Welcome to Get Goalside! I was hoping you’d be able to make it.
This week’s charity is, well. Usually I’ve come across charities that I then point to here. Wiley’s antisemitic tweets have made me realise I don’t actually know any anti anti-semitism charities.
The Antisemitism Policy Trust have a couple of guides on myths and misconceptions about Jews and antisemitic imagery and cartoons that seem informative. If you wish, please let me know of other resources or charities. By the time this email goes out, I’ll be 27 hours into a 48-hour ‘walkout’ on Twitter that some are doing.
Please, let us all do better.
In the summer of 2015, Aston Villa spent (according to Transfermarkt) nearly £60m on:
- Jordan Ayew
- Jordan Amavi
- Adama Traoré (yes, that one)
- Jordan Veretout (lotta Jordans)
- Idrissa Gueye
- Rudy Gestede
- Scott Sinclair
- Joleon Lescott
The last three of this list were bought from English clubs; the first five were bought from France or, in Traoré’s case, the Barcelona academy. The average age of the last three was 28; the average age of the first five was 22. It seems fair to say that there was something different driving these two sets of signings.
This was the brief ‘Moneyball’ era of Aston Villa, under CEO Tom Fox*, Head of Recruitment Paddy Riley**, and Henrik Almstadt***. It ended when the team were relegated with just 17 points in 2015/16, an exact half of what the team above them managed.
*arrived in 2014 after five years as chief commercial officer at Arsenal
**arrived in 2014, having previously been an analyst at the club, from a short stint at Liverpool
***arrived in 2015, having previously worked with Fox at Arsenal, and is now working at AC Milan
Were these bad signings? Were Villa unlucky? Is there a chance that a data-driven approach to football could have taken off four years before Liverpool once-and-for-all legitimised it with their Premier League triumph?
A few months after the summer 2015 transfer window closed, Brendan Rodgers was sacked from Liverpool. The Daily Mail’s Neil Ashton wrote a story, which I won’t link to, with the following title: “Liverpool's head of technical performance Michael Edwards is the laptop guru who did a number on Brendan Rodgers”.
The article has not aged well. But this was the atmosphere of the time.
(In a quirk of fate, the two teams were connected in that summer window, as the Merseyside club bought Christian Benteke from Villa for a reported £32.5m; it seems unlikely that Benteke was one of Edwards’ signings)
Villa were clearly embarking on something similar to Liverpool in selling a key asset (in Liverpool’s case, Raheem Sterling) and trying to rebuild with analytics-driven purchases. As a quick overview, here’s where those five buys are now.
- Jordan Ayew - Crystal Palace
- Jordan Amavi - Marseille
- Adama Traoré - Wolves
- Jordan Veretout - Roma (on loan from Fiorentina)
- Idrissa Gueye - Paris Saint-Germain
As individuals, their careers have gone in slightly different directions, but all have taken steps up from Aston Villa. The group as a whole have done well enough that Ayew, whose career has been with meh-to-middling Swansea and Palace sides, looks like the failure of the bunch.
That Tim Sherwood was the manager at the time means that it’s fairly well-established that he was (publicly) not happy with the transfer business in the summer of 2015. As well as the eight players previously mentioned, Villa also brought in José Crespo from Cordoba for a miniscule fee, Tiago Ilori on a loan from Liverpool, and Mark Bunn and Micah Richards on frees.
Of the five that look like fairly clear ‘analytics buys’, three of them were pretty much regular starters. Idrissa Gueye started 35 league games that season, Jordan Ayew 27, and Jordan Amavi started nine of 12 before tearing his ACL in November 2015.
Veretout started 21 games, a hefty share, and it’s only Adama Traoré — still a teenager when he joined the club — who didn’t start at all. There was only one match, a 3-1 defeat at Tottenham Hotspur, where none of the group were in the starting line-up (although Jordan Ayew came on to get Villa’s only goal).
Granted, Sherwood had left just before this, and under him, only Gueye and Amavi were regular starters; Ayew and Veretout were more bit-part. Still, they started four and three times respectively in Sherwood’s ten league games to start that 2015/16 season.
They may not have been exactly the players he wanted, but that hardly seems the sign of a manager throwing his toys out of the pram for being given dud parts.
Perhaps the problem wasn’t the analytics influence, but another, quite fundamental, part of squad-building…
As has been mentioned, Aston Villa brought in a lot of players in that summer of 2015. Quite probably too many.
Nearly 44% of Villa’s minutes in the 2015/16 Premier League season were played by people who hadn’t played at all the previous campaign. This was ahead of the second-highest figure by seven percentage points.
Teams ordered by percentage of minutes played by ‘new’ players; the 17 non-promoted teams only. To save me time going through transfermarkt manually, I’ve used ‘didn’t play in the previous season’ rather than ‘new signings’ per se, so this could include academy graduates
It seems notable that the three teams at the top of this list finished in the bottom four that season. On the other hand, Southampton and West Ham had two incredible years, although neither of those teams had more than a third of their league minutes being played by newbies.
With Sunderland and Newcastle, their high figures (37% and 35% respectively) are probably signs of squads in need of work. Villa will have fit into this category too, in the transition between 2014/15 and 2015/16 seasons, but nearly half of your minutes being played by new players seems like a sign you’ve overdone it.
This is an aspect of Moneyball that is probably overlooked. In a team sport like football, it seems likely that this degree of turnover is counter-productive. It’s often noted how international football seems a lower standard than club football because compatriots may be compatriots, but they don’t play together enough to know each others’ games. Signing as many players as Villa did in summer 2015 probably creates a similar effect.
I’m sure that the minds at present-day Liverpool, Manchester City, Arsenal etc have done work on ‘good’ levels of squad turnover. Managers like Sir Alex Ferguson may have understood it intuitively, or through experience, but by now these clubs (and others) will certainly have some kind of data on it.
I don’t know what the interpretations of this data analysis are though, or particularly how they’d apply to clubs like Villa, Newcastle, or Sunderland in those years. If your current squad is, to put it bluntly, bad, then you might not care about disrupting connections between established players by chucking half of them out. Getting new, better players in might be the most important thing.
Yet on the other hand, chopping and changing the squad doesn’t just affect the players being shifted in and out of the side. The week-in-week-out regulars will have got to know the movements and tendencies of all of their (granted, subpar) teammates. One or two new players might be one thing, but if they’re scattered all over the pitch, that’s a huge amount of the 11x11 connections that are affected.
I was going to leave it there. The theory made sense to me. A lot of turnover equals a self-inflicted shot in the foot that negates some of the benefit of potentially good signings.
But then I did the sensible thing and watched some of the games, and went to check the table for Sherwood and Rémi Garde’s reigns.
When I watched bits from a couple of Sherwood’s games, I saw a team who were a little porous in midfield through a lack of structure, and a team who just struggled to piece anything together in the final third. The midfield could be decent on the ball though, and if they tightened up a bit out of possession then there could be something about them.
The most granular I can go with data for this season (without spending more time than it’s worth on it) is just goal difference on WhoScored. And that tells an interesting story.
In Sherwood’s first ten games, despite only picking up four points, the team’s goal difference (-8) was a similar level to the bunch of bottom five teams that stretched up to Norwich on nine points. I dunno what more advanced stats might suggest, but maybe Sherwood got a little unlucky. Six teams had conceded as many or more than them, although 15 had scored more.
It’s the Garde era where it all went wrong. If we take the 24 matches that encompassed the rest of Villa’s season where there was something to play four (they were relegated mathematically in mid-April), their goal difference (-34) was the worst by far. The next-worst was Norwich, who’d played the same amount of games, on -18.
Did luck put a stop to Villa’s ‘Moneyball’ era? Could Tactics Tim, with a bit more time and a few more lucky bounces, have overseen a melding of old- and new-schools…?
We’ll never know. But it’s a hell of a ‘what if?’ to debate about.
Let’s do quick-fire bumper one after I forgot to do one last week:
- The NWSL Challenge Cup final was on Sunday. I asked before the tournament who to support, Kim McCauley suggested Houston Dash, Houston Dash won. Shout-out Kim, here’s her Patreon; shout-out this thread from The Athletic’s Meg Linehan on the women’s soccer media to be consuming.
- This review/preview of PSG’s season is excellent. It’s in French, but stick it in Google translate and it’s well worth your time.
- The Measurables podcast has put together an incredible list of analytics pros who are making themselves available for virtual office hours for people of color, members of the LGBTQ+ community, women, and others underrepresented in sports analytics.
- A lot of the open football-related coding help is in the coding language R, but PySport are doing neat stuff for Python. See below:
We are very proud to release version 1.0.0 of kloppy!— PySport (@PySportOrg) July 26, 2020
- Metadata models
- Project website with quickstart notebooks
Thanks @brunodagnino @fishnets88 @pratik_thanki @mr_le_fox
Checkout: https://t.co/XUh2cw2ub5#sportsanalytics #socceranalytics #pysport pic.twitter.com/FBoVKdAIXE
A reminder of the Antisemitism Policy Trust’s guides on myths and misconceptions about Jews and antisemitic imagery and cartoons that seem informative. Please, let us all do better.