Finding the next Jamie Vardy
We all know that 4 years ago Jamie Vardy was playing in the relative wilderness of non-league football and now he is on the verge of lifting the Premier League title with Leicester City. But as a 29 year-old he is the wrong side of the age-curve, and he probably only has another two or three seasons at the top level for which he has just extended his contract with Leicester to cover. So well done to Leicester for taking a chance on Jamie Vardy but how can other clubs learn from this and find their own lower league gems?
I must admit when I started researching for this piece I thought Jamie Vardy must have been a peach in a market of lemons and that I’d find a handful of players that profiled just like Jamie Vardy did. Then I’d be able say here’s ten more potential Jamie Vardys and if there’s any sort of hit rate on them becoming stars then bigger clubs should take risks on these lower league players instead of paying over the top for the safety blanket of ‘he already has Premier League experience’.
The truth is the only forward in recent English history I found that profiled like non-league Jamie Vardy was Luis Suarez in the Premier League. Using goal and assist data from transfermakt.co.uk I looked at 12 years of goal contribution data from the Premier League to League 2 and five years of data from the National League. In the 2012 Vardy dominated the National League, scoring 31 non-penalty goals and providing 17 assists in 3000 minutes of football for Fleetwood Town. That’s a scoring contribution of 1.44 non-penalty goals and assists (NPGA) every 90 minutes.
The only other player to pass this test at the 1% significance level was Luis Suarez in 2014
Now obviously Vardy could have been running hot in 2012 and his expected contribution could have been lower than that so I used the Poisson distribution¹ to test at a 5% significance level that his expected contribution was more than a goal a game. He passed this test and in fact his p-value of 0.0098 would have passed the same test at a 1% significance level. The only other player to pass this test at the 1% significance level was Luis Suarez in 2014 with a p-value of 0.0013, he had contributed to 1.58 NPGA per 90 over 2965 minutes of Premier League football. Widening this test to the 5% significance level adds two more players into the mix: Premiership winning Sergio Aguero also from 2014 and Premiership winning Didier Drogba from 2010. Suarez, Aguero, Drogba. This is rarified company Vardy is keeping but hardly the handful of lower league crushers I was hoping to find.
So what practical lessons can clubs gain from this? Well, Vardy is currently a sample size of one but the next time somebody is absolutely crushing a league, don’t brush it off as ‘only the National League’ or similar but go and consider them seriously. Leicester managed to pick up a striker who’s now competing for the Golden Boot and propelling them to the Premier League title for just £1 million.
One such player recently, as pointed out by Ben Torvaney of Analytics FC, was Marcel Sabitzer who contributed to 1.35 NPGA per 90 for Red Bull Salzburg last year and at age 22 he is definitely on the right side of the age curve to grow. This season he has been back at his parent club Red Bull Leipzig and he has been solid if unspectacular contributing 0.43 NPGA per 90 playing as the second striker rather than the centre forward. This is a similar rate to Vardy’s first season at Leicester and both rates speak of the volatility of goal scoring contribution and the difficulties that can be faced moving to harder leagues. Given Vardy’s progression since his first full season at Leicester, let’s keep an eye on Sabitzer to see if he can kick on too.
Lowering the Bar and Widening the Search
Now given that there isn’t a readily available pool of league crushing forwards ready to be snapped up, it seems smart widen the search pool to forwards who have still had very strong if not stellar seasons. Instead of searching for players with expected goal contributions greater than 1.0 NPGA per 90 at the 5% significance level, I lowered the barrier to players with expected goal contributions greater than 0.6 NPGA per 90 at the 5% significance level in the last 3 seasons of lower league English football. I also filtered them by age to keep only players aged 25 or younger so they still have enough peak seasons ahead to make them worthwhile purchases and finally I filtered players to have at least 60 minutes per appearance to alleviate some of the problems with Transfermarkt’s undercounting of minutes played for substitute appearances.
First up, Andre Gray contributed 42 non-penalty goals or assists in just under 39 sets of 90 minutes whilst playing for Luton in the National League. He was then bought by Brentford for about half a million pounds before they flipped him to Burnley for £9m. He’s now two wins away from being promoted to the Premier League with his current Burnley side.
Both Nouha Dicko and Kwesi Appiah had standout seasons in 2014 followed by solid follow up seasons each playing a division higher, Dicko contributed 0.71 NPGA per 90 in the Championship with Wolves last year and Appiah has contributed 0.61 NPGA per 90 across two loan stints in League 2 and also picked up a goal for on loan for Reading in the Championship. In an unfortunate twist of fate both players have missed the last eight months with ACL injuries but may be two to watch next season if they recover well.
Oliver McBurnie has had very limited minutes in League 2 this year but he contributed three goals and three assists across his 293 minutes of league football. Of course we should be wary with performances over such a small time frame as game states and the quality of the opposition can have an outsized effect on scoring rates. Nevertheless when you couple this with 8 goals and 2 assists in 11 appearances for Swansea’s Under 21s, McBurnie has had a promising start to his career and will be one to watch. As he is already on the books of a Premier League club, he isn’t quite the lower league gem we set out to find and it may be difficult for competing clubs looking for a young striker to sign him.
Finally Marcus Maddison had a lightning fast start to the National League last season picking up 4 goals and 3 assists in 5 starts for Gateshead before Peterborough bought him as the transfer window was closing. He has since had two good seasons in League 1 putting up 0.62 NPGA per 90 in 66 appearances including these goals from just his first two months at the club in 2014.
Okay so these five players aren’t the ‘next Jamie Vardys’ in the strictest sense of the term but they are five good youngish players I’ve found in a couple of hours who I think could all play a part at lower mid-table Premiership or play-off chasing Championship clubs, and bar Andre Gray who’s already had a big money move each of them could represent good value for buying clubs.
What Else Can Scouts at Football League Clubs do?
Firstly, get better data. This was just a top-level scan of goal scoring contribution and as such it is only really meaningful for evaluating forwards. At the very least clubs with resources can get match by match line-ups and create something similar to GoalImpact to start to evaluate players in other positions. Also put in the legwork to get this data for leagues such as the National League North, the National League South, the Evo-Stik leagues and the Ryman League ³, their league websites are crap and club and local news websites are the best public places for this information but Jamie Vardy did spend a year scoring 27 goals for Halifax Town in the English Northern Premier (7th tier football) before moving to Fleetwood in the National League.
Better yet, get event data from a provider like OPTA for as far down the football pyramid as they go and build your metrics for expected goals, expected assists, ball progression, elo ratings for headers or dribbling and tackling. The world’s your oyster when you have the right information available but don’t ever let the data make the decisions for you. Identify the weak links in your own squad and build a shortlist of candidates that look like they could be round pegs for the round holes in your squad. Then watch the players extensively and make judgements with all of the information, visual and analytical, and make the best informed decisions you can.
1: Although it is more realistic to model football events as a set of Bernoulli trials, without access to specific shot data its prudent to use the Poisson distribution as it is the limiting case for a set of a Bernoulli trials and we just have to assume greater variance than there may actually be. This may create false negatives when testing but it avoids the overconfidence that can stem from assumptions given to a set of Bernoulli trials. As you can see below from these 3 models of a 30 goal per season striker, the Poisson distribution has fatter tails than the sets of independent Bernoulli trials. This makes tests using the Poisson distribution harder to pass and less liable to false positives.
2: The data actually flagged up a sixth player, Matty Taylor at Bristol Rovers but it transpired that Transfermarkt have his age wrong and so he should not have passed through the age filter. This is a good warning to check the validity of your data when making decisions professionally but for a couple of hour’s amateur spreadsheeting I can live with this. On this note, Transfermarkt stop counting time past 90 minutes, a trap many of data providers fall into too. But again I’ll let this slide for the purposes of this piece as some data with small errors is better than no data at all.
3: Yes, I spent time trawling the EvoStik and Ryman League websites and although they have goal scoring tables and even separate out non-penalty goals from total goals (This is great and unexpected!!) they don’t supply appearance numbers, minutes, assists or anything else and finding minutes played on SoccerBase or club websites was tricky at best. If you really want a name from these leagues, 25 year old Sam Higgins from East Thurrock United will probably do well for another few years in the Ryman League.