- Making a Plan for the Upcoming Blank and Double Gameweeks - 22 Dec 2020
- Overcoming the urge to Transfer & Captain “Differentials” - 17 Nov 2020
- Overcoming the Echo Chamber That Is FPL Twitter - 21 Oct 2020
This is Part 3 of the Overcoming FPL series, where we analyze how mental shortcuts and cognitive biases affect our FPL decisions, and what up-and-coming managers can do to change their behavior. While the series is meant to be more approachable for newer managers, we hope that there are some lessons for even the most experienced of managers.
FPL managers frequently use terms like “Differentials” and “Upside Picks” to describe low ownership players that they hope will have big performances in a given gameweek. Picking the right player in the right gameweek can propel a manager above their mini-league and overall ranking rivals.
But is this the correct strategy? Shouldn’t we as managers be choosing the players with the highest Expected Value to either transfer in or to captain, for our teams each gameweek?
Each article in this series details a mental trap that newer, less experienced or sometimes even seasoned managers fall into. This week’s classic mistake is not trusting Expected Value data when making your gameweek captain and transfer decisions, and instead leaning into your own biases in selections.
Expected Value (EV) is defined as the anticipated (or average) value for a future speculation, investment, bet, play, etc. It is calculated by multiplying each of the possible outcomes by the likelihood each outcome will occur and then aggregating all of those values.
Expected Value has direct applications in the worlds of statistics, probability analysis, sports betting, and poker. Additionally, EV calculations can be utilized by FPL managers to determine the best captains and transfers for the upcoming gameweeks. Here is a quick example of how EV works, and a basic application of how it can guide our decision making process.
Each week our very own Jan Sienkiewicz writes the Captain Sensible article over on the FFScout Mothership. In Gameweek 8, given Spurs’ appealing fixture away to West Brom, it was a mostly straight pick between Heung-Min Son and Harry Kane for many managers. In fact, just these two assets combined for over 65% of the votes on the GW8 Captain Poll.
Many of us were faced with either this exact decision of which one to captain, or even the decision of which Spurs player to bring in if you didn’t already own (at least one of) them for some reason. So how does calculating EV help us make this decision?
xG90 = 0.49 // xA90 = 0.28
Scoring a goal 0.49 x 5pts = 2.45
Not scoring a goal 0.51 x 0pts = 0.00
Registering an assist 0.28 x 3pts = 0.84
Not registering an assist 0.72 x 0pts = 0.00
EV = 3.39
xG90 = 0.88 // xA90 = 0.52
Scoring a goal 0.88 x 4pts = 3.52
Not scoring a goal 0.12 x 0pts = 0.00
Registering an assist 0.52 x 3pts = 1.56
Not registering an assist 0.48 x 0pts = 0.00
EV = 5.08
Yes, the calculations above are extremely oversimplified. And yes, they also use previous gameweeks’ data. Many FPL managers are familiar with Expected Goals (xG), Expected Goal Involvement (xGI) or Assists (xA), and even Expected FPL points. Although they all have a word in common, metrics like xG and xGI are retroactive statistics that demonstrate past performance in a match or set of matches, while EV is meant to be a predictive statistic for guiding decisions for future performance.
So, just to be clear, we understand that using xG data for our EV model isn’t perfect because we actually need to use something like Expected Expected Goals (xxG) for any given player. Also there certainly are other factors in play that will change week to week, such as considering small sample sizes (the story of our lives), rotation, and/or who the next opponent is (Son may perform better against high lines, while Kane is more likely to score against a low block).
However, the important takeaway from this data is that based on historical performance, Harry Kane had a higher Expected Value for scoring FPL points in GW8 than Heung-Min Son.Embed from Getty Images
The basic principle behind using EV to inform our decisions is to convince ourselves that if Spurs and West Brom were to play that game 10 times or 100 times, putting the captaincy on Harry Kane every single time would be the correct decision, and the decision that would pay off with the most FPL points in the long run.
It’s fairly easy to look at the math above and agree with that statement, but it is much harder to actually put this into practice. Overcoming the variance inherent in our game that leads to Results-Oriented Thinking is crucial to trusting that decisions based on the players with the best EV will pay off with the most points returns in the long run.
In the scenario above, where we play the Spurs v West Brom game 100 times, if Son were to brace the first game and Kane blank, would it make sense to switch captains for the second game of the set? According to our EV calculations, no. Similarly, if the 1.0% owned Gareth Bale had braced and Kane blanked, would it make sense to captain him for the rest of the games in the set? Definitely not.
This is where differentials come into play. Fantasy Premier League managers love to pick differentials. There are articles, videos, and podcasts dedicated to the pursuit of low-owned players with the potential for high scores each gameweek. There are several reasons:
- When we choose a differential or upside pick and it pays off, we feel like a genius.
- When low-ownership players score big, we rise in the rankings above the “unimaginative” managers that play the “safe picks” each week.
- When we choose a differential or upside pick and it goes horribly wrong, we convince ourselves that it was a punt worth taking, and no harm done.
- Illusory Superiority.
Illusory Superiority is another cognitive bias that causes people to overestimate their own abilities in relation to others. Basically we think we can make better FPL decisions than the rest of the managerial herd. When 50% of the game, and all of FPL Twitter, captains the obvious pick in a home banker, there will be a certain percentage of managers (we know who we are!) that will turn to a differential pick, despite there being very good evidence and data for selecting the consensus choice.
In most cases, the risk of selecting a differential is simply NOT worth the reward. Too many managers seem to conflate differentials with underdogs in sports betting. In gambling, obviously betting on the underdog pays off when you win, because the longer odds make underdog bets worth it. Although unlike sports betting, longer odds do not translate to more FPL points!
To illustrate this point, let’s analyze some EV calculations again. Except now we’ll look at a differential in Gareth Bale vs Kane, with actual gambling odds (anytime goalscorer) and GW9 in mind this time.
xG90 = 0.69 // Anytime Goalscorer = +850
Scoring a goal 0.69 x +$850 = +$586.5
Not scoring a goal 0.31 x -$100 = -$31
EV = +$555.5
xG90 = 0.85 // Anytime Goalscorer = +550
Scoring a goal 0.85 x +$550 = +$467.5
Not scoring a goal 0.15 x -$100 = -$15
EV = +$452.5
So from that EV calculation, with Gareth Bale as a betting underdog, it would actually make more sense to place a bet on him than Harry Kane as an anytime goalscorer against Man City in GW9. Let’s look at this scenario again, but with even odds as it would be in FPL.
xG90 = 0.69 // Anytime Goalscorer = Even
Scoring a goal 0.69 x +$100 = +$69
Not scoring a goal 0.31 x -$100 = -$31
EV = +$38
xG90 = 0.85 // Anytime Goalscorer = Even
Scoring a goal 0.85 x +$100 = +$85
Not scoring a goal 0.15 x -$100 = -$15
EV = +$70
With even odds, Harry Kane’s Expected Value is nearly double that of Bale’s, and between the two, would clearly be the pick for our FPL teams.
A good shortcut is to take a moment and ask yourself if you’d put money on the underdog or differential player to score without the temptation of great odds.
If not, or if you’d put money on another player in your team, then the differential isn’t the pick!
Again, this is an oversimplified example meant to illustrate why trying to be clever in choosing differentials or diverging from the most popular captain picks is not a good strategy. It may work out in a single gameweek. You may spike a low-ownership player’s best performance of the season and shoot up the overall rankings, but how many weeks of taking red arrows while chasing differentials were there in between the spikes?
No matter what mental sport you are taking part in, whether it is chess, poker, or FPL, the best players are the ones who make the best decisions the highest percentage of the time. By using EV to predict outcomes instead of hoping for outcomes, we will all become better FPL managers.
All of this being said, there are two main things to keep in mind:
- The players with the best xG and EV week in and week out are also typically the players with the highest prices in the game. FPL managers literally can’t afford to have the 11 highest EV players at each position. So this logic may be better applied when looking at individual transfers or captain decisions.
- We are not all xG and EV calculating supercomputers (unless you are World Chess Champion, Magnus Carlsen). There’s also not a million dollar prize for winning FPL (SPOILER: we’ll come on to that in a future article). So may we heed the advice of Always Cheating who reminds us to #MFFA or Make Fantasy Fun Again! Pick the players you want. It’s all just a bit of fun anyways!
Every week, FPL Salah does a great job of putting out the anytime scorers odds on Twitter for the FPL community’s benefit. This is such a great resource for managers to use for team and captain selections.