Fantasy Prep 2

Conventional fantasy football wisdom is that goalies and defenders are pretty useless as far as points-production is concerned. Attention (and money) is typically more focussed on the “stars” in midfield and forward positions. To assess the evidence and potential pitfalls we can start with data from the last Premiership season (I am not clever enough to make meaningful judgements about players from other leagues who will be participating in the 2013-14 season).

Using whole season data and taking a pool of players who generated at least 1 point over that 38-week period from August 2012 to May 2013, I found that three-quarters of the 381 players who met this most basic of selection criteria ended up with less than 100 points apiece for the season. These “Also Rans” produced on average only 9 season points per £mn (using current, pre 2013-14 season prices). There were only 5 stars with 200+ season points (20pts per £mn), 15 excellents with 150-199 points (21pts per£mn) and 79 good players with 100-149 points (19pts per £mn).

Of course, given the strongly-binding £100mn budget, it does not take an integer programming genius to work out that splashing out on a few stars will quickly force you into the dark pool of sub-£5mn players. You might well be lucky and choose an “Also Ran” in a rare moment when he actually does well – that’s part of the judgement call that is Fantasy Football. But what experience and data show us is that you will (probably) get it wrong more times than right. Also the average price for a representative player is just that – an average. There are some players for less than £7mn that have produced more than 150 points last season. And there is a goalie and a couple of defenders who make the “Excellent” category (Hart, Baines and Evra).

Averages over entire 2012-13 season Number of players achieving at least 1 point
Price* (£mn) Total points Weekly points Pts/£mn GK DEF MID FWD Numbers
200+ pts STAR £11.5 228 6.0 20 0 0 3 2 5
150-199 pts EXCELLENT £8.1 170 4.5 21 1 2 7 5 15
100-149 pts GOOD £6.3 122 3.2 19 11 26 26 16 79
<100pts ALSO RANS £5.1 45 1.2 9 20 97 127 38 282
381
*as at August 2013

 

When picking a team for the new season I prefer to focus on the last 20 weeks of the 2012-13 season. Before that is too “historic”, in my view. Moreover, at least for the first few weeks of the new season, I only look at “meaningful” players – that is, players who achieved at least 60 points in that 20-week window from the previous Premiership season.

Given those filters, we have a pool of 83 players of which 12 are goalkeepers (GK), 21 are defenders (DEF), 31 are midfielders (MID) and 20 are forwards (FWD). The table below summarises the relevant information.

Weekly Fantasy Points Profile £mn/Mean
High Low Mean Std Dev
GK 5.6 1.6 3.6 1.0 £1.5
DEF 6.2 1.1 3.6 1.3 £1.6
MID 5.9 2.1 4.0 1.0 £1.9
FWD 6.0 2.3 4.2 0.9 £1.9
TEAM 5.4 2.4 3.9 0.8 £1.8
Points data based on the final 20 weeks of the 2012-13 Premiership season
Player costs are based on current prices at the start of the 2013-14 season

 

As far as “representative” players are concerned (so GK is the average of all qualifying goalkeepers, etc) it is true that, on average, midfielders and forwards are more productive as regards the expected number of weekly points. However, differences are not statistically significant (also see the chart below). Moreover, when it comes to value (the final right hand side column) searching out the quality goalies and defenders is certainly worth paying careful attention to.

The “Team” is suitably weighted to represent formation constraints and assumes the mean and standard deviation of the “average” formation (of the 8 that are feasible in Fantasy Football). For what it’s worth the 3-4-3 formation provided the highest points per-week average (3.91) with the lowest average coming from a 5-4-1 lineup (3.82). Needless to say, these differences are not statistically significant.

Formation Chart 1

In the chart above, markers indicate averages (expected values) with the vertical bars representing 95% confidence intervals. Data calculations are based on the last 20 weeks of the 2012-13 season.
 

Note that if you are successful in matching the “representative” team comprising “meaningful” players you would have an expected points total of just short of 50 points per week. That is not quite enough to get into the hallowed 2000+ zone for the season as a whole. Nonetheless, it provides a pretty solid foundation for outperformance.