Michael Bradley’s 81st minute winner for Roma away to Udinese on Sunday afternoon was an historic moment.  It enabled Roma to notch up their ninth consecutive win at the start of a Serie A season.  This feat has only been achieved once in Italy, via Juventus in the 2005/06 season, or never before, if you prefer not to see an asterisk when looking at past achievments.

With not hours enough in the week I’ll confess to not having seen much of Roma this season but I thought I would see what my numbers can tell us about their wonderful start.


Defence

A cursory look at the league table will tell us that with just 1 goal conceded in 9 games Rudi Garcia has ensured that Roma has delivered an almost perfect defensive performance.  As is our style, we need to go a little further into the numbers to see if the concession of single goal has been deserved or whether they have been the beneficiary of large chunks of good fortune.

Defensive Shot Chart

Roma Defence ShotChart

Roma’s defensive shot chart is impressive.  Although the concession of 107 shots (excluding penalties) is good, it places them in just 4th place in terms of the fewest amount of shots allowed in Serie A.  Juventus top the table with a really stingy total of just 70 shots given up during their first 9 games; Sampdoria and Inter complete the teams that have conceded fewer shots than the league leaders.

Roma’s defensive strength has been in the way that they have forced the opposition to take long range shots.  Only 21% of shots against them have come from the Prime Zone (defined below); this figure is the lowest in the league.

Shooting Zones

Their staunch defending around the areas of the penalty area can be further seen by the fact that by allowing just 27% of their shots conceded to be hit from the Secondary zone, they ensure that a massive 51% (roundings) of shots targeted at Morgan De Sanctis originated from the two outside zones.  Inter trail a long way behind in second place in this metric at 41%.

So, in terms of defensive shooting zones Roma have been superb.  In fact, so good have they been at preventing teams from taking shots at goals from favourable positions that they have the lowest average ExpG value per shot conceded in the Big 5 leagues this season.

ExpG Values

To expand a little on ExpG for those unfamiliar with the concept, we assign a probability of each shot being scored.  This probability is determined with reference to all of the information that we have about a shot; the information is obtained from Squawka and StatsZone.

There’s obviously something special going on with this team in terms of defence.  This can be seen in the chart below which plots the average ExpG values for the teams across all Big 5 leagues that have have had the lowest average ExpG values for the shots they have conceded.

Roma AvgExpG Big5

Not only do Roma lead this measure, which is quite a feat in itself, but the difference they have over the other leading teams is clear to see.  Even with removing the scale on the y axis, we can see how closely bunched the teams that follow behind Roma are on this measure.  This is always a really good indicator of a team that is head and shoulders clear of the chasing pack.

Other than the shooting opportunities they give up, if I was to point out one other facet of the defensive play that they have excelled it, it would be their ability to block shots.
Roma have blocked 41% of all shots they have allowed.  Only one other team in Italy has blocked more than 30% of the opposition shots, Catania at a clip of 32%.

This large amount of blocked shots leads to an interesting anomaly.  De Sanctis, as the Roma goalkeeper, has saved just 17% of shots taken.  Unbelievably, with just 18 saves he has actually had the fewest number of saves in Italy.  That is certainly not what you’d expect the goalkeeping stats of the team that has just conceded 1 goal in 9 games to look like.

Defensive Style

It may be as a result of the fact that they have spent huge amounts of time in leading game states, but Roma’s defensive style has certainly been one of keeping it tight and compact at the back.  They do not press high up the pitch at all, they are content to let the opposition have the ball in non-threatening positions and reorganise themselves in their own half.

Here follows the defensive actions (tackles, interceptions and fouls) carried out by Roma in their last home game, against Napoli.

RomaTackles

RomaInterceptions

RomaFouls

During this game Roma only had 4 defensive actions (1 foul, 2 interceptions and 1 tackle) in their attacking third.  This game was by no means a one off in this regard as they consistently have amongst the fewest defensive actions at the top end of the pitch in Serie A.  The proviso here is that we wouldn’t expect a team that has led for so many minutes to do much pressing.  However, I think this lack of pressing is important as it partly explains why they have been able to block so many of their shots faced this season.

Defensive over performance

However, even with the exceptionally low average ExpG defensive values that they have posted, they have still massively over performed in allowing just 1 goal to be conceded.  Based on the objective ExpG calculations Roma would have been expected to have conceded a few more goals on top of that.  Given that their goals against tally currently stands at 1 it’s not the bravest call in the world to suggest that they will see some regression in the near future.  But all I can say, is that they will……..

Outcome of Shots Conceded

RomaShotsOutcome

The above image shows the location, as well as the outcome for each shot that Roma has conceded.  Apart from the sheer volume of dots outside the penalty area what is striking is the 7 blue shots that have been taken from central locations in an around the edge of the 6 yard box.  The fact that they are coloured blue means that they missed the target.
The wayward placing of what should be great shooting opportunities for goal scoring from those shots will go a little towards explaining why their concession of just 1 goal is much less than they “should” have conceded.

For what it’s worth, according to our ExpG values Juventus should have conceded fewer goals than Roma.  Yet the Turin giants have conceded 9 (excluding penalties) to AS Roma’s 1.  Make of that what you wish.

I’ll make just one final comment on the ExpG values of the shots they have conceded.  As I have stated in all similar articles, our ExpG values do not take account of defensive pressure.  The data that is currently displayed by the likes of Sqauwka and StatsZone do not include any defensive information at all.  The data does not know whether there were any defenders pressuring the striker as he took the shot or whether the defensive team had a significant number of defenders back.

As Roma have blocked such a high proportion of the shots they have faced, perhaps the defensive pressure they are exerting on shots they allow is higher than average.  If this is the case then the ExpG values that we assign to the shots they conceded may well be over-estimated as our model is not capturing exactly how any given team defends.

 

Attacking

For me, Roma aren’t nearly as interesting going forward as they are in terms of trying to understand how they defend.

Attacking Shot Chart

RomaAttackShotChart

With just 22% of Roma’s shots being hit from Prime Locations they actually have the second lowest proportion of shots from this optimum location in Serie A; Livorno are the only team worse than Roma on this measure.  This fact certainly isn’t what we’d expect to see of a team that is top of the table.
They have slightly more shots (excluding penalties) than average with 130 against the league average of 119, and this mix of shot quantity with their shot quality results in a forecast position of 9th in the league in terms of ExpG scored.   Yes, in terms of the number of goals scored there should have been 8 teams that scored more goals than them, but there has only been one that has actually done this.

It will therefore come as no surprise for me to say that through scoring 18 goals from non-penalty situations they have comfortably exceeded the amount of goals that our model had expected them to score.

RomaShotsForOutcome

The majority of Roma’s over performance can be attributed to their long range shooting.  Their total of 6 goals from outside the penalty area is a league best total (jointly held with Napoli), this compares with a total ExpG value per our model for these shots of 2 goals.  Perhaps their style of play; compact defensive shape and then quick transitions to counter attacking football, means that due to the relative lack of defensive pressure the quality of the chances they create are better than our model gives them credit for.
However, even bearing this in mind I’d be strongly of the opinion that their goals scored tally of 18 (excluding penalties) makes them look much better than their bare numbers would suggest from an attacking point of view.

Placement of Roma’s shots

At this early stage of football analystics, not enough work has been done to be able to conclude whether a team or a player can, in general terms, repeat good shot placements.  Personally, I think it is reasonable to accept that the very best strikers demonstrate a specific skill which allows them to pick out great shot placements.  However, on a team level I’d need some persuasive evidence to suggest that this phenomenon exists.

RomaShotPlacement

The above Shot Placement chart suggests that Roma have been very effective at picking out spots that goalkeepers find it difficult to save from.  I think this goes a little to explain why their conversion rate is 14%, which is second highest in the league.  This is conversion rate is even more astounding when you consider that they have had the second fewest proportion of their shots from Prime shooting locations.

The question is whether they can continue to be as efficient with their shot placements given the distances that they have tended to shoot from.

Summary

So, after my detailed look at Roma I hope that this article will help Statsbomb readers gain a better understanding of how they have managed to capture 27 league points from their opening 9 league games.  With that in mind, here’s my wrap-up thoughts on their performances to date:

  • They are doing something very special in defence.  The quality of the average chance they concede is the lowest in the Big 5 leagues.  That is a serious accolade to have on your mantelpiece.
  • With a blocked shots percentage of 41% they get a huge amount of bodies back when defending which constricts the space available to opposition forwards.
  • But even with these terrific defensive traits they have still over performed significantly in defence to have conceded just one goal.  Given the amount of shots they concede their goals against tally will start to increase.
  • Going forward they appear to be average, with a slightly higher than average volume of shots being offset by generally poor shooting choices.
  • They seem to have led a charmed life with the number of goals they have scored, especially from long distance shots.  I’ll be hugely surprised if this continues.

 

  • marlon

    Would it be better to use total ExpG rather than average to encompass a teams defensive strategy? A team that shells might give up more low quality shots and a possession team might give up fewer, higher quality shots. To fairly compare them, total ExpG could be better than average. (Obviously in this case you were making a different point which called for Avg ExpG).

    Would be really good to see a total ExpG For/Against chart for European leagues (like Ben Mahew does with shots/Shooting %)

    • Colin Trainor

      Marlon,
      I’m careful with my use and publication of total ExpG values. They form the basis of my betting models, and so this is information that I can’t share until the end of the season.

      However, in saying that I’m not sure it was neededed for this article. I “talked through” the actual ExpG values both offensively and defensively without explicitally stating their values.
      In this piece the average ExpG value is what was really interesting for me in terms of Roma’s defence. The total values wouldn’t have shown me that the average chance Roma concede is the best in Europe (from their point of view).

  • Toshack

    Colin,
    Excellent analysis! Let’s see what regression ot the mean means for their goalscoring – as you point out.
    I was surprised to read that Napoli was not in the top four when it comes to fewest amount of shots allowed in Serie A, seeing as Benitez is the manager. I guess that being no. 3 in Italy and no. 5 in Europe for Avg ExpG conceded is more important (and should reflect Benitez’ traits better)?
    Will be interesting to see how far the Italian teams can go in CL and Europa League this year.
    Cheers,
    Peter

  • Frank

    I’ve noticed that many of Roma’s distance shots have come from ‘lay-off’ passes. I’d be curious to see where the passes are coming from when they take shots from non-prime positions compared to other teams. This article hints to not having defensive pressure data, but this may hold some answers. Obviously beating someone off the dribble and taking an ambitious shot is not going to be as effective as having a teammate draw the defense and/or goalie and then playing a pass back for a shot.

    • Colin Trainor

      Frank, it’s certainly doable to tie where passes where played from to create each chance but there is some serious amount of additional work required to get to that point.

      Unfortunately, this tends to be the case once you go anywhere beyond the most basic metrics. The data just isn’t available in an easy to digest format. Huge amount of effort is required to get the data to the point where this article could be written, but you’d be talking about an another huge body of work to answer the question that you have posed.

      • Frank

        I appreciate your reply, and the amount of research required to write an informative and interesting article. I admit I’m naive to the data mining process for soccer analytics, so I hope you didn’t take my comment offensively. I was merely suggesting another possible reason for these outlying statistics.

        I do think it’s somewhat funny that all this crunching of data is done to come up with a pretty common diagnosis. (Roma will not continue winning every game, will concede more goals, and won’t keep being so efficient offensively.) Most articles I’ve read, are written when the statistics point to a conclusion that is not widely assumed. I think that it is extremely helpful to bring to light how statistics can be used to support a widely held belief like you have done here.

        Thank You!

        • Colin Trainor

          No offence taken Frank.

          Re your point of arriving at the obvious conclusion, well yeah, when I analyse a team that has conceded just 1 goal in 9 games then I guess there is no other conclusion that could be reached.

          However, my work was targetted at trying to understand WHY they had only conceded one goal. I think that analytics should help us understand why things happened, not necessarily to arrive at a narrowly held view.
          For example, before I undertook the research for this piece I didn’t know that Roma had the lowest average ExpG value for shots conceded. That fact alone helps us to frame their results.
          When you throw in the large amount of blocks they made on the shots they have faced we can begin to piece together how Roma play the game. And that’s even without seeing any game tapes at that stage.

  • Nick

    Damn, I wish i had more time for this amazing stats stuff.

    Great article, enjoyed it!

  • Simon

    Hi Colin,

    Nice read once more. I’ve seen about half of Roma’s matches this season and these numbers are in accordance with what I’ve seen on the field. One of the weak spots for Roma this season will be the lack of depth in their squad and it will be amusing to see how the media will fully credit the decline to the (inevitable) absence of key pleyers instead of partly to that and partly to regression to the mean.

    I fully understand you are not willing to disclose all the numbers, but could you indicate (without giving the exact numbers) how the over performance in both offense as defense in ExpG2 values relates to the ExpG values? Due to the high amount of blocked shots and low amount of shots on target I assume these tend to be much lower and therefore the over performance much smaller.

    Once again, very interesting read and looking forward to see more articles about teams in the other Top 4 leagues.

    Cheers,

    • Colin Trainor

      Simon,
      Thanks, glad that the findings seem logical to someone that has seen a lot more of Roma than I have this season.

      I haven’t actually ran the ExpG2 values yet for Serie A this season, it would be interesting to see the findings (as you say).