Instead of producing a weekly Sky article, I thought I’d produce a series of articles based on how I use the algorithm and analytical tools in making decisions for my fantasy games.
This first article is how I’m selecting my first draft for both FPL and Sky
In my previous FPL article, the main points were to invest heavily in defence, being indifferent to investing in midfield or attack. Though it would be useful to have premium outfield players to ensure captaincy options.
With the algorithm being built, I can start producing my first proper draft. To do this in the forecasting model I go to the “User Dashboard” select game type to “FPL” (cell C2), covering Gameweek 1 to 6 (C5:C6), and must have played at least ten games last season (C11).
I choose Gameweek 1 to 6 as I’m interested more in the short-term fixtures but want it to be a fair amount that the team doesn’t need surgery after the first week.
My first step is to select the 11 highest scoring players by the algorithm which gives (note I’ve ignored Mendy, Lovren and Otamendi as I have concerns about them playing):
Though I think this is quite a good starting 11, unfortunately, it breaks two rules, one budget and secondly maximum of three players from a single club. As I am required to have a bench, my aim is to produce a starting 11 at a cost of £80-£82.5m.
The next step is to copy the summary tables into another spreadsheet and order each position by points/£m as shown **Become a member to view this link**
By looking at the returns of pts/£m, it looks likely that the defensive players will remain resulting in the need to remove the attacking Liverpool and Man City (due to max three players per club rule). As an initial first step I will move Aguero to Kane – this is solely as a captaincy option with him paring with Aubameyang. For each midfield player, I replace with the highest pts/£m player (non-City/Liverpool). First selected is Dendoncker. The next two options are Perez of Leicester and Ings of Southampton, I ignore both of these based on Perez moving club and am unsure of his place and Ings due to injury. The next two choices are Sigurdsson and Gross. This gives the team:
For the bench, I reduced the Upper price on in the forecasting tool (cell C9) to 6/5 and 4.5 to select my bench as Gunn, Diangana, Noble and Ings. These will be monitored more closely in pre-season to ensure they are playing.
As a final step, I run my squad through the Dynamic Programming spreadsheet (link **Become a member to view this link**). Running this through the first six Gameweeks and allowing £0.5m extra per player shows there is no significant difference in transferring in another player. This suggests the squad should be quite stable for the first few weeks. Obviously, things may change with transfers etc, but my current line up for Gameweek 1 is:
To select my Sky team, I go through a similar process, by selecting the Sky game (cell C2) and select the period to be weeks 1 to 4 (since the overhaul is after week four these are the only matches that matter).
First step is to select the best 11 (ignoring players that I have concerns about playing Stones, Mendy, etc). This gives an initial team of:
Though a good-looking team, it is over budget. The next stage is to copy the players from the forecasting model into another spreadsheet and order by pts/£m as shown **Become a member to view this link**
The first set of iterations is to swap the players with the lowest pts/£m with the player with higher ratio (so swap Sigurdsson to Milivojevic, etc). This doesn’t achieve us coming under budget, so we then move players to players with the least worst ratios. This gives a team of:
Before completing the team I need to run this through Dynamic Programming spreadsheet (link **Become a member to view this link**). It is worth noting that the above team doesn’t have a captain Wol v Man U, Villa v Everton and the last matchday of the period. To account for this in the DP model, I give these game days double points (in page P1 K5, L5 and P5 set to 2).
The model indicates that a transfer to Pickford for the final matchday would be worth 11 points (double as captain). Accounting for the captaincy would give Pickford projected points of 19.5 (current) + 5.5 (captaincy boost) = 20.5 giving a pts/£m of 3.6 (0.2 higher than Ederson). In fact, the benefit is greater because Everton have two uncovered captain days so Pickford would have forecast 29.9 points (4.3 pts/£m)
Another possible option suggested is Digne for TAA but this doesn’t improve our team enough to make this a better option.
With bringing in Pickford, there is an extra budget to enable upgrading elsewhere. This is then put through the DP model and repeat until a balance is reached. This gives me a first draft of:
I realise that Rudiger is injured, but will re-visit my team once players are added to the game, and study other possibilities considering captain days.
Hope this article has helped how I use the algorithm to aid creating my team.
Use the Algorithm Yourself
Become a Fantasy Football Hub member for just £1
- OPTA Data (NEW).
- Members’ slack Channel (NEW).
- Carls Sky Prediction Algorithm
- Ian Parrin’s Sky planning sheet.
- How To Win at Sky guides.
- Advanced fixture ticker.
- Minimum 100 extra articles each year.
- A lot more!
Join Our Slack (if you’re not already!)
Follow the Hub
View all our latest fantasy football tips on our Homepage