Transfers are difficult. The ‘Tomkins Law’ suggests under half of them ‘succeed’ while, even if you opened up your definition of success, you’re still likely to identify widespread inefficiency. Being even slightly better at them than your opposition has the potential for huge gains in a competition like the Premier League, or any other for that matter.

Data plays a big part in scouting. At an ideal football club, this would be in tandem with traditional scouting, where trained scouts attend matches, watch clips of players and so on to identify strengths, weaknesses and generally profile. Statistical analysis is likely to be used mainly as an initial filter to narrow down a list of potential candidates before further, more ‘traditional’ scouting is done to build a more comprehensive picture.

This may seem fairly simple. For example, Swansea need a new striker, so they filter and then scout, right?

But filter based on what? Goals? Non-Penalty Goals Per 90? Expected Goals per 90? Age?

The filtering process has to be contextualised to maximise the chance that this transfer is a success, and different clubs want different things from their strikers. So we start at the beginning.

Why do Swansea need a new striker?

Note: 2016 data is a couple of weeks out of date, and all data is provided by Opta apart from ‘values’ which are from transfermarkt. 

At least a part of the reason Swansea may want a new striker is to do with goals. Bar a brilliant opening 4 games to the season, Bafetimbi Gomis has hardly been inspiring, though this may be at least in part due to the whole team playing poorly.

However, Gomis’ responsibilities are not limited to goals alone. Swansea are, traditionally, an indirect, possession styled team that require a fairly competent passer and someone who can hold up play.

This is one of the reasons why Wilfried Bony was so liked at the club; despite stereotypical commentator discussion of his size, heading ability and so on, Bony was very suited to slow build up play and the relevant striker duties. For the most part, Gomis has been an able replacement:

Per 90 Stats	Total Passes	Pass Completion	Key Passes	Short Passes (<10m)
Bony 13/14	28.31043	0.7399757	0.825578	3.6807
Gomis 14/15	25.92589	0.7937916	0.689824	4.713798
Gomis 15/16	23.38192	0.7941176	0.916938	3.973397
PL Forward Avg 12-16	28.25	0.7373	0.998	3.0415

But another reason why Swansea may need a new striker is because of Gomis’ incompatibility with Gylfi Sigurdsson, arguably Swansea’s star man after a high value return from Tottenham Hotspurs. Who you play with and your individual output are interconnected, and this is something that, although difficult to quantify, should be taken into account in a holistic approach to recruitment. Since Bony left, Sigurdsson has been notably more muted in the ten role that he was thriving in last season:

And the concentration of his key passes from the archetypal ‘ten’ position has decreased too:

 

At least some of this may be because Gomis passes to that ‘ten’ position (the central area in front of the box) less than the average forward, and Bony:

Per 90 Stats	Passes to ‘ten zone’
Bony 13/14	19.60747
Gomis 14/15	15.63601
Gomis 15/16	13.83048
PL Forward Avg 12-16	15.97

The extent to which Gylfi’s drop in form is systemic, individual or causally tied to Gomis’ introduction is extremely difficult to quantify without insider knowledge of the club – are Swansea playing more down the wing this year specifically to play to Gomis’ strengths, or as an unrelated tactical change? For the sake of this piece (and because it’s my own theory based on watching games), we’ll assume the prior, but ideally coaches could be spoken to and so on.

Eder, a £5m signing as back-up to Gomis, has also been uninspiring when rarely played, and has been in and out of the squad with injury.

So, now that we know why Swansea may want to invest in a new striker (poor goal output, suffering Gylfi, lack of replacements), we can begin filtering.

Who should Swansea sign?

As a first round of minimum criteria, only forwards who have had a season (of > 720 mins) that meets or exceeds the Premier League forward average (since the 12/13 season) in the following will be included:

• Non-penalty goals per 90 (0.3224)
• Expected Goals per 90 (0.33)
• Shots per 90 (2.6725)

These metrics aren’t exempt from systemic biases, nor is it ideal to compare output across leagues, but it is somewhat unavoidable and filtering has to be seen as the best attempt to minimise the chances of failure rather than a completely perfect methodology. The players are grouped by name and season, and can have met the criteria last season or this one in any of Ligue 1, Serie A, La Liga, the Bundesliga, Eredivisie or the Championship.

As you can imagine, a lot of players ‘out of Swansea’s league’ come through the initial filter. Shuffling them out and removing forwards that I know aren’t strikers (e.g. El Ghazi), we’re still left with way too many candidates.

Let’s throw the pensioners out the window and introduce a filter for being under 28 years of age. Also, it’s worth removing players like Odion Ighalo who fulfilled criteria before making the step up to the Premier League, guys carrying long term injuries like Assombalonga and Callum Wilson, and players who have recently transferred like Clinton N’Jie.

At this point, we have just over 40 players left, which is probably just about within the grasps of the scouting system of a comfortably wealthy Premier League club. This shortlist can be passed on to relevant staff, who work on building comprehensive player profiles, filtering based on specific criteria and so on.

Name	Season	League	xG90	Nineties
Jordan Rhodes	2015	Championship	0.4431532	40.930926
Patrick Bamford	2015	Championship	0.3695821	36.207037
Mark Uth	2015	Eredivisie	0.3969865	34.53037
Luuk de Jong	2015	Eredivisie	0.5330976	33.047963
Kike	2015	Championship	0.5600647	28.632407
Max Gradel	2015	Ligue 1	0.5097079	24.293148
Lewis Grabban	2015	Championship	0.4193613	24.207222
Jelle Vossen	2015	Championship	0.4301955	23.739444
Manolo Gabbiadini	2015	Serie A	0.3366086	22.18
Tomas Necid	2015	Eredivisie	0.5091556	20.682407
Brett Pitman	2015	Championship	0.5328166	18.036111
Freddie Sears	2015	Championship	0.4714098	17.693889
Michy Batshuayi	2016	Ligue 1	0.4739613	17.017963
Anthony Modeste	2015	Bundesliga	0.4252083	16.987778
Stef Nijland	2015	Eredivisie	0.3313854	16.533889
Elvis Manu	2015	Eredivisie	0.4355947	16.447037
Andre Gray	2016	Championship	0.360435	15.477778
Sebastien Haller	2015	Eredivisie	0.4931852	14.747037
Abel Hernandez	2016	Championship	0.3892818	14.717037
Lars Veldwijk	2016	Eredivisie	0.4453268	14.529815
Khouma Babacar	2015	Serie A	0.3592972	14.440185
Arkadiusz Milik	2015	Eredivisie	0.3353546	13.681111
Luuk de Jong	2016	Eredivisie	0.5510093	13.259074
Vincent Janssen	2016	Eredivisie	0.5857012	13.092037
Kolbeinn Sigthorsson	2015	Eredivisie	0.4185775	13.063333
M’Baye Niang	2015	Serie A	0.340347	12.965741
Cheick Diabate	2015	Ligue 1	0.4332191	12.96537
Sloan Privat	2015	Ligue 1	0.4705423	12.188889
Robert Muhren	2015	Eredivisie	0.3683008	12.010926
Fernando Forestieri	2015	Championship	0.5094556	11.570185
Marco Djuricin	2016	Championship	0.4282863	11.512593
Arkadiusz Milik	2016	Eredivisie	0.4621074	11.301667
Michiel Kramer	2016	Eredivisie	0.3778392	11.048333
Bas Dost	2016	Bundesliga	0.430471	10.72537
Pierre-Michel Lasogga	2016	Bundesliga	0.3757492	10.617037
Andrea Belotti	2015	Serie A	0.5878321	9.736111
Jurgen Locadia	2016	Eredivisie	0.4565808	9.536296
Borja Baston	2016	La Liga	0.5368348	9.356481
Duje Cop	2015	Serie A	0.3739273	9.342593
Viktor Fischer	2016	Eredivisie	0.4877755	8.768333
Ruud Boymans	2015	Eredivisie	0.5783796	8.690926
Matt Smith	2015	Championship	0.5989744	8.515
Lucas Joao	2016	Championship	0.435138	8.492778

But let’s say that Swansea haven’t scouted any of these players yet, and want to buy someone in January for sure. Remember how they want a player who can link-up with Sigurdsson? Let’s further filter, excluding those who:

• Have a pass completion lower than the average PL striker (0.7373)
• Pass to the ‘ten zone’ per 90 less than the average PL striker (15.97)

Name	Season	League	xG90	Value	Contract Until	90s
Andrea Belotti	2015	Serie A	0.587832	3.5m	2020	9.736111
Fernando Forestieri	2015	Championship	0.509456	2.1m	2019	11.57019
Viktor Fischer	2016	Eredivisie	0.487776	3.85m	2017	8.768333
Jurgen Locadia	2016	Eredivisie	0.456581	2.45m	2019	9.536296
Arkadiusz Milik	2015	Eredivisie	0.335355	4.55m	2019	13.68111
Stef Nijland	2015	Eredivisie	0.331385	700k	2016	16.53389

Four of the six remaining players come from the Eredivisie, which is a league Swansea will be less wary to buy from than some given Bony’s success.

Of particular interest to me is Andrea Belotti. After a 2014/15 season with ~ .59 XG90 and ~ .5 NPG90, the Italian forward moved from Palermo to Torino in the summer for £5.25m. This season, he’s still posting ~ .4 xG90, but scoring only around .13 NPG90. He may potentially be a bargain if his new club see him as a bit of a failure after a poor spell of finishing, and is still only 22. Also:

New favourite player? pic.twitter.com/X836xrJVVd

— Ben (@Torvaney) November 27, 2015

Nijland, too, may be a bargain given the looming end of his contract.

Locadia’s 2014/15 season would have made it through to this stage too had the cutoff been 7 nineties, while he also played more than 2000 minutes with ~ .38 XG90 in 2013/14, and, like Belotti, is still only 22. He’s had a pass completion of over 80% each of the last three seasons, and has played over 30 passes per 90 in the last two. The amount he passes per 90 doubled once Luuk De Jong arrived at PSV, along with a 50% rise in the amount of passes he played from that ‘ten zone’ in the 2014/15 season compared to the one before, which may imply he’s had to drop deeper when playing than before De Jong arrived. Also, three of his five goal this season have been headed, which may mean he could work in both a wingplay oriented Swansea and a slow build up one. These are more initial flags than concrete comments on Locadia’s ability, but they are the sort of insights that can be passed onto scouts, who can then check with their own methods, and used to flesh out a profile.

From the limited amount I’ve seen of these players, I know that there’s a range of striker styles and suitability within the 6 player sample. Fischer, for instance, is probably more of a left winger, while Forestieri may be more of a second striker than someone suited to holding up play. Though there may not be many games before the transfer window opens, scouts and recruitment staff have access to services like Wyscout with which they could plunder through gameclips to assess these things before it does.

Final Comments

Ideally, someone tasked with filtering would like to have loads of scouting time available to delegate and a large shortlist too, to try and alleviate those systemic biases. Say, for example, Michy Batshuayi is explicitly told not to pass into that ‘ten zone’, or Marseille simply don’t use one, this doesn’t necessarily mean he wouldn’t be good at it, yet he doesn’t make it into a final 6 filtered rather crudely because of a hypothesised time constraint.

As far as Swansea go, if they do decide to go for a striker in the January transfer window, I’d be surprised if they hadn’t already started scouting most of the larger shortlist, and almost all of those 6.