There is an ongoing argument around the relative merit, or otherwise, of points per game (ppg) or points per 90 minutes (ppg90).
The fundamental point of the game is to score the most points over the season. Ultimately we want to have players in our teams who score high points in every game they play – I hope we can all agree on that much at least?
It seems to me that the rate at which points are scored is a useful metric, but I’m more than willing to acknowledge various caveats. Points alone provide no context. Dimitri Payet gets the same amount of points from dribbling around five defenders and scoring, as does James Milner for banging in a penalty, so too Alexis Sanchez from being touched by the shot from Alex Iwobi. Any goal credited is given the exact same amount of points – a goal is a goal is a goal.
The same for assists; a cross from open play can result in three points, or a deflected shot, or a foul before a penalty, or a simple sideways pass before a long range thunder-smash. So yes, there can be a tremendous amount of luck when looking at points – especially if we’re looking at just a small sample.
So if a player scored a lot of points last week does that mean that they are more or less likely to score points next week?
That is the big question, to which there is no simple answer.
Underlying Stats v Actual Points Scored
Some people will prefer to ignore points scored and instead look at the underlying numbers. The latest theory seems to be that the players who are most likely to score goals have:
- A high number of attempts at goal (Minutes per Chance or MpC)
- A high accuracy when shooting (Shot Accuracy or SA)
- A high number of touches/shots from close range (Shots in Box or SiB and Touches in Box or TiB)
Which makes complete sense, and it’s unlikely that anyone would argue against it. If a player is doing well in those three categories then even if they are not scoring many points in Fantasy Football, you would expect them to start doing better soon. Similarly, if a player is not doing well in those categories, then even if they are currently scoring a lot of points, you would expect them to stop doing so well soon (regressing to the mean).
I don’t believe that football is a game of luck. It is unpredictable, certainly, but over the course of the season – the best team will win. The best players will score the most goals, the best teams will concede the fewest. The ball is round, and bounces in a (theoretically) predictable way – it is just a question of which players are better at controlling it.
There is no ‘random’ element built into the game. This is not to say that fantasy football is as simple as picking the players who take the most shots, or have scored the most points in the last six weeks – but both of these are likely to be things which could be used to predict who will do best next week, and the weeks after that.
A criticism being levelled against pp90 is that it says what “is happening” or “has happened” rather than what “will happen”. Which is undoubtedly true, and there is no value in predicting the past. But an identical criticism can be levelled against any metric – nobody cares if a player has shot 200 times last season, if so far he’s shot just twice. Whether it is the eye test, points per game or MpC we are trying to do the same thing; we are trying to predict future performances based on past performances. The real question is whether the past performances (good or bad) are sustainable?
Capoue v Redmond
For examples like Watford’s Etienne Capoue, who has high points (48), and a high overall total PPG (6.85) despite low numbers of shots (9), there is a counter example like Southampton’s Nathan Redmond who has low points (24), and a low total PPG (3.42) despite a high number of shots (20).
All of Capoue’s stats show that, so far, he has scored at a very high proportional rate and it is unlikely to be sustainable. If you had decided to listen to the stats from the first for games (I know, I know, small sample size) you’d have heard 1.6 shots per game, with 1 SiB per game from just 3 TiB (which itself is a huge spike from a 7 TiB in GW3, compared to 1 in both GW1 and GW2) and said “he can’t sustain that”. And you’d have missed the goal and the assist in the fourth game. And probably the goal in GWK 5 too. Alternatively if you’d noted he was scoring five PPG in the first three games, you might have been more tempted, and you’d have profited from those goals in GW4 and 5. If you were looking to buy him now, you might be tempted by his overall PPG, but a closer look at his recent form shows a PPG of just 2.6 in the last three games which is much less appealing. Though for the price, you might hope there’s still value there.
Let’s say instead of going for Capoue you went for a player like Nathan Redmond, who from the first three games had 2 SiBpG from 5 TiBpG and was having a chance every 31 minutes, rather than Capoue who had to wait 55 minutes per shot. In which case you’d have been frustrated throughout the subsequent four weeks, despite his underlying statistics staying healthy – in GW4-6 he had improved to 2.6 SiBpG albeit from 4.3 TiBpG, and his MpC was down to just 29. Obviously it’s not that simple, but his PPG went from 2.6 in GW1-3 to 0 in GW4-6 (and remains at 0 as he’s done nothing since his goal in GW1 despite most of his statistics getting better as time goes on).
Final Thoughts
Obviously these are two fairly extreme examples, and most players will fall somewhere in the middle. Most players who take relatively high number of shots, with a high proportion from inside the box, and show a general involvement in the penalty box, will score more often than those who don’t. And those players will also show a strong PPG.
It’s very tough to predict who will score well in any single Gameweek, but generally simpler to predict who will score well over the next six or so. There are many players who have stats which flatter to deceive (e.g. Redmond, Andros Townsend and Ross Barkley) but these players don’t fool the PPG/90 metric. As with so many things, the answer is in moderation – there is no single statistic that gives a simple answer. I
don’t think it’s right to say there is no link whatsoever between points scored in the past and points being scored in the future, but it’s daft to suggest that you could predict the future using nothing but PPG (or PP90). I think it useful for showing a trend of things which have occurred, which resulted in Fantasy points.
If, after 20 games, the choice was between…
1. a player with consistently high PPG despite consistently low chances per game, or
2. a player with consistently low PPG despite a high number of chances
…I know which I would pick.
7 years, 6 months ago
Thanks for this. Another excellent stats discussion will ensue hopefully.