I only started playing Sky fantasy football this season after listening to the excellent “Sky fantasy podcast”.  The attraction, is the unique feature of multiple captain switches and a set amount of transfers.
What I didn’t know was the best formation for the team or how to distribute my £100m budget.  I listened to the podcast and read various posts, and the advice seemed to be, spending big up top with attackers.  So I minimised spend on defence and played a 3-4-3 formation.
Now I seem to be picking up a vibe about spending a bit more on defenders and less elsewhere.  With an International break looming, I thought I’d investigate possible strategies on best formation and spend.
Metrics and initial view
It is quite difficult to know the best metric to use to compare players, but my preference is the number of points per game by value of player.  The points per game, makes adjustments for players that have missed a game through injury or being rested etc.  The chart below comes from the <a href="https://cms.fantasyfootballhub.co.uk/sky-fantasy-football-toolkit/">Sky toolkit</a> that plots midfielders who have played at least 3 games.
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2159" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig1-1.png" alt="" width="659" height="369" srcset="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig1-1.png 659w, https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig1-1-300x168.png 300w" sizes="(max-width: 659px) 100vw, 659px" />
The solid blue line is a linear regression line.  The r2 figure (bottom right hand corner) indicates how much of the variation is explained by the line.  The closer to 1 means a good fit and 0 means there is no relationship.  Here the value is 0.2 which suggests other factors such as position type (CDM, CAM, winger, etc), quality of opposition could explain the variation.  As such we can&#8217;t draw information from this chart.
Analysing top players by position
The next stage was to see if there is a better fit when just looking at the top 20 players by position.  The rationale being our skill as players may be able to identify these in advance.  This produces the following:
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2160" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig2-1.png" alt="" width="647" height="364" srcset="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig2-1.png 647w, https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig2-1-300x169.png 300w" sizes="(max-width: 647px) 100vw, 647px" />
The r2 figure is now 0.57, still far from ideal but acceptable.  There is no visible relationship for goalkeepers, but for outfield players the summary statistics of the regression lines are:
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2163" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig3-1.png" alt="" width="231" height="94" />
How does this help?  Well, we can plot these 3 lines to view more easily:
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2164" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig4-1.png" alt="" width="604" height="347" />
Firstly, defenders are the most valuable commodity, followed by midfielders and finally strikers.  This suggests the best formation to be 5-4-1.
The next part is where to spend the budget and this is determined by the gradient of the lines (eg the co-efficient of the x value).  So invest budget in upgrading strikers, followed by midfielders and finally defenders,
So to select a team, the first stage is to identify the cheapest players that fill the formation.  Assuming you have budget left over, invest as much as possible in the striker.  Once you invested as much as possible here invest the rest in midfield etc.
Is this robust enough to use?
The short answer is no!
This is only based on seven matches. Some teams may have had easier fixtures and distort regression lines.
I found it surprising that the analysis showed 5-4-1 was the best formation.  During the first 7 game weeks there has been an average of 2.47 goals per game.  This compares to 2.80 last season and 2.75 over the last 8 seasons.  This indicates an unusual period of matches with fewer than expected goals giving defenders an inflated worth.
Last season’s data
I have last season’s data (points and matches played) for players with this season’s prices.  However, it maybe worth noting that this would exclude all players from relegated teams and have no data for the newly promoted teams.
Just looking at players who played in at least 10 matches, we have a better fit of the data.  The two graphs below are for defenders, the left with all defenders and the right one taking the top 25 of defenders.
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2165" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig5-1.png" alt="" width="647" height="202" />
The summary statistics of the regression lines are:
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2166" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig6-1.png" alt="" width="244" height="154" />
This shows a very similar story to what we have observed at the beginning of this season.  However, the r2 is higher which gives us greater confidence in the conclusions.  Producing a graph of these lines shows a slight nuance we need to take into account.
<img loading="lazy" decoding="async" class="alignnone size-full wp-image-2167" src="https://cms.fantasyfootballhub.co.uk/wp-content/uploads/2017/10/fig7-1.png" alt="" width="628" height="409" />
This is that top value defenders and midfielders run close together and their regression lines cross at around the £8.5m figure.
The analysis suggest going again with one striker, and as the gradient is the steepest, then invest as much as possible in this player.  It would be reasonable to play a 5-4-1 and investing all remaining money in the midfield.  If you have maxed out investment in midfield and have money to invest in defence which could take a player above £8.5m, then switch to a 4-5-1.
Conclusion
The analysis showed a different picture than I expected, and against the strategy I have been following.  There are a few reasons for this which are:
<ul>
<li>People can identify miss priced players, and adjust appropriately.</li>
<li>Captains receive double points, so having expensive players here can reap rewards.</li>
<li>Higher priced players may have more volatile scoring, eg score a lot more against weaker clubs so have more value through skilled transfer dealings.</li>
</ul>
Last season, Kane returned on average 0.75 pts per game per £m in price, which rapidly tails off thereafter.  If you can identity who will top the charts for this metric then you may gain a slight edge.
However, the scoring system has changed this season for defenders so if they don’t keep a clean sheet, they receive an etc point.  This may also be a cause of the regression lines so far this season and last season.
So far this season, listing players by their points/game/value has Aguero being the best performing striker at 0.7 and in 35th place (Kane is 45th).  There are only 7 matches and so the list will change.  However, it does seem to indicate that blindly going with 3 strikers isn’t an automatic choice and playing 4 or 5 at the back could be feasible.
I will be binning my 3-4-3 formation.  Not sure what formation I’ll be moving too, as I need to consider how to transition with transfers to a new formation.
Let me know what formation you are currently using and if you are planning on changing.

Moving away from 343 – Why you should ditch your strikers in Sky

7 Day Free Membership!

Community & tools

Live rank

Fixture ticker

Vote