obvious, but it’s often overlooked in the same way that footballer’s personalities are overlooked or brushed aside. This is the guy that sets the stage for every discussion you have within your club for the life of their contract. If they are closed off to new ideas, then that will have ripple effects for years. This is especially important to know for clubs that have undertaken a project to become more modern.

It sounds like a cliché, but man management matters. If you have a big squad to cope with things like European competition, you’ll have fringe players that are mostly there as cover in case someone gets injured. Some managers hate having big squads and that can cause huge issues if they aren’t also able to handle squad personalities well.

That said, sometimes a new manager is so good that you are willing to accept possible personality clashes in exchange for better performance. This is exactly the type of thing you really want to know ahead of time. What are the trade-offs we have to make when hiring this person and is it worth it?

Is your new coach a good teacher? A good communicator? A good leader? All of these things matter in general, but they become very important if you have a young squad, your club coaches need to learn their style to train academy players in, or if you have a lot of big personalities in the squad that need managing.

The person you are hiring deals with other people constantly. You really want to know before you hire them how that is likely to impact your club as a whole.

  • Their Staff

This one can be strangely overlooked, but the staff that comes with a head coach can be very important in executing their style, and they can also fill certain roles your head coach as an individual might not be very good at (like man management, logistics, etc).

Here’s an example: some clubs allow new managers/head coaches to only bring two staff members with them when they join. The theory is that institutional knowledge inside the club is important and they want to maintain that, so instead of losing knowledge every time they have to change coaches, they limit potential change. This also forces the new coaching staff to communicate more with club staff as a whole, which generally should be seen as a good thing.

The problem here is that every coaching group is different. In some groups, the most important person after the head coach is the guy who does the fitness training. In others, you have explicit roles and subject matter expertise. (Like the head coach, an attacking coach, a defensive coach, a set pieces coach, the GK coach, etc.) Breaking that when you don’t have ready made roleplayers to fill the needs of the tactical style is a problem.

Finally, over the last few years we have heard of certain coaches waging war on their medical staffs. In some cases, having a trusted physio on staff is a hugely important dynamic because it means new managers get more comfort and clarity about player injuries and when they are likely to return to performance. In the high pressure world of football, a second trusted medical opinion for important players absolutely matters.

I generally agree with minimising change at clubs and at providing a good way for new managers to impart their knowledge to long-term club personnel. However, you need to be sensitive about breaking a whole that is greater than the sum of its parts.

  • Their Style

This is perhaps the most important, obvious thing you get with football coaches but also the most misunderstood.

The common myth: Football coaches can change/learn new styles.

I covered this last year, explaining that how coaches learn is very different from how most of the population learns, and therefore it is difficult and time-consuming for them to take on new things. Expecting wholesale stylistic change is a near impossibility. Coaches can’t learn almost any of what they need to know about new tactical styles from books, so where is the information and execution coming from? Who is teaching it to them? What training sessions are they using to impart this knowledge to the players?

To be fair, some coaches are far more adaptable than others. Part of the education at Italy’s Coverciano is to learn to adapt your tactics and coaching to different requirements and not to be too married to one style. Italian coaches often seem more pragmatic and adaptable precisely because of how they are educated for their licenses. Most coaches can’t do that and expect any degree of success. On the other hand, plenty of people criticise Italian coaches for being too adaptable and too willing to change – the criticism cuts both ways.

So yeah, the vast majority of coaches are strongly married to whatever style they have displayed in the past and are unlikely to change much after you hire them, no matter how hard you wish it were otherwise.

With a particular style comes a whole host of other things including the most pressing and expensive concern: recruitment.

Tactical styles require players that fit the style. It’s easier if they also understand the style, but you need players with the right skill set for any chance of success.

The recruitment team needs to know the tactical style the coach wants to execute. Then they need to talk to the new manager and pull out specific role requirements for each position on the pitch and compare those needs to current playing staff. Once they do that, you can construct a recruitment plan based on the new requirements compared to squad weaknesses.

There is no generic football style. Different managers require different players. This is what makes changing managers regularly almost catastrophically expensive for clubs, especially if you let those managers also run recruitment.

Some tactical styles require pace and endurance at every position on the pitch. Size is a “nice to have,” but you will take a smaller, faster guy over a bigger one every time. Tony Pulis’s requirements are basically the opposite. Pace matters up front, but everywhere else you need beef.

Jurgen Klopp and Pep Guardiola’s teams both press. But their positional requirements for who they can recruit are almost surprisingly different.

Conclusion: Your recruitment team really needs to understand the style of play of your new manager in order to succeed.

Conclusion 2: If you do not recruit for your new manager’s style, you will fail. Especially if it differs significantly from the style that your team has played in the past.

The Case of Crystal Palace and Frank de Boer

In practical terms, the part about style circles immediately back to Crystal Palace’s situation this summer and the hiring of Frank de Boer.

As a manager, I think de Boer is actually pretty decent. He has proven he can coach a defense over the years, and though his attacking style is often regarded as boring, given the right players he has been successful. Palace hiring FDB was a bit of a risk, but also a decent shout if they wanted to raise the ceiling of the club. De Boer is perhaps not as good as hiring Marco Silva or Roger Schmidt might have been, but those guys went elsewhere, and Schmidt at least would have had even greater recruitment needs to succeed.

What de Boer also brings to the table is a very clear tactical style. His clubs are going to play a variation of Dutch/Ajax press and possess football. To pull this off, you need players who are understand how to execute this style, especially in the center of the pitch. And unlike the German zonal style, individual players probably have higher style-IQ requirements because it’s largely a man-marking system.

Right, so clear tactical style. Very unlikely to adapt it. Data analysis also shows zero indication he’s capable of tactical variation. Because of all this, you need to provide him players that can succeed and PROBABLY quite a bit of time to teach the team his system.

Remember, Pochettino’s first season at Spurs was bad defensively. He needed a year to imprint the learning on the squad and also multiple transfer windows to get players capable of playing the system at a high level.

So to support their new style that comes packaged with their new coach, Palace bought Jairo Riedewald (age 20), and loaned Timothy Fosu-Mensah(19) and Ruben Loftus-Cheek (20). They also added Mamadou Sakho on deadline day.

This is a complete and utter failure to recruit for the stylistic needs of a new manager/head coach. They also brought in Dougie Freedman as Sporting Director this summer, which was nearly as interesting/baffling as their recruitment when it comes to joining up style/head coach/club needs.

This is why clubs need to clearly understand the impact their choices on head coach/manager have on their future. If Palace weren’t going to do a lot of recruitment this summer because they didn’t have the budget, they should have gone a different direction with their coaching hire. Save the money you’re going to have to pay from sacking FDB and his staff in the first half of the season and use it in the next transfer window.

Hiring FDB and not recruiting for him is basically lighting money on fire.

To be fair to Palace, the squad as a whole is decent. There’s just this huge problem in that it doesn’t fit de Boer’s tactical needs at all. This is something every single football club needs to be aware of when making new managerial hires.

Later this week, I’ll discuss an improved process we've developed for coaching hires that delivers a better chance of immediate success and a brighter future.

Serie A-ttack!

  The prevailing perception of Serie A is one of a league typified by the art of defending. Just the phrase 'Italian football' conjures up VHS-quality memories from decades gone by, images of Arrigo Sacchi's AC Milan teams moving as one fluid unit and myths about Paolo Maldini somehow defending without ever tackling. For many years some of the stereotypes did have truth to them. Times have changed though, and Serie A has changed along with them. Before we get there we need to look back at the league's recent past. When viewed from a macro perspective, Serie A’s statistical profile sticks out like a sore thumb. Italian teams take lots of shots. For the past five seasons the league has been at or near the top of overall shots per game amongst the top five European leagues. Something which has persisted as the rest of their peers on the continent have diminished in this regard.

Shots Per Game in the top five European Leagues
On top of that, Serie A teams have always loved a longranger. Amongst these leagues they’ve taken the highest percentage of their shots from outside the box in each of the last five seasons. Add all this together and what you get is a lot of teams with inefficient attacks. Yet this looks very different in 2016/17. As you can see above the overall shot volume remained high; it was the 4th highest of any season across this timeframe. Yet the shots from outside the box dropped more in line with the wider trends. Still the highest compared to their European peers but far more sensible. A drop in percentage of shots from outside the box of near enough 9% from 12/13 to now is quite an astonishing change. When visualised in terms of shot distance it becomes even more evident.

Average Shot Distance (metres) in the top five European Leagues.
This shift is reflected in the league’s wider expected goals production. For many a season Serie A was generating very uninspiring non-penalty expected goals per game numbers. 2015/16 especially was this odd one where creation across the board dropped and teams just weren’t taking as many shots. 16/17, by contrast, saw Italy fall way more in line with their peers. The mind-boggling shots totals persisted but now a majority of them were good shots.
Non-penalty Expected Goals Per Game in the top five European Leagues.
Italian domestic football has ostensibly thrown off the fetters of its associated stereotypes and become a more multidimensional league. There is a clear impetus towards better shot locations and a specific, modern style of attack. So who is leading this charge? Roma Luciano Spalletti's return to Roma for his second spell only lasted a season and a half due to club politicking (he’s now in charge of Inter Milan - more on them later) but boy howdy was it an impressive old knock. He joined in January of 2016 and by the end of the 2015/16 season they were already putting together a good attack. In 2016/17 that good attack became a great one.

Roma
Season Average shot distance (metres) Non-pen xG
Non-pen xG per shot

14/15 20.82 45.4 0.088
15/16 18.94 60.9 0.105
16/17 17.71 74.6 0.113

The fulcrum of this was Edin Džeko, who had an obscenely good season for a striker. He was the league’s top scorer with 29 goals and was getting 5.2 shots per90 in the league, with 4.6 of those coming in the box. For comparison: Gonzalo Higuaín - not a bad striker himself as you’ll be aware - was putting up 4.1 shots total. A couple of other players were close to his shots totals but no one came near to his monstrous amount in the box. These are peak-Cristiano Ronaldo numbers. In fact, it’s the highest since 2012/13 at least.

Seasons with most shots in the box Per90. (Top 5 European leagues, 2012/13 - 2016/17). Minimum of 1500 minutes played.
Player Season Minutes Shots Per90 Shots in the Box Per90
Edin Džeko 2016/17 3050 5.22 4.6
Cristiano Ronaldo 2013/14 2536 7.67 4.47
Robert Lewandowski 2015/16 2653 5.12 4.27
Sergio Agüero 2013/14 1527 5.07 4.18
Sergio Agüero 2014/15 2528 5.27 4.09
Gonzalo Higuaín 2015/16 2963 5.53 4.04

Whether these eye-watering stats can continue without Spalletti or Mohamed Salah remains to be seen. New manager Eusebio Di Francesco did fine work at Sassuolo, but he has big shoes to fill on that running track at the Stadio Olimpico. Juventus The Old Lady may be ever-dominant but that doesn’t mean she rests on her laurels. The first couple of seasons under Massimiliano Allegri were quite good from an offensive perspective, yet the defence and its stellar expected goals against numbers were doing most of the heavy lifting on the way towards consecutive titles. 2016/17 and the addition of Gonzalo Higuaín layered an extra level of attacking power on top.

Juventus
Season Average shot distance (metres) Non-pen xG
Non-pen xG per shot

14/15 18.99 52.4 0.088
15/16 19.54 55.7 0.094
16/17 17.73 66.4 0.114

That they somehow managed to do this without sacrificing any of their defensive obstinacy (finishing with a measly expected goals against tally of 25.3) was all the more impressive. They don’t take as many shots as Roma (15.3 per game vs 17.8 for the boys from the capital) but they’re every bit as deadly. Both sides are also pretty indifferent about how they get those shots, as long as they get them. In terms of percentage of their shots as headers Juventus rank 8th in the league and Roma 7th. Napoli Napoli are something akin to poster boys for this 'new look' Serie A. Manager Maurizio Sarri took over from Rafael Benítez in 2015 and ever since then they've become darlings in pockets of the football community for their slick pass and move style. Away from how easy on the eye they are though, their numbers and how they've changed are quietly intriguing.

Napoli
Season Shots Per Game Non-Pen xG For xG / Shot
Avg Shot Distance (metres)

13/14 (Benítez) 15.2 57.6 0.101 20.84
14/15 (Benítez) 16.4 59.7 0.097 19.78
15/16 (Sarri) 17.3 64.0 0.099 19.91
16/17 (Sarri) 17.7 70.0 0.105 19.40

The difference between Benítez and Sarri is there but it is more subtle than you might expect. Sarri's iteration operate as this sort of emblematic bridge between the older Serie A style of attack and the modern incarnation. They bomb shots in but are also geared towards higher value shots than their forebears. The truly impressive part here is when you realise that they lost Gonzalo Higuaín - the league's top goalscorer in 15/16, responsible for 45% of the team's goals - to Juventus in the summer of 2016. Far from suffering because of it they actually somehow got better. Dries Mertens out of nowhere had easily the best season of his career, scoring 28 goals (1 behind Džeko). That tally was a bit hot compared to his individual xG but the overall team numbers don't lie. They were the ne plus ultra of egalitarian attacking football. Inter Milan One of Serie A's overarching storylines of late has been the relative decline of Inter and AC Milan. Both were once regulars at the top of Italian and European football but recent seasons have seen them struggle to break into the league's top four, or even the top six. AC Milan overhauled their squad this summer with many new and young players, meaning they'll likely be radically different than their uninspiring 16/17. However their city brethren are starting from a slightly different base. Inter cycled through four different managers on their way to a 7th place finish, contributing to the image of a club in constant turmoil. However, the underlying numbers painted a much kinder picture. Their attack came on leaps and bounds in every department, coming close to par with their title-challenging ambitions. All this despite them being shifted around between managers like a hot potato. There's clearly great potential here. With Spalletti in charge for a full season that potential might actually be properly maximised.

Inter Milan
Season Shots Per Game Average shot distance (metres) Non-pen xG
Non-pen xG per shot

13/14 14.5 19.19 56.2 0.102
14/15 16.2 19.76 53.5 0.088
15/16 13.6 19.32 52.5 0.102
16/17 16.7 17.63 65.5 0.104

__________________ Check out a Serie A match next time there's one on TV and you'll see the difference that is apparent in the numbers. Teams throughout the table are now getting noticeably smarter with how they attack, as well as the action just becoming a bit more open and fun to watch.  The old days of endless long shots against packed defences are in the rearview. It would be unreasonable to expect Juventus' stranglehold on the title to be broken given how long they've been sovereign for but, regardless, the league looks as competitive as it has in some time.  Whatever the end result it'll likely all be a good bit of craic along the way. __________________ If you have any questions about the article you can DM me on twitter @EuanDewar. Enjoy the season.