As I am writing this, Barrie McKay is being sold to Nottingham Forrest and former boss mckayMark Warburton for a rumored £500,000. When I read this, my thoughts immediately went to the video The SPFL Radar made of the compilation of key passes McKay made last season that one would expect would lead to a goal, but did not. Now, I have written countless words discussing the short comings of the strikers Rangers had last season and I will not add to that here. However, the video had me check McKay’s expected assist stats from last season, where I saw he had a 7.84 xA and 0.28 xA per 90. “Pretty good,” was the first thought that came to mind, but then I thought “compared to what?”

By now, most who have been following my journey through stats and Scottish football are probably familiar with terms and concepts like “expected goals” “expected assists“, and “goals per 90 minutes“. If you are not, check out those links above. There seems to be more acceptance of these ideas around Scottish football as time continues to pass. Well, I am about to throw a whole new concept at you. But you have nothing to fear, it builds on what we know rather than something entirely new.

We know expected goals is a better indicator of future success than things like shots or even goals, but it often can be a bit abstract. If I tell you that Adam Rooney had an expected goals of 10.60 in 2016-2017, you can surmise you could expect him to score around 10 or so goals based on his performance. Maybe he will overachieve and scorerooney, maybe he will underachieve and score less, but we could reasonably expect 10 goals. However, does that mean Adam Rooney had a good season? A bad season? A mediocre season? Expected goals by itself does not tell us much when it comes to a player’s performance compared to the league.

While baseball does not have the hold on Europe that it does on North and Central America, one cannot deny that it was an early adaptor of analytics and stats. I have discussed this before when applying the concept of the age curve to the SPFL and I am going to borrow from baseball here again. Baseball has come up with some stats in Runs Created and On-base Plus Slugging that they have realized are better indicators of performance than traditional stats like Batting Average, which certainly sounds like a familiar phenomenon to what football is going through now. They report these stats in traditional percentages, but they also compare a specific player’s stats to the league’s average in that stat with stats like wRC+ and OPS+.

These baseball stats gave me an idea. I would certainly like to know how a striker’s expected goal stats or a midfielder’s xA stats compares to the rest of the league. I decided to apply the same methodology baseball stat nerds apply to wRC+ and OPS+ to football. Now, perhaps proving we surround ourselves in an echo chamber on Twitter, I contacted Rangers Report to see if this idea was crazy or if it made sense. He let me know he was working on something similar with goals and borrowing the ice hockey stat of Goals Above Replacement. He has written about this and you should read it.

To do this, I first needed to determine the average expected value I was going to use. I also decided to use xG per 90 numbers so minutes played would not skew the numbers. For all “attacking” players (as classified by Transfermarkt) that played at least 400 minutes and took one shot last season, the average xG per 90 was 0.18 in the SPFL Premiership (perhaps tellingly that was the Non-Penalty Goals per 90 minutes for that same group as well).

Now that we have an average, we can compare your favorite player’s xG per 90 with the average. We divide a players xG per 90 by the average and then multiply that by 100. Heart of Midlothian v Glasghow Rangers, 1st, February, 2017This gives us what I am, tentatively, terming as a player’s xG per 90+. If you have an xG per 90+ of 100, you are average, at least when it comes to your xG per 90. Every number above 100 is a 1% better than average, anything number below 100 is 1% below average. Let us use Adam Rooney again as an example, last season Rooney averaged a 0.33 xG per 90. We divide 0.33 by 0.18 and multiply by 100 and we get an xG per 90+ of 140, meaning he was 40% better than the average “attacker” in the SPFL last season. Got it? Ok, let’s look at some numbers and see what we can get from them.

xG p 90+_NPG p 90

The above chart compares attackers from the SPFL Premiership xG per 90+ and Non-Penalty Goals per 90 who have at least played 400 minutes and taken a shot last season, and we see many names we would expect towards the top. Dembele, Griffiths, Sinclair, Moult, Boyce, etc. However there were some things I noticed and first was Esmael Gonclaves, aka Isma of Hearts. The Portugese striker was brought to Tynecastle in January and with the controversy that surrounded the Jambos and manager Ian Cathro in that time, Isma put together some impressive stats during a time when not many impressed in maroon. In fact, Isma had a xG+ of 219.7 (meaning he was 119% better than the average striker in the SPFL last season), which was second among SPFL attackers. Combine that with a very impressive Goals per 90 of 0.41 in his time last season, and one could think that given a full season at Hearts, Isma could be among the top scorers next season with that performance. Of course, all this is dependent on Isma not wringing Ian Cathro’s neck on the touchline, which is not a given.

Something else I noticed on this graph was Alex Fisher. The recent Motherwell signing had a xG per 90+ of 145.6 (or he was 45.6% better than the average attacker last season). He also had the highest Goals per 90 of this group at 0.66, but finishing can be fickle and can fluctuate up and down. However, we see Fisher having similar xG per 90+ numbers as Liam Boyce, Louis Moult, and Adam Rooney. Last year, Fisher only played around 800 minutes for Inverness Caley Thistle. We would expect him to get more playing time next year, seemingly as the replacement for Scott McDonald for the Steelmen. While he may see some regression in goal scoring, if he can continue these good xG per 90+ numbers, we can guess he will be able to add more goals for Motherwell.

In addition to using this metric to compare players and their xG, we can also compare players xA as well. We calculate “xA per 90+” the same we do for xG. We take the league average xA per 90 of 0.13 and divide a player’s xA per 90 by that average and then multiply it by 100. In this chart, we see the xA per 90+ and Assists per 90 for every SPFL Premiership player who has played at least 400 minutes and have one assist.

Dashboard 1

Perhaps surprisingly, we see Leigh Griffiths with the highest xA per 90+. I have discussed Griffiths being more than a goal scorer before and we have more proof here that Griffiths is more than a poacher, with Griffiths xA numbers 212% better than average with an xA per 90+ of 312. We also see Ryan Christie second in this metric, with an xA per 90+ of 229. Numerous people have been very impressed with Christie’s time at Aberdeen last year, and the winger will be back at Pittodrie on loan this season. With such impressive numbers, Christie could be  replacement for Johnny Hayes the Dons are looking for this season.

Another surprising name we see right next to Christie is Motherwell’s Elliot Frear. The English midfielder arrived at Fir Park in late January and put in some impressive frear.jpgperformances to help the Steelmen avoid relegation last season, and his xA per 90+ of 220.8, third highest in the SPFL. With a full season with the ‘Well, if Frear can replicate these numbers, he could put together a great season with Motherwell.

Finally to bring everything around full circle, we again look at Barrie McKay. When we see his xA per 90+ numbers, we see he had a 109 xA per 90+. That means McKay was 9% better than the average player in the league when it comes to xA, which was 17th best in the league. With this new metric, I am hopeful that stats like expected goals and expected assists will become more than abstract concepts and a tool we can use to compare players contributions, like McKay’s, the rest of the league. I am also open to suggestions on the name or any other comments you have about it.

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.

2 thoughts on “Identifying Possible Break-Out Players By Making League-Wide Comparisons

  1. If you were analysing a non league football team and only had the data on that team what would you focus on? xG xA and those against your side? Love to hear back!!

    Like

    1. John,
      If you only have data on your team and your opponents each match, your teams Total Shot Ratio (Shots For/Shots For+Shots Against), Total Shots on Target Ratio (Shots on Target For/Shots on Target for+Shots on Target Against), and Danger Zone Shots For and Against for your club. xG might not be effective for non-league sides and without data for the whole league it would be tough to impossible to come up with. I’d stick with those shot metrics for your team. Good luck!

      Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s