Goal scoring per 90 mins in the League of Ireland Premier Division

The League of Ireland has just resumed after what one could say is a mid-season break. With the league changing its structure slightly next year, there has been more pressure on teams to avoid relegation. This piece will not be a half-season report or a tactical breakdown but rather an introductory statistical piece of how to properly evaluate goals scored.

Last weekend, I collected some player data for the Premier Division and ran a couple of tests on goal scoring data. We all know that currently Cork City’s Sean Maguire (who’ll be leaving the league at the end of the month for Preston North End) and Bray Wanderers' Gary McCabe are ahead of the rest in the goal scoring charts.

There’s a problem though with taking these numbers at face value. That is because at first glance it can be deceiving. We have no comparison to measure the goals score to. We do not know how many shots or how many minutes each player took and therefore, we cannot exactly be sure that the player we think is good, is actually that good.

A few years ago Benjamin Pugsley from Statsbomb.com (a sports analytical blog) introduced a metric called Per 90 (short for Per 90 minutes). Why would we use Per 90? As Pugsley put it, for this reason: “It gives us context when evaluating players who play wildly different minutes over the course of a season.”

Basically, it’s taking an average contribution of what we can expect from the player for the time they are on the pitch. Essentially what he did was compare two player’s metrics at face value and then over a per 90 minute basis. The results were very different and should open eyes for coaches, scouts, players and fans alike.

This is what I aim to do with the data collected from https://www.transfermarkt.com/. Unfortunately, I only had access to goals scored but the same can be done for assists, passes, touches and so forth. 

When all that data is collected, we can build a profile for each player in the league and compare players to one another and see how they match up statistically. For example, see the radar graph below. The more area covered by the graph – the better the player is. It’s as simple as that.

Let’s call this Player X’s profile.  I know who the player is and what position he plays in. Based on the radar though, we should also be able to identify that. The first developer of these radars for football (as far as I know) is Ted Knutson (@mixedknuts) and the breakdown of each metric can be found here (http://statsbomb.com/2016/04/understand-football-radars-for-mugs-and-muggles/).

Either way, this player is a striker. This can be seen from his high shooting and goal conversion %. He’s also good at dispossessing opponents, creating assists and taking a decent number of shots per game.  Moving on, let’s look the Premier Division and its players.

Premier Division

Let’s take the top 15 players who have played at least 1,000 minutes and compare their goals per 90 minutes played. The reason 1,000 minutes was chosen to install a sense of reliability (1,000 minutes played equals to just over 11 games played).

As the season ends, I would suggest using a higher number of minutes played. If we take players with little minutes played, the chance that luck played can be quite substantial.

Comparing the graph above to the top 5 goal-scorers we can see that David McMillan (Dundalk) and Dinny Corcoran (Bohemians) did not make the cut due to their minutes played being just under the 1,000 mark.

Maguire and McCabe score goals nearly at a rate of one per game on average. After that it drops quite a bit to around a goal every two games for Karl Sheppard (Cork City) and veteran striker Rodrigo Tosi (Limerick). Interestingly, Sheppard is fifth in the goal scoring charts, yet based on the minutes he has played so far this season, he is good enough to sit in third place.

Derry City’s fine performance in front of goal can be attributed to two of their younger midfielders (Barry McNamee & Aaron McEneff) as well as 23-year-old striker Nathan Boyle.

Despite another down season for the Hoops, they boast the joint second most players (3) in my 15-man goal scoring chart. They are Gary Shaw, Brandon Miele & Graham Burke and are all under 25 years old.

What if we were to break the age groups in two categories (under-25) and (25+) and display the top ten players from each age category? Who would show up then?

U25 Year Olds

Maguire again shows up first on the list and is a noteworthy outlier. This is one area where Maguire is certainly good and that could have caught the eye of the recruitment team at Preston North End. Though as he will be leaving soon, who’s there to replace him?

No one really that’s under 25 years old but that’s not necessarily a problem. 26-year-old Sheppard has been scoring goals at a great rate as well, so I presume he’ll be their main man from July 1^st onwards unless manager John Caulfield has spotted his ideal replacement for Maguire.

What about the rest of the chasing pack? Why is the difference so small between the rest or indeed why was Maguire so much better than the rest? It could be due to several factors such as experience, playing style and/or tactics used by their club managers. I would argue that playing time, wouldn’t affect this as I have already accounted for that (>1,000 minutes played).  

Looking at the league table and the players in the charts above, Galway United’s Ronan Murray (http://galwayunitedfc.ie/ronan-murray-signs-galway-united/) stands out for me. At 25 years old, he’s been around the block a bit in the lower English leagues. For a team struggling with relegation, a 0.38 goals per game ratio is not terrible.

As the graph shows, he’s in line with other strikers in the league, yet is being let down by his teammates' contributions. It’s hard to say exactly how, as we are only evaluating the players on one metric. If more were available, we would be able to evaluate players by positions, age profiles and so forth.

26+ Year Olds

I have already talked about the top three players, so let’s focus on the rest or the older age group as a whole.

First things first, the drop off for the older age group is more obvious. This could be due to the age of the players as some a nearing 30, on 30 or even well into their thirties. While it does not explain everything, it matches with previous research into the contributions of older players (http://statsbomb.com/2016/07/player-aging-attacking-players/) in football.

The oldest player of this bunch, Tosi, has done remarkably well for a 34-year-old striker. Interesting will be if he can keep up this contribution. Let’s wait until the end of the season and keep an eye on him.

What about Goalkeepers? Can we use this metric here too? We can indeed.

There are so far only two goalkeepers, Mark McNulty (Cork City) and Peter Cherrie (Bray Wanderers), who have played all the minutes available for their clubs this season.

McNulty has by far been the best shot stopper (using this metric) conceding on average a goal ever two games. Cherrie meanwhile is on the other end of the spectrum conceding just over a goal and a half a game.

The best young goalkeeper right now is Galway’s Conor Winn (25). Even though Galway are having a tough time at the moment, he could be a prospect to watch out for.

Using a metric such as Per 90 can be useful, yet it isn’t everything. This might be able to show us who’s good at scoring/saving goals but not why that might be. In order to examine that, I would need more access to data, specifically shot data such as shot volumes (number of shots players are taking/goalkeepers are facing) and shot maps (below). This allows us as analysts/coaches to assess how the individual players are performing.

To help illustrate this point, I have taken some self-collected data from the German Bundesliga. For example, let’s take Player A and Player B. For the sake of the argument, they are both scoring at a rate of close to a goal per game. The colours match up as follows (Orange = Miss/Saved Shots; Yellow = Goals).

Player A is taking his shots from far out (around a 2% conversion rate) whereas Player B takes his shots from what we as analysts call “good or even great locations”. While I assume it’s obvious/general knowledge, it’s important to emphasise that the closer you are to the goal, the likelihood of scoring is greater. Therefore, we would say that Player B is more likely to continue his goal per game performance than Player A would.

             Player A Shot Map                          Player B Shot Map

Conclusion

To do more accurate analysis of players, we would ideally need a more centralised “hub” of all the match data that is collected. Maybe it already exists and I just haven’t been able to find it.

The best out there at the minute are either Whoscored.com (https://www.whoscored.com/) or Squwaka (http://www.squawka.com/football-stats/english-premier-league-season-2016-2017). Having a resource such as this, would be of real benefit to clubs in the LOI as you can quickly review player numbers and therefore assess performance of players more accurately. 

Until that happens, it is imperative that we use Per 90 as the metric moving forward. This allows clubs to properly evaluate players in several different metrics. A little bit of due diligence before we sign players should not offend anyone. The last thing anyone wants is to waste their club’s precious financial resources.

Lastly, penalties are normally removed from goal scoring records such as this goals per 90 metric. This is due to them skewing sample sizes because of their approx. 75% conversion rate. As penalties were categorised differently under transfermarkt.com, they might influence respective players’ numbers slightly.

Alex Rathke

Performance Analyst

http://alexrathke.net/