Dousing the Fire, Fanning the Flames Gameweek 3

Three weeks into the new season and everything is pretty much back where we expect it. Man City and their rejuvenated cross-town rivals are flying high, Arsenal fans are panicking about their perceived lack of striking options and I've decided to re-vamp the projection model a bit. Chelsea's seven goals conceded and Leicester's flirtation with the big boys, however, show that there`' still room in this league for variance (however fleeting) and it's in that spirit that I've made the latest changes.

I hope the tables are fairly self explanatory, but as a quick background, the focus has been to (a) keep our search for potential flags as wide as possible and then narrow them down with narrative and (b) acknowledge that this entire process is fraught with uncertainty. On the former point, you'll see that I've included four different "projection" metrics to reflect the different identifiers we have for a player's success. For goals, we're now looking at a player's touches in the box, total shots, shots inside the box and shots on target separately to better illustrate why a player may be over or underachieving. For assists, we have the number of chances created by a player and the number of passes they complete in the opponents final third. Those metrics will stabilise at different rates and will need tweaking as more research is done to see if, for example, players from top teams need less shots to score, but for now they give a decent set of warning flags or prompts for potential investment.

On the issue of uncertainty, I was influenced by a recent discussion in baseball where two pundits (presumably new to the stats world) were arguing that a given player was better than the other because he was worth 6.1 fWAR (Wins Above Replacement) versus the other's 5.9. That metric is extremely complex and even its creators would be quick to note the level of uncertainty regarding some of the inputs that go into it, so to argue superiority based on a rounding difference is likely misjudged. I realised I was guilty of that too with my wildly imprecise (and vastly inferior to fWAR) metrics so I've opted for the more visual-based approach to give a more approximate judgement rather than focusing too much on decimals.

Despite being limited to just a cameo role in the opener, Alexis Sanchez has already jumped to the top of the xG table alongside teammate Giroud, with particularly strong indicators with his touches in the penalty box and shots inside the box. The price is obviously high and Sanchez faces competition from the likes of David Silva and Eden Hazard in that elite midfielder category but right now, no-one is doing more for his team and it's just a matter of time before the points start to follow the underlying stats. Troy Deeney is someone I was largely unfamiliar with coming into the season but he's shown plenty of promise in the opening month and deserves attention in that third forward role (though you might want to wait until after this week's trip to the Etihad). Only Giroud has more shots than the Watford man and his eight SiB put him 5th among his peers. The downside is you're not going to get a price rise from Deeney with Premier League debutant Caullum Wilson and old faithful Jermaine Defoe making some waves in the budget market, so while the numbers say Deeney is a nice pickup, at this early stage he might be more obscure than you need to go. It's a similar story with Diafra Sakho, who at 6.5m has performed well but may be too cute of a pick right now.

The next group of names don't need much introduction with Memphis Depay, Aaron Ramsey, Harry Kane and Juan Mata all having plenty of pedigree and promise. All are very ownable, though I wouldn't let this "underachievement" sway you one way or another. If you were high on these players before the year then the data suggests there's no reason to panic but if you preferred the likes of Mesut Ozil or Giroud coming into the year then feel free to ride those players a bit longer.

The last player to highlight is Christian Eriksen, not because he's particularly undervalued but because he illustrates the potential value of the new data table. You can see that his shots on target number is large, suggesting he should have scored more goals than he has, yet the other metrics are not so optimistic. A look at his stat line shows six shots, three in the box and four on target in two games of action. The four shots on target are a nice haul and put him tied for 5th among midfielders but taking a step back we can now see that if he averages three shots a game with just half coming in the box, he's unlikely to keep racking up healthy shots on target totals and thus his expected goals will likely fall. Players do appear to have the ability to hit the target more often than others but a 50% rate for a player who takes plenty of efforts from outside the box is likely unsustainable. I'm hoping that these new tables will help illustrate these nuanced cases a bit better than the old format.

Riyad Mahrez is one of the rare cases where despite heading the "douse the fire" table, he still represents an excellent investment. The only reason for the perceived negativity is that he can't continue to be this good for the remainder of the season, but at his price range he doesn't need to deliver anywhere near his current returns to offer excellent value. Whether or not teams will figure Leicester out is another question for another post but for now, enjoy the ride and the incredible price rises coming his way. The same is true to a lesser degree of Everton's promising duo of Romelu Lukaku and Ross Barkley who have started very nicely, although a look down the road on the fixture list makes this pair more of a "hold" than a "buy" proposition.

The defenders here don't really need comment as they are obviously not going to sustain their goal scoring pace for the year, though it's worth noting that both Vicent Kompany and Russell Martin have posted quite healthy xG rates and can probably be relied upon to deliver some offensive value, making them both potential ownable.

Santi Cazorla is no stranger to underperforming his underlying stats, having notched just a single non-penalty goal last season on an incredible 27 shots on target. This year it's the turn of his assist totals to suffer with nothing to show for his 12 created chances which are second among all midfielders. Despite this solid performance and rejuvenated performance last year, it's hard to see a scenario in which the Spaniard outperforms both Ozil and Ramsey with any consistency, both of whom are available for the same price. An interesting name to keep in the back of your mind is old favourite Gylfi Sigurdsson who has nothing to show for the season to date but has already racked up solid numbers, appearing in both fan the flames tables. Andre Ayew's hot start along with the emergence of Jefferson Montero will likely overshadow the Icelander's return to form but he might be there to capitalise if some of the hot new things start to stutter.

Marc Albrighton has the dubious combination of double digit gameweek and recognisable name on a promoted team that can often lead to widescale overreaction. His numbers are okay, perhaps even encouraging given his 5.1m price tag but he remains a promising bargain bin option rather than a potentially undervalued diamond. Yaya Toure plays at the other end of the spectrum and remains one of the league's better players but continues to confound the model, which generally finds him to be in a constant state of overachievement. This year it's his assist total which looks inflated, although the discount is probably a bit less than suggested as we can safely assume that Man City players (read: Sergio Aguero) convert created chances at a higher clip than the league average. The Ivorian is a tough man to project but with his recent price rise, the numbers would suggest he's tough to get overly excited about.