Gladbach: luckiest team in Europe or a a blind spot in expected goals?

Borussia Monchengladbach are back on the rise going from a league average Bundesliga side just a few years ago to comfortably qualifying for the Champions League and pushing for 2nd this season. They have rebounded from losing Reus, Dante, Neustadter and others after their great 2011-2012 season and climbed back toward the top of the league. There is something strange about Gladbach though, and you see it when you compare their shot numbers to Wolfsburg, who have a had a similar run.

Gladbach Points Rate

 

Gladbach TSR

The Foals haven’t neared 50% TSR (total shot rate, team shots/total shots in game) meaning they are repeatedly being outshot by their opponents. This is not how a Champions League team plays, in fact no team in a top 5 league currently in a UCL spot has a TSR of 50% much less Gladbach’s 45%.

So how can we explain Gladbach’s great season, which has them in 3rd place with the 3rd best goal difference?

It doesn’t have much to do with distance, they don’t take closer shots than the average Bundesliga team: 11th in close shot %, 12th in average distance. Defensively they do allow a tough shot:

Even with distance accounted for, we can’t explain Gladbach’s numbers. Take a look at this chart of last 3 years in the Bundesliga, it shows teams who have outperformed their expected goal* total on the offensive side:

*Expected goals is a model that factors in shot location, body part and game state to give a number showing how many goals that shot is typically worth.

This year, the Foals are in the middle of the pack but the previous two seasons they put up the 2 largest over-performances of the 54 seasons I have data for.

On defense, it’s a similar story:

Gladbach have the 2 largest under-performances of expected goals in the past 3 seasons. They have come this season and last season, with the 2012 year above middle of the pack.

Combining those two numbers we see that Gladbach have had 3 of the 4 top seasons where the goal percentage outpaces the expected goal percentage

How do they do this? Expected goals might be missing something, Gladbach could be the luckiest team in Europe or a combination of the two. Let’s look at the luck factor first.

Over the past 3 seasons Gladbach has a expG%/actual goal% ratio of 1.20. Here is the list of individual seasons across the top 5 leagues that reach that level of outperformance:

Sunderland 2012/2013 (1.23)

Man City 2010/2011 (1.21)

Levante 2013/14 (1.30)

None in France or Italy

So Gladbach’s 3-year average “outperformance” was equaled only 3 times in the other 331 single seasons I have in my database. When other teams are reaching your 3-year level approximately 1% of the time over a single year, I feel confident saying there must be more going on than just blind luck.

What factors are there that could lead Gladbach to perform so much better than their shot location data suggests?

DEFENSE

 

We see here that Gladbach are great at most per-shot defensive metrics. The distance is accounted for directly in the expected goals model and the others are regressed but still accounted for in part. Blocking shots, keeping balls off target and a low G/SOT have varying R squared values around between .3-.5 and go into the model but as we saw, Gladbach blow past the model each season. Do the Foals and Lucien Favre (manager since 2011) have a strategy or tactic that gives them more control over opponents rates than I expect and less left up to luck? Let’s take a look at the passing stats to see if they are doing something special there.

The first thing you notice as you peruse the pass numbers is Gladbach do not bother the opponents very much. They have allowed the 3rd, 1st, and 4th most completions per game and top 3 completion % against in each of the last 3 years.

At first glance, this ease of passing is distributed over the entire field. They have been the easiest team for opponents to pass against in all three thirds of the field (own, middle, attacking) in each of the past two seasons. The only exception is Koln has been easier to pass against in the midfield this season. This seems to imply Gladbach are putting zero pressure on the ball and simply sitting back.

These numbers don’t really show them ever pressuring the opposition, even as they move into the final third. We need to look deeper to find Gladbach’s defenders.

Within the final third, we see it gets harder and harder to complete a pass as you move toward Gladbach’s goal. This following graph is using a z-score, so a 0 is comparable to league average in each area. You can see that the Foals go from nearly 2 standard deviations easier to pass against to about .7 standard deviations on the tough side as opponents close in on goal. We finally find the Gladbach defense: they are packed in close protecting their goal.

This packed-in defense also stops intrabox passes (passes that start in the box and end in the box):

No team is harder to complete that extra pass inside the box against than Gladbach. Michael Caley in his ESPN piece wrote about how “extra-pass” shots can be scored at a higher rate: http://www.espnfc.com/blog/tactics-and-analysis/67/post/2427824/barcelona-arsenal-attacks-excel-due-to-danger-zone-passing. This makes sense as someone playing a shot immediately after a pass is generally going to have more space as the defense wasn’t closely marking him or paying full attention to him just seconds earlier giving him more of a window to get a shot off. Thus intra-box pass completion rate, block rate and pass completion rate close to the goal is likely the closest proxy we have to defensive pressure on shooters which is the biggest weakness of most current (mine included) expected goal models. Gladbach rate near the top of the league in all three of those categories, which means that all those shots they allow are harder than we’d expect on first glance.

Example pulled from one game (not claiming this is representative at all, it’s just one example)

Dortmund win the ball back from a Gladbach attack and look to have some space on the counter with 2 on 3 at the worst and maybe a 2-on-2 with pace:

Gladbach’s entire team sprints back full-speed to set up and stop the counter

a few seconds later, they set up their typical defense:

You can see all 10 men behind the ball in their own third. Dortmund wind up completing a lot of passes but none of any danger as they pass the ball around out of the reach of the Gladbach block before trying a speculative ball over the top. Lucien Favre has gotten his entire team to buy into getting back and defending in deep areas (6 different players are in the top 32 in blocking shots) in what seems to be a very effective system.

OFFENSE 

Offensively Gladbach are a very deliberate team. In the 2013-14 season they had the lowest attack/own third ratio of any team in the Bundesliga. They spend more time passing the ball in their own third than any other team. That’s not where you usually see good teams:

This slow, deliberate pace allows them to complete a huge % of their passes and complete more passes per game than everyone shy of Bayern the past two seasons.

In total these past 3 years their offense has outscored significantly what we’d expect from their shot numbers alone. Let’s take a look at their metrics:

They take shots from way out yet routinely rack up good goal/SOT and SOT% numbers. Let’s see how they do this.

Counters?

Probably not. It makes intuitive sense that a team that sits so deep might break at pace on the counter and create some clean looks at goal but the numbers don’t really bear this out. WhoScored.com credits Gladbach with six counter goals this season (in a 5 way tie for 4th), but only two last year (t-14th) and three in 2012-13 (t-15th). They don’t score a lot of goals from the counter.

Passing Style 

Let’s look at the pass #’s again to see if they are doing something different passing wise, like they were on defense:

Here we see that as they move closer to goal in the attacking third, their pass completion z-score increases. Inside 25 yards they are 2 standard deviations above normal. Gladbach also have the best % in the league when it comes to intrabox passes.

Gladbach don’t get the ball into box at massive rates but when they do they generally play it in from the sides.

This helps keep the ball (sides much higher completion % than center) at the expense of creating dangerous chances (over 3x more likely to create a chance from the center of the pitch). However, they don’t rush the ball into the box with crosses when they are wide. No team completes fewer crosses per game (per WhoScored) and no team plays a higher % of backwards passes from wide positions than Gladbach. This indicates a team that picks its spots and values the ball. Only Bayern has a higher completion % when playing a medium or long pass and no team has more assists on those type of passes than Gladbach. The Foals have 20 assists on passes of 15 yards or more, Wolfsburg have 17 in second place while the average Bundesliga team has 8.

What have we learned?

Gladbach games feature a ton of completions. The Foals do not pressure opponents into turnovers, allowing teams to move the ball up the pitch with ease until they get close to the goal. That’s where the Foals buckle down, apply pressure, and force opponents to a lower shooting and passing percentage than most teams. On offense, the team moves the ball at will completing an extremely high percentage of passes. Playing up the wings helps with this but there is no rush to swing crosses into the box. Gladbach complete a great percentage of passes on the attack, especially passes that travel 15-35 yards (which they use more than any other team in the league). No team is better at completing passes when inside the box, indicating Gladbach uses their deliberate pace of attack to move several players up the field. This leads us to believe that Gladbach are taking more one-timers or shots as the result of passes before the defense has re-adjusted. Their offensive numbers are the opposite of the defensive numbers as they improve their passing percentages close to goal and generate cleaner chances than the average team.

Expected goals models can certainly take this into consideration at some point, my summer improvement will hopefully have some sort of passing style factor but for now it’s enough to realize that Gladbach’s distinctive style does not come through in expected goals models. Their over-performance is not simply luck (though how much is luck/repeatable skill is certainly up for debate). This would have been nice to know the past two years when they repeatedly hurt my bankroll every weekend, going forward I will be much warier when going against the Foals and they are an important case study for anyone wanting the quantify the game.

Advertisements

Thomas Tuchel: flexible styles to control the center in the final third

Thomas Tuchel is the new manager for Borussia Dortmund. What is his style, how effective is he, and what will change at Dortmund next season? There are many quotes and summaries out there about his style, including several from the man himself: “There’s definitely a style that’s been attributed to me, that we brought to the table at Mainz: pace going forward and attack-minded football. I prefer certain qualities, an active playing style, bold defending and pacy play in attack.” is how he described himself. He’s been described as someone who is constantly looking for new ideas and ways to improve his tactics from teams and coaches around him. He has supposedly been visiting with basketball and volleyball coaches during his year off to see if he can glean anything from them. By all accounts he is an intellectually curious man, let’s see how that translated to the pitch at Mainz.

Effectiveness
I only have data for the previous three seasons so some of Mainz’s best seasons will be cut out of this analysis. Tuchel took over in 2009 right as Mainz were promoted to the Bundesliga:

2009-2010: 47 points, 36 GF-42 GA, -6 GD
2010-2011: 58 points, 52 GF-39 GA, +13 GD
2011-2012: 39 points, 47 GF-51 GA, -4 GD
2012-2013: 42 points, 42 GF-44 GA, -2 GD
2013-2014: 53 points, 52 GF-54 GA, -2 GD

Very similar goal difference except for the brilliant 2010-2011 season where the team won 7 straight to start the season. Mainz and Freiburg both came up in 2009, both have similar monetary limitations (Freiburg even more so) and have generally overperformed them. His Mainz record can’t be described as anything other than a success. They are one of the bottom 5 teams in revenue and have performed like a middle of the pack team since promotion.

Style

One training strategy Tuchel uses that has become widely reported is his “rhomb-training” where the pitch is adjusted to discourage low-percentage passes. Most of the articles about him mention something similar to the following:

“Tuchel’s focus in practice is on “rhomb-training” with the pitch dimensions being cut to resemble a diamond, ruling out long passes down the touchlines.”

From: http://www.dw.de/rule-breaker-tuchel-takes-on-job-of-replacing-klopp-at-dortmund/a-17503661

Let’s see if his 2 most recent Mainz teams really ruled out long passes down the touchlines, first in 2012-13:

Here we see his teams take their training to the pitch and joined Bayern and Dortmund as teams that rarely use the long ball down the touchline. In 2013-14 however, it was a different story:

Mainz hit more passes to those spots than the average Bundesliga team in 2013-14. Is this ineffective coaching or did Tuchel simply change his system? I think he changed, this next chart shows the main difference:

The red line indicates final third control or the ratio of Mainz passes in the opponents final third to opponents passes in the Mainz final third. You can see this ratio drop significantly throughout the second season (which starts halfway through the chart). In 2012-13 Mainz generally made about as many passes as their opponents did in the final third but for the large majority of the next season, they were making only 75% or so as many passes. The blue line is the % of passes forward that are long balls, with the “1” on the left axis signifying a league-average amount. With more possession and territory being ceded, Mainz moved the ball up the field using significantly more long balls. The defense was also sitting noticeably deeper in 2013-14. In 2012-13, the Mainz defense won the ball back an average of 40 yards up the field, which dropped to just over 38 yards up the field the next season.

In the 2012-2013 season, Mainz attempted fewer than the league average in long balls while the next season they led the league in long balls as a proportion of the total. Sitting deep, ceding possession, and hitting long balls are things Dortmund has done very little of in the past three seasons.

High press

Klopp has been well-known for his gegenpressing, which involves attempting to win the ball back high up the pitch. Some of the criticism of him this season has claimed he has no plan B after this strategy and that the cumulative effects of this (and losing his original fitness coach) have led to fatigue and a performance drop. Tuchel brings a different approach to the high press.

2012-13:
2013-14: 

In the 2012-13 season we see Mainz actually has a similar pressing rate to Dortmund up in the top half of the league, but then the next year they press (press here referring to takeaway rate) less than any team in the Bundesliga. Tuchel is clearly not fixed on pressing high up the pitch. Mainz pressed less, played deeper, and had the ball less in 2013-14 but did keep one defensive metric near the top of the league, we will get to that next. We can see from the changing possession, pressing, and passing styles that Tuchel cannot be accused of lacking a plan B, he is a very flexible coach.

Centrality: the Tuchel goal?

Tuchel’s teams in these two seasons do a great job of forcing opponents possessions to the wings while moving their possession through the center of the pitch. This is important because chance creation is more than twice as likely from the center as it is from the wings.

2012-13

2013-14

Mainz also do a great job of getting the ball to the center of the pitch when on offense.

2012-13

2013-14

I’m not willing to go so far as to say this is the fulcrum of Tuchel’s ideology, but when possession, press intensity, long ball rate, and territory all changed significantly, this remained near the top of the league. In both seasons, Mainz consistently denied opposition access to the center in danger areas on the pitch and got into those areas when on offense. This means Mainz faces low percentages of attacking balls played into the box (3rd and 2nd lowest the last two seasons) while being one of the better teams at getting the ball into the box at a high rate (6th and 1st). Their forays into the final third are much more dangerous than their opponents.

Shots, Goals, and Expected Goals
In both seasons, Mainz had a similar total shot rate. In 2012-13 they took 46.9% of the shots in their games, and the next season they took 46.9% again. Mainz increased their shots from 10.7 to 13.8 per game but saw their shots allowed rise from 12 to 14.8. The pace of shots increased from 2012-13 to 2013-14 but shot quality (expG/shot) was extremely close league average for Mainz and their opponents in both seasons. The less active defensive approach in 2013-14 led to more shots from both teams in total.

Mainz and their opponents had basically league average shot profiles as far as distance was concerned but angles saw some slight deviations. In 2013-14, Mainz forced opponents to shoot from sharp angles more than anyone outside of Bayern and in 2012-13 they took shots from central angles more than anyone else, this is likely related to the fact Mainz has the ball in the center so much and forces opponents out to the wings.

Exp G and G ratio:

Mainz ranked 7th in goal% over these two seasons. They outperformed their exp G% at a level higher than any team other than Gladbach and (who are simply a bizarre statistical case that warrants a deep dive but that’s for later). Some call this overperformance “manager efficiency”, crediting Alex Ferguson for dragging a United side that didn’t have league winning expG numbers to titles. It’s certainly not as simple as any difference from expG is manager quality, most likely Tuchel was had some element of luck in outperforming expG.

Game States

In the 2012-13 season, Mainz were simply awful after falling behind, being outshot by a greater margin than any other team in the Bundesliga. They coupled that with getting a worse quality of shot than any other team in the league. A ridiculously lucky conversion rate (twice the expected rate) covered up their level of performance, which was the worst in the Bundesliga. When taking the lead, Mainz would also sit back and cede the majority of the shots (opponents took nearly 60% of total) but would pick their spots with precision as they took a higher quality shot than any other team (over 50% better than the average shot).

In the 2013-14 season Tuchel fixed this problem. There were only minor differences between Mainz at the different game states. They were the same team ahead, even, and behind, one that is basically 50/50 to get the next goal.

Have we learned anything?

Keep in mind this is only a two-season look, so no conclusions should be taken as firm. With that said, I feel confident saying Tuchel is a very flexible manager. In 2012-13 he played a reasonably high-intensity press, Mainz and their opponents spent equal time in the opponents final third and his teams hit few long balls. The next season, the press totally disappeared; Mainz spent way more time defending in their own third, and played a lot more long balls. The counter attack was used much more. Both seasons saw similar results. Tuchel is not an idealogue like Klopp is.

He’s also a problem solver. His team was awful when falling behind in 2012-2013, they were lucky to get back in as many games as they did. Whether the entire style change was because of this or not, the next season Mainz were much better when falling behind.

He values the center. No team did a better job of getting to the center of the pitch in the final third and denying it to their opponents than Mainz over the two-season sample. Dortmund’s playmakers like Kagawa, Reus, and Mkitaryian will have plenty of chances to get the ball in dangerous areas.

There will be a change in style, the passion and non-stop running of Klopp’s press will not seamlessly transition to a like-minded manager. However, with opposing teams simply letting Dortmund have the ball and Dortmund not really knowing what to do with it maybe something new will benefit the side that will be looking to climb back into the Champions League spots next season. It will be interesting to see how Thomas Tuchel adjusts and tinkers with a side with much more financial power than Mainz.

Analyzing passing patterns against Marseille

I have made a few posts recently on here looking at Marseille’s distinctive defense but can’t help myself from dropping a third, and hopefully wider ranging and more informative post. I’ve looked at how other teams pass the ball against Marseille throughout the game as they strike a distinctive profile in many ways. This is very chart and graph heavy so if you hate those, this is not the article for you. If you can’t get enough of those, take a deep breath and jump in.

What the Marseille defense does well

-Keep the opposition offense from possessing the ball.

This is tied in with the offense, if you have an offense that completes a high % of their passes you keep the ball away from the opposition defense. However, PSG have a higher offensive pass completion rate and Lyon and Bordeaux have similar rates to Marseille and yet OM is in a class of its own as far as preventing completions.

-This is because they take the ball away quickly, especially at the start of the game: A takeaway is an interception, tackle, or ball recovery where the defense wins possession back. A few tackles that knock the ball out of play may be in the numbers but that will not significantly change the conclusion that Marseille are winning the ball back at a rate well above the rest of the league. They have the highest rate of takeaways in their own third and the midfield, and only PSG takes the ball away more in the opposition third.

-A big reason they take the ball away so much is their pressure forces teams to hit long balls up the field, which are easier to intercept. When the opposition has the ball in the midfield or their own third there are fewer options open due to the higher pressure which results in longer balls which are harder to complete and easier for Marseille to win the ball back.

-These long balls are hard to complete, especially against Marseille at the start of each half.

Just over 30% are completed, well ahead of the 2nd place team. This is likely because long balls against Marseille are less likely to be measured passes toward a clear target but more likely to be aimless boots vaguely forward to escape the pressure and avoid a more costly turnover deep in their own half.

-Opponents struggle to string long moves together

This is a pretty cool stat. It’s not one I am 100% confident in the accuracy of because it doesn’t factor in balls lost by dribbles and possessions ending in shots (a streak can continue after losing the ball that way) but it is still a rough way of seeing who is forcing you to move quick. And all of these stats emphasize the fact that Marseille make you move forward and do it fast or they will get the ball back. Most teams respond to this by hitting a long, inaccurate ball forward that is often won by Marseille.

So what goes wrong?

-With so many players pushed up the pitch and spread across the pitch, early in games Marseille can be cut open right down the gut.

The high-reward part of the press brings huge amounts of turnovers early in games as a frazzled opposition deals with the onslaught but there is still a high risk that if you get beat, the opposition will be in a good position and is able to put you at risk of conceding (middle of the pitch passes result in an assist or a scoring chance at 3 times the rate of a pass from the wings). Once the opposition make it into the attacking third, they take advantage by completing passes to the box at a very good rate.

This is a tradeoff almost every manager would make, a little extra sweat when the opposition breaks into your box is worth the big possession edge you are gaining, and the number of offensive chances taking the ball away will create. This is not the main tradeoff however, as anyone who watches Marseille or who has noticed the abundance of 1st half charts so far will know, it’s the fact that the game lasts 90 minutes.

The in-game fade

Marseille’s defensive pressure falls away as the minutes tick by. There is a small fade at the end of the first half, after halftime they pick back up but the fade returns stronger than in the first half as the 70th minute nears. The opposition controls more of the ball as the game goes on (game is broken up into 4 “quarters” min 1-22, 23-45, 46-68, and 69 on):

Monaco, Lyon, PSG, and the Ligue 1 average are added in for reference. Teams in general (and those other 3) allow fewer completions as the game goes on. Marseille is the opposite. The next two graphs show how the pressure eases and the opposition has an easier time moving the ball around as the clock runs toward 90:

Takeaway rate drops
which goes hand in hand with the opposition completion rate increasing

What was a ferocious press in minute 1 that forced opponents into a lower pass completion rate and a higher rate of turnovers than any other defense has become league average at both by the end of the game. This would be ok if it was part of a strategy, maybe Marseille is sitting back more, bringing more players deep and shutting down the openings that come with a poor press. This is not the case. The long balls that were so hard to complete in the opening minutes of both halves (see chart above-32.4% is the number), become much easier as the game goes along:

Long balls are completed at 43% in the second and fourth “quarters”

This leads to opponents spending way more time in the attacking third than they did at the start of the game, as they make over 2.5 attacking third passes for every 1 own third pass late in the game: Only 3 teams (Nice, Rennes, and Bastia) allow a higher ratio from minute 68 on. The other 4 teams in the top 5 allow a ratio of 1.47 in that time span. The opposition having so much possession in the final third against a defense who can’t take the ball away anymore spells trouble, which comes in the form of total number of passes into the box completed (chances created and assists have a similar trend). I think that satisfies the claim that it’s easier to move the ball against Marseille as the game goes along. Another possible downside of the full-blast pressure from the opening whistle is it robs you of some tactical flexibility as the score changes. As noted soccer tactician Nigel Tufnel has mentioned before, when you need that extra push over the cliff, Marseille are already at eleven. They can’t increase the intensity of their press anymore to try and change the game.

You can see Lyon doesn’t change anything while Monaco and PSG immediately crank up their pressure to win the ball back more after falling behind. Marseille’s falling rate can be explained by the fact that even game states happen proportionally closer to the start of the game while trailing game states happen closer to the end, when the fade has started. The fade is powerful enough that Marseille has no response late in games, they can’t change their defense much because they have already gone to eleven. The ratio of attacking third completions/own third completions, % of passes into the box and long pass% allowed all look the same when Marseille is behind as when they are tied.

What about when they take the lead, do they sit back and become harder to break down? They get less takeaways:

The % of attacking third passes that result in assists or chances created increases (from 6.8 to 7.8%), the share of completions in the attacking third that end in the center of the pitch (where a chance is 3x more likely to be created from) drops but drops equally to the league average drop. (Marseille share drops from 29 to 26%, league average drops from 23.7 to 21.4%). However, two important stats do change when Marseille take the lead. First, share of possession. In the table below it is represented by share of total completions:

You can see that Marseille let the opposition have the ball a small but significant amount more when they take the lead. The other metric that changes is the share of opposition passes in the final third drops even though the league average rises in that game state.

So we see that Marseille do sit back, concede a bit of midfield and own third possession at the benefit of keeping teams out of the final third. However, many of the in-game trends that plague Marseille overall are also true when they lead, as their defense flags late on even if they have a lead. The completion % allowed rises late in the game and no one allows more chances as a percentage of total passes in the final quarter of a game with a lead than Marseille. This trend of tiring is getting worse as the season gets longer as well. Compare this season long plot:

to this plot of just game 21 onwards:

Opponents are setting up shop in the Marseille third even more as the year goes goes on. 3.5 passes for every pass in their own third as the game ends compared to 2.5 for the entire season. The pressure is dying off by minute 65 now.

Since game 21, Marseille is winning the ball back in the midfield and in the opposition third like an average team. The rate for their own third is similar as well. The gas light is on and the tank is nearing empty. This article is about passing patterns but those are useless if they don’t lead to shots and goals. Both of those categories climb as the end game nears.

Can we pinpoint certain players who drop off?

Sort of. We can look at tackling + INT per 90 rates in the first 60 minutes of games and compare it to the same players rates in the final 30 minutes.

Why I say sort of, is we don’t know if there is some shift to the players roles as the game goes along. If there isn’t and the players have the same roles and responsibilities throughout the game then it looks like Imbula, Morel, and Romao struggle to keep taking the ball away for all 90 minutes. Fanni, N’Koulou, Ayew and Dja Djedje do not have the same problems at all.

What have we learned?

-How effective Marseille’s press is, especially at the start of games, in disrupting the oppositions moves, forcing them into low-percentage long balls, and taking the ball away

-That if you can break the press, you have a good chance to get the ball in the center of the pitch in the attacking third and create a chance or get the ball into the box

-That the press dies down as the game goes on, those long balls turn into better propositions and the opposition can begin to set up shop in the Marseille final third as the game comes to an end.

-That Marseille can’t crank the pressure up anymore when they fall behind as they are already at eleven. They do break with the Bielsa run-at-all-costs philosophy a little bit as they sit back slightly when they take a lead. -That some players see their takeaway stats slip more than others late in games.

Are there any recommendations we can make?

Marseille have been very good this season. They sit in 4th in the table but have a good claim to being the 2nd best team in the league based on expG. They very well could make the Champions League, which would be a great result and mean it’s hard to criticize too much but there are a few things I’d wonder:

-Are Imbula, Morel, and Romao’s fitness levels not high enough to play this style for 90 minutes? If that’s the reason they are dropping off in the final 30 minutes, should 1-2 subs be used strictly to keep the high pressure up. -Does the intensity really need to be cranked to 11 from the get-go? It makes for thrilling watching but it also seems better suited to a 70 minute game and not a 90 minute game. The final 20 minutes, Marseille struggle to contain any opposing offense. If they want to keep up the heat for 90 minutes, they need better conditioning, use their subs simply for defense, or lower the intensity from the get-go.

-If they need to keep the intensity so high at the start of games and can’t condition or substitute their way to keeping plan A going strong, there needs to be a more distinct plan B. Plan B might be attacking less when they have the ball and having more tepid possession but something needs to be done to stop the late-game onslaught.

-If Marseille add a European competition in the UCL or the Europa League next year, things need even more of a rethink. There needs to be 4-5 more players added, increased conditioning from a few key players, they need to hold onto Ayew and his great defensive work, or step back the pressure bit. If they are already wearing out 33 games into this season, next year at this point they could easily be at 43 or more games. Hopefully you now know more about the Marseille press than you ever wished to before. Any questions, comments, arguments, or reactions are welcome on a comment here or @SaturdayOnCouch on twitter.

A deeper dive into the Marseille press

What exactly does the Marseille press do? We all can see the hectic effort they put in and the thrills they give neutrals with the helter-skelter games they produce, but what exactly is it getting done. In my last post, I showed there are some significant signs of slowdown among the Marseille side as they tire as the season goes along and tire in-game as well. What didn’t show up were the results of the press really. Sure, they forced takeaways at a high rate, but not near the top teams in the league which left some people scratching their heads. Several people on twitter seemed to think sheer takeaway numbers and where a teams takeaways occur might not reflect the true defensive effort Marseille are putting forward. So today, I added pass data to the mix and I think a much fuller picture is starting to emerge of how Marseille go about playing defense. First, an overview.

Here we see starkly why Marseille might lag in the raw number department: they face massively fewer passes than most of the rest of the league. Marseille face 309 passes per game, nearly 100 fewer than the average Ligue 1 team. The average Marseille game features, in total, nearly a quarter fewer passes than a typical PSG game. In the early going of games, teams complete barely 60% of the number of passes they do against a normal opponent.

When this is accounted for, their takeaway stats rise as they are taking the ball away per pass more often than most teams. For example, here is the takeaway per 10 passes attempted in the opposing third:

Now we see that instead of hovering around 3rd-5th in the press stats, Marseille and PSG are taking the ball away in a league apart from the other 18 teams. This is true in each third of the pitch, with PSG and Marseille’s edge slipping as you get closer to the defending teams goal:

This is pretty solid proof that those two teams are pushing their defense high and ball-seeking.

The in-game fading effect we saw yesterday is certainly still there. Marseille roar out of the gates but by the end of the game are taking the ball away at a standard rate all over the pitch.

One interesting stat that also slips back to league average by games end is the share of opposition passes being played in Marseille’s third. To start the game, Marseille allow a significantly higher share of passes in their own third, then after halftime pretty much track with the league average:

Clearly, Marseille are tiring or changing strategy in-game. That we established last time and get more proof of this time. However, with the added pass data we see that each pass against Marseille is more dangerous than any other team bar PSG. Up next, I’ll see if there is anything more we can learn about this press.

Taking a look at this year’s “Bielsa Burnout”

Reading this article from Mike Goodman on Grantland got me wondering about Marseille. The common thinking on Bielsa is he pushes his players too hard and can be hard to handle so after 2 years or so he moves on. Goodman titled his piece “Bielsa Burnout” and said that Bielsa’s hard-pressing, heavy-running strategy was why Marseille “had nothing left for a comeback” against PSG on the weekend in what was a crucial win for the Paris side. He said the Bielsa philosophy is to try to pin the opponent back as far as possible and this is consistent with what we see on the pitch and from most other articles about the man seen above on a a cooler that just has to be full of mini-Powerades, likely red. In the Blizzard issue 5, Jonathan Wilson said: “Although he accepts other schools of thought exist, Bielsa’s vision is non-negotiable. Repetition on the training ground translates into constant pressure and attack in matches.” On fusion.net this year Wilson said “The ideal Bielsa game will feature most of his players in the opposition half most of the time. That’s why he often uses a back three: to get as many players as possible, as high up the pitch as possible, so the pressing can be as focused and as potent as possible.”. Wilson also mentioned the inevitable burnout that comes with the early highs: “But then, a few months in, the trip comes to an end. Players find themselves mentally, emotionally and physically exhausted.”

I love watching Marseille and hate that my bet on them to win the league is fizzling out so felt it’s time to investigate this. Goodman’s statement about Bielsa’s strategy leaving nothing in the tank didn’t ring true to me, as teams who are conditioned to play high-pressing game surely should have an advantage as time ticks away as their opponents are tiring after having to deal with a relentless swarm they are not used to. Not satisfied with my admittedly basketball-influenced half thought, I took to Excel to see if I could tease out any answers.

The Bielsa Press

First, I wanted to see if this legendary mythos about the Bielsa press was really true. It’s been mentioned so many times by so many respected writers, it should be true, but I wanted to see for myself. The first metric I looked at was takeaways in the opposing half. Tackles won + interceptions. Simple enough.

4th in the league is solid, but seems low based on the acclaim the man gets. Seeing Evian so high, I figured they fact they concede so much possession might be inflating their numbers so I looked at % of takeaways in opponents half next.

Still nothing amazing. I looked further upfield to see if maybe Bielsa’s high press is a really high press. I looked at all interceptions within the oppositions third.

On twitter, Paul Riley said that I may be missing something by only counting what OPTA deemed interceptions and successful tackles and not counting ball recoveries. There is no overlap between the three categories from a visual inspection and I originally did not include ball recoveries because the description sounds like it can sometimes be counted if a pass gets deflected and the passing team recovers it. Some googling has turned me over to the camp that is is used to register a change in possession. So, thinking this might dynamite the whole thing I added them in:

Marseille actually drops further. Is the Bielsa press not as fearsome as we thought?

Not so. Maybe it’s PSG who should be getting the plaudits for the high press. Despite Marseille not getting the takeaways we would expect from the plaudits they get, maybe there is still something to the man-to-man full-field pressure they apply that wears down opponents early in games and then eventually leads to Marseille running out of gas late in games and late in the season.

The Bielsa Burnout

First let’s look at how their press changes as the game goes along. To do this I looked at the average distance a takeaway is made at. I compared these numbers in the opening 30 minutes, middle 30, and final 30 minutes of games. If Marseille’s energy is sagging late in games due to so much running throughout the game I would expect their takeaways to start coming closer to their own goal.

I added the rest of the top 4 to see if score effects are responsible for some of this and to provide context. Marseille certainly fall back as the game goes on, they start off taking the ball away a yard and a half further up the pitch than an average side and wind up equal at the end of a game. They fade more than the average team, but nothing that is shocking. Similar to Monaco.

Up next I looked to see if their defense is sinking back over the course of the season. If the cumulative effect of the running is large, then we should see Marseille falling back as the hard miles pile up on the likes of Ayew, Imbula and Romao. I broke the season down into thirds, games 1-10, 11-20, and 21 on.

We see that once again Marseille have fallen from a team who presses higher up the pitch than most to a league average team at this point in the season. Once again, it was noticeable but not shocking. Now I looked to see how the effects combined. If the cumulative effect is wearing them down, and they tire quicker than most teams during games, it makes sense that as the season goes on the later stages of games will become exhausting. The next chart is a breakdown of the just the last 11 games (not 10 as the title says) and where the takeaways come from as the game goes along.

This chart is the strongest evidence of a “Bielsa Burnout”. They have had 3 losses and 4 draws during this time so it’s unlikely to be a game-state problem. Marseille start off pressing slightly higher up the pitch than the league average team, but by the end of the game they are winning the ball back nearly a yard behind the average team.

With ball recoveries added in and the data broken up further we can see some more interesting things:

And we can look at takeaways in the opposing third by time bucket as well

This looks like the press builds and builds until the hour mark at which point Marseille stops winning the ball back deep in the oppositions side.

How does this effect the big stats

It’s not a big deal if you are winning the ball back closer to your own goal if you aren’t allowing shots or goals, but the burnout is starting to effect Marseille’s defense in those areas as well.

This chart means that at the start of the season they were holding their opponents to just over 80% of their normal expG totals. In the more recent stretch, opponents are getting nearly 120% of their average expG totals when they play Marseille. This chart comes before the PSG game which would make things look even worse.

If it was this simple, we could wash our hands and say boy Bielsa shouldn’t push them so hard, there is nothing left for the end of the season. Alas, it doesn’t seem to be because the other side of the ball is thriving.

Marseille are improving their expG totals in their last 10 games, not what you’d expect from a team dead on their feet. To complete the head-scratcher they improve as the game goes on, scoring 63% of their goals in this most recent 10 game stretch after the 60 minute mark and deserving it:

Over 50% of their expG total comes from that time period as well. Maybe this is unrelated to the defense (Batshuayi sub effect?) but it is a puzzler.

What have we learned?

The Bielsa Burnout might be a thing. Marseille are sagging deeper defensively and have trended downwards both during the game and during the season, with the end of games recently seeing them deeper and deeper in their own half. At the same time, their offense has been overpowering during that same time period, scoring 12 goals in 10 games after minute 60. If the defense is dead on their feet and the offense is scoring at will, is Bielsa using his subs to open the game up at the end and trying to hit on the counter? The games have certainly opened up, as they have allowed 7 goals in that post-6o minute zone as well over the last 10 games to make it 19 totals goals in just over 300 minutes of game time. It’s made for thrilling viewing and fascinating analyzing. We now know their press is fading late, will the late offensive output pick up the slack enough to see Marseille into the Champions League? It will be fascinating to watch and find out.

If that’s not a sure enough conclusion for you, I apologize. As with most soccer data experiments, the game is much too complex to figure out in Excel. I understand Marseille more now than I did before this and have ideas for future experiments (Athletic in Bielsa’s era is first up) and something else to watch in the upcoming games. I hope you do too.

European Top 10 Lists

A smattering of top and bottom lists involving teams shot data on offense and defense. These may be interesting to you, may be not, but I love looking at them because the more I watch and think and look at season stats the more I think the table lies. Soccer is such a low-scoring sport and such a sport of small margins that I think luck plays a larger luck in the final table than it does in most other sports. That’s is a future experiment to determine table vs performance variability in soccer vs baseball, football, hoops, hockey, etc but for now I think it’s safe to say at the least we can learn a lot about how teams play by looking deeper. Some of these stats have more repeatability than others, for example SOT% on offense is as you would expect is pretty repeatable year-on-year. Goal/SOT for offenses have much more randomness. Non-blocked SOT% for defenses has a large amount of randomness as well. A more detailed post on which of these are repeatable and which aren’t will be coming later but for now enjoy the Europe-wide tables.

Offense

Exp G per game 

Top

Bottom

Shot on target% (with distance rank, lower=better)

Top

Bottom

Goal/SOT

If Hamburg and Evian switched goal/SOT rates (and Hamburg are taking shots from closer than the French side) the Dinosaurs of German football would have 19 more goals and would not be sweating relegation. Evian would have 18 less goals and almost certainly headed for Ligue 2.

Exp G/S – who is getting the best/worst quality shots

Top

Atletico Madrid lead Europe in header% at nearly 30%. Roma are the bottom team at just over 5%. Atletico is nearly 6 times as likely to head a shot as Roma are.

Bottom

Teams outperforming xG

Do not mistake this for a lucky table. For example, Real Madrid’s shot distance and ability to put balls on target and in the goal is a several year trend which means you expect them to overperform their unadjusted expG. At some point I will post the adjusted expG numbers in which I expect Real Madrid to fall off this list. Caen and Evian on the other hand I do not. Spurs I think have been very lucky this season as well as can be seen here. Lyon and Wolfsburg are two very interesting cases as both have had surprisingly strong seasons and sit in 2nd in their leagues right now. Neither get amazingly close shots, yet put the ball on target extremely well and score at a high rate. Wolfsburg especially I feel has become overrated due to the beatdown they put on Bayern, I’ve heard a podcast where Napoli was dismissed in the Europa League because they had to face Wolfsburg. The Wolves are playing very well, but aren’t quite at the level of elite European side yet.

Teams underperforming xG

I’ve already talked about poor Hamburg here. They have since added a absolute beatdown at the hands of Leverkusen and are 3 points from safety. They have never been relegated from the Bundesliga and I hope they aren’t this year. They can complain a bit about their absolutely putrid luck at scoring the ball when they put the ball on target, a shocking 13.9%. Hamburg take shots from closer than Wolfsburg but the when the Wolves put the ball on target it is 2.5x more likely to go in. Granada, Burnley, QPR, Metz, and Almeria all sit in the relegation zone.

Defense

Exp GA per game

Top

Dortmund are the surprise here as every other team on this list is likely headed for the Champions League.

Bottom

Spurs have allowed a closer average shot than any other team in the top 5 leagues in Europe.

SOT allowed %

Top

Bottom

Bayern Munich stands out here like a sore thumb. I would guess they allow such a high SOT rate and have such a low block rate (the two go hand in hand oftentimes) due to how they play. They want to limit shots against most of the time and are trying to get the ball back with high pressure instead of getting men between the ball and the goalie. So the very few shots they allow tend to come with fewer players than most teams have back.

Goal/SOT

Top

Bottom

Exp GA/S

Top

Bottom

Block%

Here is where the English sides just absolutely crush the opposition as I’ve talked about before: https://saturdaysoncouch.wordpress.com/2015/03/02/english-defenders-getting-stuck-in-shot-anatomy-of-europes-big-league/. 8 of the top 10 get stuck in while Juve and Udinese have some of the English grit and toughness. Everton and Hull are more than twice as likely to block a shot as those soft-defenders at Celta Vigo , Freiburg and Bayern Munich. Cool to see the massive difference in approaches that Bayern and Juve take to becoming the best defenses in Europe shown so clearly here.

Top

Bottom

Non-blocked SOT%

Top

Bottom

In a category that is almost totally random when you look at the big picture as we can see below:

we see one team doing it year after year. Juve has now made it 3 straight years right near the top, obviously tied into their forcing teams to take such long shots. Or maybe the groundskeepers simply paint the boxes super huge in Turin and fool the opponents? Can’t count it out with Torino at #3. A strange nugget from this category is a shot against Celta Vigo is 43% more likely to miss the target than one against Eintracht Frankfurt, despite Frankfurt giving up a longer average shot.

Teams outperforming raw expG (allowed fewer goals than expected)

I don’t think this is a list of “lucky” teams. There are a number of adjustments that need to be made to the expGA tally before you can conclude which teams are actually lucky or not. Juve’s ability to consistently have low SOT% and low G/SOT numbers means they aren’t simply getting lucky year after year. However, Gladbach is one I am comfortable saying fortune is smiling upon them. The Foals goalie Yann Sommer has saved nearly 87% of on-target shots. They are generally coming from a long distance away but the 10th place distance team, Guingamp, allows on target shots to be goals at a .292 rate. If the Foals allowed goals at that rate then they would have allowed almost 20 more goals this season and certainly would not be locked into next years Champions League and the $25+ million that will come with it. Maybe this is how they get back to their rightful place of dueling with Bayern: a sizzling goalie and poor opposition shooting helping them to the UCL then they kick on. Wishful thinkings maybe, but would be nice to see. And if you like reading about the history of soccer at all I highly recommend Tor! by Uli Hesse, it’s an absolutely brilliant book that has filled me in on all kinds of German soccer history I knew very little about. Great read, not a slog at all.

Teams underperforming expG (allowed more goals than expected)

No surprise for people who have followed the Bundesliga to see Dortmund way up in the unlucky category. They and Napoli could miss out on big-time UCL money because they can’t keep teams from pouring the ball into the net.

Are Spurs* bad?

sorry Harry, not you

*except for Harry Kane?

I am not a Spurs fan and watch only snatches of their games. So my perspective as an outsider, listener to experts (including all those on the 606 Phone-In) and peruser of league tables was that Spurs had improved from last season. My impression was Pochettino is a much better manager than Sherwood and Kane is on fire so of course Spurs are doing better than last season. They are doing fine in the table, only recently falling out of the race for the Champions League and made the League Cup final as well, sliding them solidly into the “solid season” portion of my mind. Then I started looking a little deeper at the stats and it became clear my perceptions were not correct. Spurs are not a good team. They might even be a bad team. And they are heading in the wrong direction when you look at previous seasons. You might not agree with me, but let me make my case.

Exhibit A: Spurs get awful shots

The raw shot numbers and raw goal numbers are misleading in Spurs case. They fire at the 6th-highest rate in the league and have scored 50 goals, giving the illusion they have a strong offense that can generate good opportunities alongside good volume. They do not. Spurs shot quality ranks 18th in the league mainly because they take shots from a long way away (distance in feet):

Unsurprisingly, they take 2nd most long shots in the league (70 feet or more)

…and the 2nd fewest close shots (30 feet or less from the center of the goalmouth)

When you look at where Spurs are shooting from, you wonder how they don’t have one of the worst offenses in the league. The main reason is they are good at putting the ball on target (7th) despite taking those bombs. When they do put the ball on target, it goes in a lot (3rd in the league). If Spurs were just league average at putting the ball on target and getting it to go in when on target, they would have 7 fewer goals.

Exhibit B: Spurs allow great shots

Spurs allow fewer shots than they take (by about 1.5 a game) but the shots they allow are as high quality as the ones they take are low quality. They allow the closest average shot of any team in the league (measurement in feet again):

With this comes a high number of close shots:

and a small number of long, speculative shots:

For every 10 long shots Tottenham allow, their opponents get nearly 9 close shots. Meanwhile, on offense Spurs get just 2.5 close shots for every 10 long ones they take. Unsurprisingly, Spurs allow a higher shot on target% than any other team. They also block very few shots (19th most in EPL). This shouldn’t really surprise anyone, they have allowed 45 goals. This is basically just clarifying the fact that they haven’t been unlucky to allow those goals, their defense is just not any good.

Exhibit C: Game state stats help Spurs shot numbers

Looking at traditional, unadjusted numbers like TSR are somewhat favorable to Spurs. They rank 7th in total TSR. However,  teams generally shoot more when they are behind: https://saturdaysoncouch.wordpress.com/2015/03/03/game-states-across-europe/

and Spurs are behind a good amount more than the teams around them in the table. Liverpool and Southampton have both led nearly for nearly twice as many minutes as they have trailed this season while Spurs have trailed more than they have led, leading to them spending a good amount of time in a negative game state which helps with their raw shot total numbers. West Ham, Everton, and Swansea have also spent more time ahead. In an even game state, Spurs have taken 163 shots and allowed 158, which is 8th in the league in total shots ratio but as close to 11th as they are to 7th.

Those three pieces of evidence: they get awful shots, allow great shots, and have their shot totals boosted by trailing a lot, seem to imply that Spurs should be 8th at best and maybe down around the low-mid teens if you think they get a bit of bad luck. Why aren’t they so low? Of course, it’s Harry Kane.

If you removed Kane’s numbers, the rest of Spurs would drop from 7th to 17th in SOT% and 3rd to 15th in G/SOT. That’s not really a fair comparison because someone else would be getting at least the chance of getting the higher quality striker shots but it tells you how Kane-reliant the team has been. A fairer comparison might be comparing expG to actual goals.

Is Harry Kane this good? No one is this good. His shooting and scoring rates match up favorably with some of the best in the game but scoring almost 2.25x your expG is not sustainable. I can’t find a post from the end of last season but I’m almost positive either Ted Knutson or Michael Caley talked about Kane converting shots at an elite level for a significant sample size and that was before this madness. With that vague recollection done and dusted, I am ready to say Kane is a very good striker. He’s very good, but he’s also probably going through the best stretch of his career right now.

So where does that leave Spurs going forward? They can’t get good shots, allow very high quality shots and can’t dominate the shooting rate when playing at an even game state. They’ve got a great striker who is running hot right now carrying them, and their goal difference is still just +5. What happens if Harry regresses, cools off, or gets injured? What happens if Spurs run into a few hot goalies and their goal/SOT rate falls to middle of the pack and they face teams that start shooting well and convert goals at a high rate? I’d say they could “pull an Everton” except if you look at the underlying stats, Everton are arguably playing better than Spurs right now. Wigan played at a level somewhat comparable level to Spurs back in 2012-13 and they got relegated. That’s obviously the worst-case scenario and Spurs have a much better ability to get the ball on frame than Wigan did so relegation is not really a worry (though I will be taking a peek at how long the odds are next season), but a bottom half finish is what you expect when you play like this. Spurs have tons of work to do to get their team anywhere near the Champions League.

Spurs expG% trends (ratio of exp GF to exp GF + exp GA)

Other EPL teams with similar expG% to this years Spurs team and their final spot in the table:

On twitter someone mentioned looking at Saints to see how Pochettino’s full year there matches up with this season at Spurs compared to how last season at Spurs correlates with this year:

Spurs already got very bad shots last season (though they got a good amount more per game last season) but the major change is they give up almost 60% more close shots this season. Saints have increased their close shots by almost 50%.