What is The Algorithm?

The Algorithm is a statistical spreadsheet tool developed by our resident statistician Carl Weeks. It uses various spread betting odds and expected bonus points to determine the expected FPL points for each player. It can be applied to help select captains, plan transfers and predicting who will be the highest scoring defender, midfielder or striker over a set amount of weeks (and more!). 

In this new mini-series, Carl will be using the tool to see who it selects to be the best 11 players for Blank Gameweek 31 and Double Gameweek 32 – ideal for those using the Free Hit Chip.

Introduction

With the quarter-final of the FA Cup fast approaching, a crucial part of the season is upon fantasy football managers. There will be blank and double Gameweeks and so the community have been full of discussion about various strategies. The main strategies involve using the free hit chip in either GW 31 (the blank Gameweek) or GW 32 (the double Gameweek).

Following Wills excellent youtube video involving the algorithm for team selections, I will produce two articles on how to use the algorithm to select teams for using the free hit chip in GW31 and GW32.

As the free hit chip is only used for a single week, it would make the most sense to maximise the quality of your starting 11 and minimising the value on the bench. For this article, I assume that you fill your bench with a total value of £16.5m and so can spend £85m (assume have some value in your current squad) on the starting 11.

Setting up the prediction tool

The first stage is to open the prediction tool (

Download link hidden for members only
) and go to the page “User dashboard”. Here ensure cell C2 is set to “FPL” and cells C5:C6 are both set to 32 (the tool is a week ahead of the FPL Gameweek, hence 32 instead of 31).

The players with the highest expected scores for the Gameweek, appear for each position. The second list of players is ordered based on points per £m instead of just points. I suggest opening a blank spreadsheet and copy and paste the value of these tables (followed by copy paste-special formats – or format the cells yourself), you should have something like:

 

Initial team selection

To aid the creation of a team I have built a spreadsheet

Download link hidden, please login or purchase a membership to view.
, where the values above have been copied across. For each player, there is a select option where a zero represents not selected and a one represents being selected. The first step would be to select the expected highest scoring keeper and the ten expected highest scoring outfield players. This gives:

There is a message in the spreadsheet (just below C17) that lists Liverpool as having more than 3 players. There are actually six players selected from Liverpool, and we are currently over budget.

Iteration to meet the player constraint

Our aim is to maximum the expected score of our starting 11. The above achieves this objective but breaks the constraint of only three players from a single team. The next stage is to remove a Liverpool player that minimises the impact of the expected score. For this we look at the lowest scoring Liverpool player for each position and look at the option available:

Lowest scoring defender is VVD at 5.1 pts. He could be replaced by Balbuena or Alonso (but since they are doubts) the next player is Rice with an expected score of 4.4 pts, so this switch is costs 0.7pts.

The lowest scoring midfielder is Mane with 6.1pts, he could be replaced by either Fraser (4.5pts) or Rice (4.4pts). Moving to Fraser would be the best option at a cost of 1.6pts.

The lowest Liverpool striker is Firmino with 5.2 pts. He could be replaced by any position, so either Rice (4.4pts), Fraser (4.5pts) or King (4.5 pts – discounted Arnie as he has been sub recently). Moving to Fraser would be the best option at a cost of 0.7 pts (when looking at more decimal places).

The best option is moving Firmino to Fraser is the best option when looking at more decimal places. Since there are no other Liverpool attacks and we will still need to remove Liverpool players, we can also see that VVD should be replaced by Rice. Giving a team of:

We still have too many Liverpool players, so need to go through the same process. The lowest scoring defender is Roberston (5.4pts) and he can only be replaced by a defender. The best option is Luiz 4.4pts, giving a cost of 1.0 pts.

The lowest scoring midfielder is Mane with 6.1pts, he could be replaced Luiz (4.4pts), Pedro (4.2pts) or King (4.5pts). The best option would be King at a cost of 1.6 pts.

The best option would be to replace Roberston with Luiz.

Final Results – The Algorithm XI

This gives a final Algorithm XI team of:

This team maximises the expected points per player while meeting the constraints of formation, max players from a single team and within budget.

This is an illustrative team, and you may wish to add or subtract players, but hopefully this demonstrates how you can arrive at a feasible team using the outputs of the model.

I will produce another article for the double Gameweek and will be basing the calculations on my own team (eg. My budget). This may be a bit more complex as we’re more likely to bust the budget constraint.

What about Higuain?

People may ask why he’s not in the spreadsheet. Well this is because the default setting is to include players who have played at least 10 matches. This can be changed in the prediction tool in cell C11. If this is lowered to five, then Higuain will appear with an expected score of 4.4 pts (so wouldn’t make the team).

Good luck in the coming Gameweek.

 

The Algorithm

If you’d like to play around with the Carl’s algorithm yourself, please use the link below.

The Algorithm – FPL & Sky Prediction & Planning Toolkit

 

View all our latest fantasy football tips on our Homepage