Goalkeeper choice: what the stats tell you

The question of how much we spend on our goalkeeper duo at the start of the season rears its head each campaign.

This time round, there appear to be a few combinations that are in the mix, but the debate seems to boil down to if it is worth paying for a premium keeper, or is a 4.5 goalkeeper enough? There are many permutations beyond this (e.g. rotation v the “king and heir” such as 4.5/4.0), but that’s the central question.

Displaying the current multitude of approaches out there, the results of a straw poll I ran on Twitter strikingly demonstrates that FPL managers are utterly divided on this point.

For premium keeper selectors, the key driving factor behind ownership of a few of these high-priced options – in particular, the premium keepers in David De Gea (DDG) and Ederson – is the assumption that they will ‘cover’ a defence through offering security of starts better than owning a premium defender equivalent. For example, I have assumed from the outset that owning DDG instead of Chris Smalling (6.0) is the optimal route in to the Manchester United backline.

For those with rotation and “king and heir” strategies, I think the inherent value for money in a 4.5 such as Lukas Fabianski (fuelled by saves) is the dominant factor in that choice. Nick certainly belongs to this camp. This tends to mean going with a premium defender instead of the expensive guy between the sticks.

There are, of course, some other sub-sets in the community – for example those going 5.0/4.5 with a rotation, or having a 4.5 behind DDG.

So, with all this in mind, the question we should answer becomes clear: which is the better set-up according to last year’s data?

To see if we could answer this question, we turned to our statistical genius Ed (Prokoptas), who created an ingenious way to mine the data and tell us what the optimal combination was in 2017/18.

Method

The first thing to say is that this is very difficult to do manually. It’s incredibly time consuming to interrogate the data by adding up DDG + different defenders’ scoring through 38 gameweeks, then adding up all the other keepers in the same vein.

We’re happy for others to have a go at this.

Instead, an interesting way we thought we could test with was through using a randomisation model, where the ten most defensively consistent sets of GKs and DEFs are cross-matched.

In a modelled simulation, each goalkeeper was combined with the highest-scoring defender from each other team (excluding their own side) – which meant we ran this simulation ninety times over.

The average outcome records a statistical indication of what you might have scored via a season of points.

We then exported the mean result of each set using GK as the constant, and PPV as the variable. We used non-minutes adjusted data for this analysis to avoid the appearance points skew (as explained by Nick in our “points per cost”piece).

This outcome was then ranked for each keeper to assess the relative values of a cheap keeper and premium defender approach against that of a premium keeper and cheaper defender.

This should help us understand premium GK/cheap DEF vs the opposite, and then highlight potential outliers.

Outcome

The average yield metric shows the average points per million a keeper would have offered you, after being crossmatched once with every defender. For example, some simulations Fabianski may have been 30, other times 27, and so on – the combined average is the yield. The pool is the goalkeeper + best defenders, excepting the representative from their own team.

The four cheapest options finish 1st, 2nd, 3rd and 6th. In contrast, the premiums finish 4th, 7th, 8th and 10th, with the mid-price options finishing 5th and 9th.

It seems that the cheaper Goalkeepers produced a higher average points yield paired with a counterpart premium defender.

Additionally, DDG outperformed his price tag. He provided almost as much value as a budget GK (and indeed outperformed Lossl) despite his 5.5m cost. It’s worth noting, too, that all premiums that were not DDG proved poor value.

We also reran this simulation with DDG at his current price, but with the same performances for him and everyone else: interestingly, his value dropped down to 6th (below Pickford and Lossl). He would therefore have been the best value premium ‘keeper last year at 6.0 – but still outperformed by all 4.5 goalkeepers.

Analysis – DDG

Davesaves you value? (Photo: The Independent)

Let’s look at the Golden Glove winner himself in more detail first.

Earning this season’s uptick, DDG provided strong value for money last season. Doing a quick comparison between last season and his three season average underlines what a good (and slightly anomalous) season DDG had. Saves-wise, between 2014/15 and 2016/17 he averaged 83 saves per season, whereas for 2017/18 he managed 115: an increase of 39%(!). Similarly, DDG also kept notably more clean sheets (18) than the average over the preceding seasons (average 13). This translated into a notably higher number of points last campaign than previous seasons, with his 172 points dwarfing the 139 average points he’d scored over the last three seasons: a 25% increase.  

This may indicate that, with the 0.5 price shunt, DDG may struggle to provide value commensurate to his price this year. Nonetheless, there are historical examples of keepers in top teams sustaining decent form, so the performance compared to value dilution might not be something we should be too worried about. One example is Joe Hart, whose 2010/11 season was eerily similar to DDG’s 2017/18 season: 18 clean sheets, 115 saves, 175 points. The next season, he recorded 17 CS, 99 saves, and 166 points. Another is Petr Cech, whom had a 159 point season in 15/16 (101 saves, 16 CS) that was pretty similar, which dropped down to 134 PTS next year. Granted, there’s a slightly reduced number of points in the follow-up season for both, but the expected regression could be limited.

The same is true of other premium options, who were comprehensively outstripped by their cut priced counterparts in our model.

Analysis – cheaper keepers

Rebutting the assumption that the high probability of starts for premium ‘keepers leads to effective points generation, the data set we used shows that the better combination is a 4.5 keeper and a premium defender. In effect, it was better to own Fabianski and Valencia rather than DDG and Mawson.

This is because, despite the season DDG and others had, the cheap keepers still provided the most value relative to their pricing. This extrapolation feels intuitive – budget GKs still rule the roost in this data, as well as in other metrics we’ve showcased (e.g. Talisman Theory).

Picking the right 4.5 can be the keystone to defensive points; getting a Pope or Heaton in your side early and hitting 9-10 clean sheets, plus saves, is the recipe for FPL keeper success. 

By extension, this underlines how premium defenders just do a better job of returning value, too, which might also be in favour of going for a cheaper keeper (as long as you have the right one). Again, this synergises with Nick’s points per cost work, showing how defenders are the true outfield value kings of FPL currently.

There could be some reservations about the security of starts for premium defenders, but the highest scoring defender on teams is usually pretty assured of appearing. You just have to be a bit more active in monitoring the trend with these players rather than with a premium keeper. Our model also accounted for this through using non-minutes adjusted data.

Whichever way we spun it, the premium defender and budget goalkeeper approach still consistently came out tops.

Fun fact: if you used per-start numbers the gulf would have been even larger in favour of the cheaper keepers.

Just a note, too that the time on field for the guys mentioned above shows just how unlikely it is that you’ll need your backup keeper. Last season, Courtois missed 3 games, Ederson 2.5 (injury), Pope 2.5, Loris 2, DDG just the one and Fabianski, Ryan, Pickford & Lossl all played every match; collectively, they played 97% of the time. Perhaps, if you’re going with any strategy that has one of your keepers as a “backup”, it might be worth investing the 0.5 elsewhere and going with a 4.0 fodder pick. This may mean you’ll need an FT to fix an issue if it crops up – but you’d do that anyway if anyone else got injured, too.

Conclusion

Fab Forever? (image: Lukasz Fabianski Official Twitter Page)

The numbers seem to come out in favour of a cheaper keeper being the superior choice. This is because cheap goalkeepers paired with premium defenders are favoured statistically.

Nick tends to sway towards owning cheaper goalkeepers, and would point to his “player value for money” work for why, in a stats-driven sense, you’d be mad to not take advantage of what the 4.5 options offer. He had a successful season with the likes of Phil Jones covering DDG, as well as investing early in premium defenders such as Ben Davies. For him, an option such as Fabianski or Wolves’ Rui Patricio might be the port of call in anticipation of clean sheets plus saves generating ample points to attain small gains per match.

With this all said and done, my view is that the stats only (as ever) provide one part of the picture. A significant factor in owning a goalkeeper from a premium side is week in-week out security with regards to defensive starts: if your 6.0 Smalling is suddenly dropped, you might wonder why you didn’t invest in the sure thing that is DDG. If defensive starters are a lottery in premium sides, then the keeper is the winning ticket. To me, whether we acknowledge it or not, psychological factors play an influential role in decision-making with regard to paying for a premium goalie. These are normally qualitative factors that are difficult to measure quantitatively – such as peace of mind that you’ll have a secure starter 99% of the time.

However, this isn’t to say my approach completely ignores the stats: I’m applying what they say to my own strategy. In this case, my strategy is to start with the trusted option of DDG, with a view to swapping to a cheaper option and reinvesting outfield at first wildcard. By that point, this year’s Pope/Heaton plus the premium defenders likeliest to appear regularly should be identifiable. Equally, I might be able to save 0.5 by swapping to Ederson, say, if City are having a solid season defensively.

In the case of a 4.5 keeper v a premium one, then, it seems that the stats indicate that cheapies provide us better value for money, meaning we should be investing in the defence rather than between the sticks. Men like Fabianski, who has come in the top three for saves for the last three seasons, provide a fantastic amount of points per value. There are caveats beyond the stats, though, meaning it’s for individual managers to decide which option best suits their strategy when picking their FPL ‘keeper.