Vertical Passing 2.1: Distance and Clustering

After defining what a vertical pass is in Vertical Passing 2.0, and being satisfied with the results, the next step is to start breaking these passes down and looking at the different types of vertical passes.

This came about after a short conversation with @SaturdayOnCouchwho has also done some interesting things looking at different pass types over on StatsBomb. This piece will use a similar method to the ‘Finding the Best Pass in the Bundesliga’ , as it’ll break down the vertical passes and then look at how players compare to the average completion rate across the four leagues.

Firstly it’ll just be a simple look splitting the passes into two different distances, before looking at clustering the passes and finding the most common vertical passes.

Distance

Looking into the distance of the vertical passes, I kept it simple and just had passes that were under 15 units in distance and over 15 units in distance. I mostly just wanted to separate the players who send long line-breaking passes from those who advance the ball slightly, the more detailed look into different types of vertical passes will come with the clustering.

The player who attempted the most longer distance (>=15 units) vertical passes was Bayern Munich’s Mats Hummels, which shouldn’t come as a huge surprise. However, the most impressive numbers for longer distance passes comes from Inter Milan’s Milan Skriniar.

Using the average completion of these longer distance passes as the expected completion sees Milan Skriniar complete 14.475 more passes than he was ‘expected’ to – the most in the four leagues used. He completes these passes at a rate of 84.6%, almost double the 47.5% average. You can see a map of these passes below:

Click to Enlarge

The top 5 for most passes above expected can be seen below.

Defenders

Player Team Completed – Expected Passes
Milan Skriniar Inter Milan 14.475
Kalidou Koulibaly Napoli 12.575
Francesco Acerbi Sassuolo 12.475
Sebastian Langkamp Hertha BSC 10.9
Mats Hummels Bayern Munich 10.9

Midfielders

Player Team Completed – Expected Passes
Dani Parejo Valencia 10.425
David Silva Man City 10.1
Asier Illarramendi Real Sociedad 9.75
Kevin De Bruyne Man City 9.25
Fernandinho Man City 8.15

At the other end of the scale, the player who had the lowest difference between expected and actual passes completed was Liverpool captain Jordan Henderson. Henderson has come under a lot of flak lately, particularly given Liverpool’s win percentage with and without him, and these numbers don’t help his case.

Henderson’s completion rate for longer distance passes was 25%, only just over half of the average, leading to him completing 9.9 passes less than ‘expected’.

Obviously, this isn’t the best way to measure as those who pass more will get penalised more.

I toyed with looking at ‘extra’ passes and ratio of completed passes compared with the expected completions. While the ratio will be normalised, even with filtering for a minimum number of attempts, the results tended to favour those who play fewer passes more accurately, than those with a high number of attempts with above average completion.

See the relationship of both attempts and ‘extra’ passes or completion is more useful, and will be seen later on with some graphs.

Another problem is that it’s still lumping in a pretty big range of passes given the difference between a pass with a distance of 15 and one with a distance of 50+. Bands of distance would have been a more appropriate approach, but with that being said Henderson’s longer distance pass map doesn’t make for great viewing:

Click to Enlarge

The top 5 for most passes below expected can be seen below.

Defenders

Player Team Completed – Expected Passes
Danilo Udinese -7.4
Ciaran Clark Newcastle -7.3
Tommy Smith Huddersfield -6.55
Djene Dakonam Getafe -6.425
Aymeric Laporte Athletic -5.975

Midfielders

Player Team Completed – Expected Passes
Jordan Henderson Liverpool -9.9
Marcus Rohden Crotone -6.6
Stefan Ilsanker RB Leipzig -5.6
Daniel Baier Augsburg -5.525
Nabil Bentaleb Schalke -5.175

Doing the same process for passes with less than 15 distance and throwing it all into a big graph can be seen below:

Click to Enlarge

It helps show just how great Milan Skriniar and Kalidou Koulibaly have been playing vertical passes out from defence, but doesn’t do a great job with short distances for attacking players. Lionel Messi, Kevin de Bruyne and Christian Eriksen are 3 names that seem out of place for completing less passes than expected.

This is presumably because their short distance passes are coming in and around the final third, likely packed with opposition bodies. I even excluded passes that start and/or end in the box to try and even it up, but it had practically no effect.

This is where clustering should become useful to differentiate between the different types of vertical passes.

Clustering

I used k means clustering to group the passes into 15 different groups (I’m still not sure on this number, I just used it as 10 felt too few and 20 felt too many), resulting in the following cluster centers:

Click to Enlarge

As you can see, there’s some pretty interesting passes and separating our vertical passes into these clusters can further help find how players pass.

I’ll be referring to the clusters by the numbers labelled on the pitch as I can’t really come up with names based on the zones. I’ll also be grouping some clusters together when they seem like the same pass just from the other side, so 4 and 8 or 5 and 9 for example.

It’s also worth pointing out because we’ve divided up by so much, plus there’s a lot less vertical passes than regular, some of these sample sizes are really small.

Final Third – Types 4, 5, 8 and 9

After mostly looking at defender distribution for vertical passes I thought I’d start the clustering section with some attacking passes.

Type 8 and 4 passes are one you typically associate with number 10’s or playmaking wingers, finding a bit of room in the half space and then playing a pass between the defender and centre-back to either a winger or overlapping full-back.

With that in mind, it should come as no surprise that David Silva dominates this type of pass. The Spaniard has recently been getting lots of (deserved) praise and these types of clever passes in and around the box are a huge part of his game.

The below graph shows the number of attempted passes of these types and completion rate. Silva’s completion rate for the number of passes attempted is pretty ridiculous. (I apologise in advance for the size of the graphs, it was the only to make sure they weren’t a complete mess.)

Click to Enlarge

Praise should also go to Arsenal defender Sead Kolasinac as him and Inter’s Danilo D’Ambrosio seem to be the only defenders who stand out. Kolasinac has the 3rd highest number of extra completions, completing all 14 of his attempted passes of this type.

While it’s only for one type of pass, it’s also strange seeing Jorginho fall below the average completion for anything involving passing.

Arsenal pair Aaron Ramsey and Alexis Sanchez are also below average for completion, with the latter getting some criticism for his final third passing in recent games.

Roberto Firminio is also in an interesting position, usually flanked by players like Mohamed Salah and Sadio Mane, you’d have thought he’d have both more attempts and a higher completion rate.

Overall, there doesn’t seem to be many interesting talking points, the players who excel at this type of pass and do it often are pretty much the usual suspects.

Looking at younger players wasn’t very fruitful either, the player with the most attempts was Hector Bellerin with 18. A lot talented youngsters have good completion rates (Richarlison, Julian Brandt, Goncalo Guedes, Andrea Petagna for example) but have no more than 12 attempts.

Similar to how David Silva dominated the previous passes 4 and 8, type 5 and 9 is dominated by Lionel Messi. Interestingly his completion isn’t significantly above average, but he’s attempted it 18 more times than next closest Cesc Fabregas. Just take a look at the graph below:

 

Click to Enlarge

Like Jorginho for types 4 and 8, it’s surprising to see Fabregas be anything other than great for something passing related.

Another interesting name in a good position is Luis Alberto. Despite his time at Liverpool feeling like ages ago, he’s still only 25-years-old and seems to be improving at Lazio.

Using StrataBet data (up to date as of 7th December) and defining expected goal contribution as the xG value of his chances, chances created and chances created where he’s provided the second assist, Luis Alberto has an impressive xG contribution of 0.776 p90, while his actual numbers are slightly more than this at 0.911 p90. Removing shots he’s taken and focusing just on those he’s created (both primary and secondary) sees him have the 8th highest xG created in Serie A with 0.621 p90.

It’s worth pointing out 0.300 of this 0.621 is from set pieces, which is the highest in the division. Given that he’s only started to come into form this season it’ll be interesting to see if he can maintain it over the course of a season. With that being said, given Tottenham have been linked with a move for West Ham’s Manuel Lanzini, presumably as a rotation option to Christian Eriksen, Luis Alberto could represent an interesting alternative.

Alberto’s xG  contribution is marginally behind Eriksen’s (0.823 p90), but he actually creates more chances and secondary chances p90. Alberto has created significantly more from set-pieces than Eriksen this season, (0.300 xG p90 vs 0.066 p90, 1.442 chances p90 vs 0.576 p90) it’s mostly xG p90 where Alberto falls short of Eriksen.

Alberto seems to be enjoying football at Lazio and might not want to play second fiddle to Eriksen, but given the likely fee involved with Lanzini, it could be an interesting avenue to explore.

There’s also a few deeper midfielders who give good accounts of themselves above, Southampton’s Oriol Romeu, Manchester City’s Fernandinho, Watford’s Abdoulaye Doucoure and Inter’s Matias Vecino. Then despite Milan’s season looking disappointing, 20-year-old Franck Kessie performs these passes superbly, particularly compared to other under 24’s, as can be seen below:

Click to Enlarge

It’s also impressive that Stuttgart’s Timo Baumgartl has attempted 8 of these passes despite being a centre-back, let alone completed all of them.

Middle Third – Types 1, 2, 3 and 13

After having the uncomfortable experience of producing a graph involving passing and not have Jorginho leading the way, you’ll be glad to know in the middle third we’re back to normal.

For pass types 1 and 13 Jorginho comfortably leads the way, followed not all that closely by team mate Marek Hamsik.

As you can see on the graph below, the stand out players are the two previously mentioned Jorginho and Hamsik, as well as Granit Xhaka (despite being slightly below average completion), David Silva, Kalidou Koulibaly and then for all the wrong reasons Gabriel Pires.

Gabriel Pires also had some interesting numbers in Vertical Passing 2.0. From his stats he seems a player very much willing to take risks, but these risks may not always pay off.

His xP Rating of 0.783 is pretty shocking, he only completes 27.3% of his passes into the box and 43.5% of his attempted vertical passes. However, his xG Contribution of 0.490 for a Leganes side that has only managed 11 goals in 14 matches is pretty impressive. It seems as though he has talent, but he’s very raw. Seeing how these numbers change over the season should be interesting.

Staying in La Liga, while passes type 2 and 3 are dominated by centre-backs, one midfielder who performs well is Celta’s Stanislav Lobotka, as can be seen below:

Still only 23-years-old, Lobotoka joined Celta for £4.5m in the summer and has made a great start to La Liga life. The above shows he’s more than capable of making vertical passes in midfield, add in a completion rate on all vertical passes of 83.3% plus an xP Rating 0f 1.137 and you have a very talented young midfielder.

With Lobotka, Pione Sisto, Maxi Gomez, Jonny and Emre Mor Celta possess some exciting young talent, coupled with being more attacking this season (see the below tweet), they could be a fun team to keep an eye on.

Filtering the above graph for under 24’s makes Lobotka stand out as a midfielder even more, while Sampdoria’s Lucas Torreira attempts more his completion rate is slightly below average. Stuttgart’s young defender Benjamin Pavard also performs well for this type of pass.

Click to Enlarge

Defensive Third – Types 6, 10, 12, 14, 15

Most of the talked about vertical passes so far have been from defensive players, with Vertical Passing 2.0 pointing out the talented young Stuttgart defenders Timo Baumgartl and Benjamin Pavard. With that in mind I’ll keep this part brief as it’s much of the same, usually dominated by Kalidou Koulibaly, Milan Skriniar, Miranda or John Stones.

Starting with pass types 12 and 14 (which might actually fall into the middle third) see’s Koulibaly really shine:

Click to Enlarge

However, filtering for under 24’s sees Timo Baumgartl stand on his own:

Click to Enlarge

Another young player who seems to be above average completion for most pass types, with not huge but still respectable number of attempts, is Villarreal’s Rodri. The 21-year-old midfielder has an xP Rating of 1.133, 3.926 interceptions and successful tackles p90 and only performs below average in pass types 4 and 8, but even then only marginally. Between Rodri and also 21-year-old Pablo Fornals Villarreal could build a nice midfield partnership.

Interestingly Baumgartl excels with the previous pass type, but what looks like the easier/shorter version of those types – types 6 and 10 – he actually performs below average on, as does team-mate Benjamin Pavard. Again these pass types are occupied by the usual suspects, although Davide Astori has attempted this pass the most and with a good completion. Sevilla’s Clement Lenglet also has some pretty impressive numbers, particularly when comparing u24s.

Click to Enlarge

I’m going to leave out the graphs for pass type 15 as not many players attempt it (only 8 have >= 10 attempts) but all you need to know is John Stones and Milan Skriniar are head and shoulders above everyone else for it. Both attempting over 20 with a completion rate over 90%.

I’m also going to leave types 7 and 11 as they’re mainly goalkeepers. Goalkeeper distribution can be interesting but it’s a different discussion altogether.

Bringing it all together

While it’s interesting to break these passes down and into different types and see which players excel at them, I wanted to summarise it all and have a simple way of showing how good players are at completing their vertical passes.

What I decided to do was add together all the extra completions that a players makes for each pass type (calling it Net Extra Completions) then divide this number by the number of attempted vertical passes to get the percentage that a player over (or under) performs on their vertical passes.

Throwing this all into a big graph gives the following:

While you can spot things that have been mentioned throughout (Koulibaly, Stones, Jorginho, Skriniar and Umtiti are good at vertical passes, Henderson and Gabriel Pires under perform) this isn’t all that useful. So let’s trying breaking it down and picking out some highlights.

Defenders

Click to Enlarge

Outside the names you’d expect in that top right corner one player that has great numbers is Hoffenhim youngster Kevin Akpoguma. He only has 495 minutes under his belt, but the 22-year-old is among good company in that top right corner, even attempting more vertical passes and having a better ratio than Stuttgart pair Timo Baumgartl and Benjamin Pavard.  Filtering for those under 24 helps show how impressive Akpoguma’s numbers are as well as highlighting some more talented young players.

The previously mentioned Baumgartl and Pavard look impressive, then similar to Akpoguma RB Leipzig’s 18-year-old Ibrahima Konate has impressive numbers albeit from a small number of minutes (474). Another Hoffenheim player Stefan Posch also has good numbers, with slightly more minutes than Akpoguma having played 578.

There’s also a gathering of young defenders, including Chelsea’s Andreas Christensen and Tottenham’s Davinson Sanchez, who have a great ratio but don’t attempt vertical passes all that often.  19-year-old Leverkusen defender Panagiotis Retsos in particular has a great ratio, joining from Olympiakos for ~£15m in the summer he looks as though he could be a good buy for a Leverkusen side already stacked with young talent.

Midfielders

Click to Enlarge

Looking into midfielders and Napoli’s Jorginho leads the way with both a huge number of attempts and a great ratio. Fabregas isn’t far behind the Napoli midfielder but his ratio isn’t that much over average.

One name that has popped up as impressive throughout most pass types, and again overall, is Matias Vecino.  Having an xP Rating of 1.140 is hugely impressive, both him and Milan Skriniar seem to have greatly strengthened the spine of Inter’s side and helped manager Luciano Spalletti get off to such a great start.

Filtering for u24’s also gives some interesting results.

Click to Enlarge

Again Lobotka and Rodri shine, while Gabriel Pires struggles. Lucas Torreira and Bournemouth’s Lewis Cook have good numbers and despite signing a new contract in the summer, you’d expect more clubs to be in for the former in the next 12 months.

22-year-olds Bryan Cristante and Jean-Philippe Gbamin both look interesting prospects despite under performing the average here. Another Napoli player in Piotr Zielinski and his countryman Karol Linetty have impressive numbers too.

Another interesting name from Serie A is Erick Pulgar. When I first started looking into vertical passes with the older method, Pulgar’s name popped up as having an impressive number but his xP Rating was marginally below average. This season, however, his vertical passes are still impressive with the new method while his xP Rating has risen to an impressive 1.049. Whether this is just a good start that will later even out remains to be seen, but still only 23-years-old he could be a player worth following.

A final name to mention is that of Fabian Ruiz. The 21-year-old Betis midfielder has recently been linked with Stoke and allegedly has a release clause of £14m, if he can keep up his current level of performance it might be worth a club activating that release clause. He’s only played 533 minutes, and while he doesn’t attempt that many vertical passes the ones he does attempt he’s great at completing. His xP Rating of 1.051 is impressive, as is 1.857 attempted passes into the box p90, 1.857 successful take-ons p90 and 2.702 interceptions + successful tackles p90.

Conclusion

Breaking down passes further using things such as clustering can create some interesting results, as well allowing to better quantify vertical passes and see how players perform at completing them.

Again I think it’s important to emphasise these vertical passes don’t link with performance. Teams who pass vertically relatively more than others don’t perform better because of it, it’s more stylistic. It could be a useful way of finding players for teams who pass vertically however, for example we’ve seen where Kalidou Koulibaly makes his vertical passes and how he performs with each type which could make it easier to narrow down the search for similar passers – even more so if the clustering is extended to all passes and not just vertical ones.

Clustering is something I’ll most likely use again as it can bring some interesting findings, as well as making it easier to compare players and teams. Some other examples of using clustering can be seen on Chance Analytics looking at shots, again by @SaturdayOnCouch over on StatsBomb and @FootballFactMan clustering the location players received the ball to compare Gylfi Sigurdsson and Ross Barkley.

This article was written with the aid of StrataData, which is property of Stratagem Technologies. StrataData powers the StrataBet Sports Trading Platform, in addition to StrataBet Premium Recommendations.*

*Chance data used for xG from StrataData, not the passing data.

Leave a Reply

Your email address will not be published. Required fields are marked *