Our team of Pro Pundits, Hall of Famers and guest contributors offer their Fantasy Premier League (FPL) insight, tips, team reveals and advice throughout the season, with only Premium Members able to access every single article they write.
Here, former FPL champion Simon March highlights the problem with using quick polls to make team decisions.
To prepare for this article I asked a statistician friend of mine “What’s the biggest problem with polls” to which he replied: “An over-reliance on Robert Lewandowski”. I regretted not asking him in writing.
Polls – or surveys which collect and aggregate opinions – have always been a central aspect of the FPL community. Fantasy Football Scout’s ‘Captaincy Poll’, in which readers vote for which player they expect to be the best captain during the upcoming Gameweek, is perhaps the best known among these but they have always existed in many forms, whether it be a straw poll in the comments section or aggregated votes via a Twitter poll.
This is how we tap into the collective wisdom of the Fantasy community and navigate our most pressing FPL-related quandaries. However, polls and their output are often more complex than they might first appear.
Results viewed uncritically or without proper context can lead us astray and, equally, knowing when polling information may be misleading can offer a significant advantage to FPL managers who are trying to break free of the pack. Therefore, this article will focus on how polls can be useful to FPL managers in the right circumstances but also why they should never be blindly trusted.
How Polls Are Useful
As stated, a poll is an aggregator of individual opinions and journalist James Surowiecki – author of ‘The Wisdom of Crowds’ – is one of many to argue that collective opinions tend to be superior in quality to individual ones.
According to Surowiecki, the group tends to be right more often because, when asking multiple people, you can avoid many of the biases that exist in a single person. Over a wide and random group, those biases should even themselves out and present a filtered, objective and unprejudiced answer.
Furthermore, people might reach the same conclusion but for different reasons. Poll participants could each individually know a piece of relevant information which, when aggregated, forms a more exhaustively well-reasoned answer than could be expected of a single person. This effect is what people usually refer to when they talk about ‘collective wisdom’ or a ‘hivemind’.
In an FPL context, the value of this collective wisdom is obvious. Take the aforementioned ‘Captaincy Poll’ for example.
The results bring together the opinions of FPL managers who might have examined underlying statistics, watched matches, studied the opposition, looked at historical data, obtained training videos, followed fitness updates or, indeed, are simply tapping into their own experience and going with a gut feeling after years of playing the game.
Amalgamating these elements into a single source of data is what makes this, and well-conducted polls in general, so potentially powerful.
The Problem with Polls
However, there are caveats to the usefulness of polls, the overarching one being that output is only as good as the input. It’s a problem when the poll doesn’t get enough responses.
Without getting into the weeds of ‘statistical power’ or significance formulations, the general rule is that getting more responses to a poll will make the result more accurately representative of the wider population. They will be less skewed by individual biases.
This probably seems obvious but it’s interesting how often we are satisfied with a small number of responses. FPL forums are full of these kinds of exchanges:
FPL Xylophone: “Player X or Player Y”
FPL Noddy: “Player Y”
FPL Emu: “This ^”
FPL Xylophone: “Cheers”
FPL Oven: “lol”
This type of poll is unscientific and, unless FPL Noddy and FPL Emu happen to be FPL savants, the output is barely better than not asking the question at all. All we are doing is substituting our own biased opinion for somebody else’s biased opinion.
Indeed, we often pitch these questions just to confirm our own opinions and shouldn’t pretend that they are anything more than that.
Selection Bias
But even polls that aggregate a wide number of opinions can be problematic, such as the ‘Captaincy Poll’ due to selection bias.
This is any form of bias that affects the selection of poll participants. So, for example, if you stood outside Camp Nou and asked fans: “Who is better, Lionel Messi or Cristiano Ronaldo?” the response would likely be overwhelmingly Argentine. Ask the same question to a properly randomised collection of football fans, the response would likely be a lot more even.
The problem with even large polls in the FPL community is that we often consume the same information and, thus, are prone to ‘groupthink’ – the tendency of groups to reach a conclusion based on our subconscious desire for conformity.
We’ve seen this countless times before. On Monday, somebody will mention an obscure differential that they are considering and, by Saturday morning, people are triple-captaining that differential. Groupthink can be extremely pervasive and it almost certainly permeates FPL polls.
Another factor is that, while you might ask everyone the same question, not everybody answers the same question. For example, if you run a poll asking who you should choose out of Player X, Y or Z, people will overwhelmingly respond by choosing the one player that they personally own even if, objectively, they might think another will do better.
Instinctively, what we want to happen tends to influence what we believe will happen. Equally, we think more often about the players we own and therefore tend to elevate their potential in our minds.
Recency bias
Finally, if Player Y has just scored a hat trick, people will select him as the best choice for the next Gameweek even if that haul was an aberration based on unsustainable factors and he has a tough opponent up next.
This is partly due to recency bias, where we put greater value on recent events. Consequently, when we respond to polls, especially early on in the week, we tend to be disproportionately influenced by what has just happened and our votes often become skewed because of this.
Conclusion
A well-constructed poll can be an effective way to aggregate the broad knowledge and expertise from within the FPL community and direct it towards making a specific decision. They should occupy a worthy role in a manager’s decision-making repertoire.
However, not all of them are equal and polls that have fewer votes or a skewed selection of non-random participants are subject to selection bias.
Managers should try to observe poll outcomes through a critical lens before allowing them to influence their own decisions. It’s like whenever an economic bubble bursts yet a few contrarian investors still get extremely rich. If you can spot when groupthink has irrationally set in, a huge advantage can be gained by going against the results.
