Venturing into the underlying stats is a dangerous game, and should be approached with the utmost caution. I know of only three other people that have journeyed as far into the caverns of last season’s hidden metrics as I have. Two of them still haven’t returned, and as each day passes fears grow that they may be permanently lost; stuck somewhere in the ether between Milivojevic’s non-penalty xG and Ryan Fraser’s xA.
The third has emerged, but close family and friends report that he is a changed man. The last I heard, he had rented a garage space on the outskirts of Peterborough where he uses his own bodily fluids to document the xG and xA of every player currently active in Serbia’s league system. Needless to say, I’m hoping to get him on the podcast any week now.
But apart from a newly developed phobia of Gylfi Sigurdsson, I’ve come out of my exploration relatively unscathed. The following article is an overview of my findings with one question at its core: how reliable are underlying stats when trying to predict FPL outcomes?
First, we must distinguish midfielders and forwards from defenders. Midfielders and forwards do not profit significantly from clean sheets; they are chosen by FPL managers to score goals and register assists. As such, they need to be treated differently to defenders. We’ll come back to defenders soon, but for now, let’s look at midfielders and forwards.
I started by looking at the top players for combined xG+xA. I’m going to refer to “combined xG+xA” as “expected returns” – mostly for brevity, but also because it sounds fancier.
These are the top players for expected returns last season.
Then I looked at the top scoring FPL points scorers among midfielders and forwards.
These are the top FPL points-scoring attackers in the 18/19 season.
Nice one. So far, so simple.
Next, I wanted to see how closely the two mirrored one another. 15 of the top expected returners also feature in the top 20 for FPL points scored: a crisp 75%. Not bad, but it’s the top 12 that really impressed me. Of the top 12 FPL points scorers, 11 of them feature in the top 12 for expected returns too. The incredibly blonde black sheep is Gylfi Siggurdsson, who resides in 14th for expected returns.
So what? I hear you protest. All the underlying stats do is show us what has already happened, so of course there’s going to be a correlation between them and FPL points.
You raise a fair point – albeit a little rudely. It’s true that underlying stats are a report of what has previously occurred; not a prophecy of what is expected to happen. To get to the bottom of my question, I had to dig deeper.
If a player has impressive expected returns earlier in the season, are they more likely to be among the top FPL points scorers come the end of the season? With the rope tied firmly around my waist, I waded further into the cavity of numbers, decimal points and acronyms.
I began by looking at Gameweek 3’s expected returns. Remember Gameweek 3? ‘Twas a simpler time, where players like Theo Walcott and Alex Iwobi looked like viable options. How naive we were.
This table shows the expected returns after Gameweek 3.
I’ve left the defenders in here as an illustration of how much can change: soon they’ll all have been weeded out.
As we can see, there are some familiar names popping up already. Mohamed Salah has established himself at the top of the pile; a pedestal that he remains upon until the end of the season. Mane, Kane, Wilson, Aguero, Fraser and Pogba also make the cut; a sign of things to come if ever there was one. But alas; we’re not quite there yet. Theo Walcott and Alex Iwobi should be nowhere near a list like this.
By Gameweek 5, the underlying stats look more familiar.
There are no longer any defenders in the top 20 for expected returns (we’ll take a leaf out of Klopp’s book and ignore James Milner), and the likes of Raheem Sterling, Raul Jimenez and Eden Hazard have joined the party.
The players in this list are all pretty good – even the ones that don’t end up in the top 20 FPL points scorers. Arnautovic, Ings and Mitrovic, for example, were decent budget picks for large segments of the season.
By Gameweek 10, there are 14 players in the top 20 for expected returns who will go on to be among the top 20 FPL points scorers come the end of the season. That’s pretty big.
Here’s what the combined xG+xA table looks like after Gameweek 10.
It’s not like all of these were obvious picks by Gameweek 10, either. Sigurdsson didn’t really blip on anyone’s radar until the second half of the season and many took their time to warm to Raul Jimenez. Jamie Vardy was another who was largely ignored until Brendan Rodgers took over.
But 70% of the players listed above would go on to become fantastic FPL assets. Indeed, all of the top 10 scoring midfielders and forwards feature on Gameweek 10’s expected returns table; a figure that must surely be considered when we study the underlying stats of next season.
Can underlying stats help us to pick a good defence?
Using underlying stats to measure the success of defenders in FPL is a trickier business. The key metric is xGA – but xGA is a team metric, not an individual metric.
Man City had the best xGA last season, but only one of their defenders features in the top eight scoring FPL defenders come the end of the season. That’s because there are other factors at play: attacking threat, rotation and bonus points being some of the key ones.
To cut a very long story short, the eight highest scoring defenders are Andy Robertson, Virgil van Dijk, Trent Alexander Arnold, Aymeric Laporte, David Luiz, Marcus Alonso, César Azpilacueta and Lucas Digne. They all share the following in common:
By Gameweek 6, Liverpool, Man City and Chelsea comprised three of the top four sides for xGA – an early sign of the defensive resoluteness they would go on to demonstrate.
If you had looked at the xGA table back then and chosen a backline based on their immunity from rotation and their attacking threat, you would’ve likely had your eye on at least seven of the eight top scoring defenders in the game.
Can we use the underlying stats to predict what’s going to happen in FPL?
So what have we established? Did I risk my sanity for nothing? Was my phobia of Gylfi Sigurdsson acquired in vain? I hope not.
What I’ve established, at least, is that underlying stats are an incredibly reliable way of deducing which players are most involved in front of goal – be it with chance creation or chance conversion.
Though the stats can’t legislate for Mitrovic’s ineptitude in front of goal, they can at least tell me that Mitrovic has been getting those high-quality chances more frequently than other players. Activity in front of goal invariably leads to FPL points.
But there is a big caveat: sample size. Though Gameweek 3’s expected returns table gleaned some truth, it was plagued with plenty of red herrings too. It was only after Gameweek 7 or 8 that the key assets started to establish themselves.
With that in mind, let’s not be too hasty next season. Like a good spag bol, underlying stats need time to simmer on a low heat before we gobble them down. If last season is anything to go by, your patience will be handsomely rewarded.