Selling a good player is one of the biggest risks you can take with your fantasy football team. Lots of people sold Sigurdsson last week, then he scored. You can see why people did it. Whenever there’s a highly owned player like Sigurdsson in the FPL game people will be on the lookout for an alternative, someone to give them an edge over rivals. There always seems to be an alternative, too. With so many players playing in the Premier League each week there will almost always be somebody who scored more points last week and have a better fixture for the week ahead than the player you own.
We’re certainly not knocking anyone who swapped out Sigurdsson last week. There were good reasons to do it and we were contemplating ourselves whether a revitalised Stewart Downing had a genuine case for long-term inclusion over Sigurdsson this season. Indeed, this dilemma was what drove us to have a deeper look at both players and resulted in this post. Downing has had significantly more shots than Sigurdsson this season. In fact, a lot more; 29 to 13. As you know, at insideFPL we also use 'Expected Goal' data to measure the likelihood of a goal from each shot and Downing is ahead in this regard too, although not quite so much; Downing’s expected goals (xG) is 1.77 to Sigurdsson’s 1.33.
What is the case for Sigurdsson then? Perhaps it’s one for quality over quantity. Even with ‘expected goal’ data we can only measure so much. We can’t measure how much defensive pressure was on the shot, where the goalkeeper was, or exactly how easy the shot was. We also can never truly determine how good the player is. A basic assumption of most shot models is that all players are equal, but this is not the case.
Judging a player’s goalscoring ability is notoriously difficult and typically requires a year or more of data and 75-100 shots before a statistically valid conclusion can even start to be drawn up. Looking at the pair’s shooting history however is revealing. Over the last three seasons (and prior to this season) Downing and Sigurdsson have taken a remarkably similar number of shots; 164 for Downing, 160 for Sigurdsson. From these Downing has scored 4 goals to Sigurdsson’s 15. Taking the quality of these chances into account via xG and shot location, shot type, pass type, etc. we can see Downing has an expected goals of 10.3 compared to Sigurdsson’s 12.1, meaning Downing has converted his chances at a torrid 39% rate whilst Sigurdsson is at a very decent 124%. Historically it’s clear Sigurdsson has been a far better goalscorer.
What may explains this further is the typical quality of their shooting opportunities. An average shot from Stewart Downing over the last 3 years has had a goal expectancy of 0.06. Sigurdsson’s over the same time is 0.08, about 20% better. This is important when you think about the difference in football between half-chances and ‘big chances’. A team/player can have a ruck of half-chances and never look like scoring, whilst a player or team who get just a single big chance can easily put that away to win the game. The ‘quality gap’ this season between Downing and Sigurdsson is more pronounced. Despite having many more shots, the average quality (goal expectancy) of Downing’s chances is only 0.06 compared to Sigurdsson’s 0.10.
A theoretical explanation for importance of 'shot quality' in football is described here by Mark Taylor. Mark simulates a match between two teams: Team A create 2 very high probability chances in a game with a goal expectancy of 60% each, versus Team B that take many more shots (12) but at a lower goal expectancy (10%). Both sides have the same overall goal expectancy of 1.2 (2*0.60 = 12*0.10) but in the simulation Team A wins the game more often. They’ll never score more than 2 goals but will rarely blank either. Team B fail to score a lot more.
Shot and key passes numbers are very widely used across fantasy football these days. We didn't intend for this article to be confirmatory evaluation of either player, more to point out that simple shot volumes might not tell the whole picture and perhaps we’re finding a new way of looking at the same numbers. Accounting for the quality of shots using expected goals and then mixing in some basic probability theory in the same way Mark did we can boil down both player’s shooting numbers. The probability of Downing scoring a goal in any one game is 15% whilst for Sigurdsson it's12%, much closer than their pure shot totals would suggest. Factor in a little of their historical shot conversion rate and there’s really only one winner.