Unstructured Vs Structured Data: Why It’s Important to Fantasy Football’s Future
Consider the following hypothetical scenarios:
Player A has scored seven goals in his last nine Premier League games and 23 for the season. His next game is away against Chelsea, who have the league’s best defensive record, at Stamford Bridge. Six of those seven goals have been scored in home matches, and, in fact, Player A has only scored twice away from home all season.
Player B has scored one goal in the last seven games, and he has hit just five goals all season. In a recent interview with the Guardian, his manager has talked about the player’s difficulties since arriving from Boca Juniors in the summer. But his manager has said he has faith in the player and that performances have improved of late, particularly since the player’s family has arrived from Argentina to join him in the UK.
Both of those scenarios might influence us to make a decision with our fantasy football team. Scenario A might make us think about removing the captaincy from that player for a week or two, given his struggles on the road and the fact Chelsea (in this hypothetical situation) have the best defence in the league. This is an example of structured data. It can be compartmentalised, modelled and easily understood, and it’s the backbone of most people’s fantasy football strategy.
As you can see, Scenario B is a bit more complex. This player has been poor all season, but perhaps things in his social life have changed. It takes time to settle in a new country, and maybe we can see some evidence of that settling in. If we take that risk, we might get a bigger reward. This is what is termed as unstructured data. It can’t be compartmentalised – it does not show up in the stats. We might use it for fantasy football strategy, but not always consciously.
Computers getting better at analysing unstructured data
So, why is this important? Well, the use of unstructured data is what sets us apart from stats-based modelling of fantasy sports. Using stats – pure stats – can be beneficial, but it can be aided by our intuition and analysis of unstructured data. This is a rather crass and rudimentary example, but consider choosing between two players with equal stats and performance levels. If you knew that one player was going through a messy divorce and the other had a happy home life – which would you choose?
But here’s the thing: Computers are getting better at analysing unstructured data. IBM has been harnessing the power of its AI machine, Watson, to work with fantasy NFL predictions in the United States. Watson, and other AIs, can scan thousands – perhaps millions – of news articles, sifting through commentary (unstructured data) to make its predictions. It has the capacity to analyse huge amounts of data, and it is getting better at analysing the unstructured elements.
Watson is not perfect – far from it. But you would be foolish to think that they won’t master it one day soon – sooner rather than later. The consequences for fantasy football could be huge, and that doesn’t necessarily mean in a good way. Imagine logging into a computer each Premier League Gameweek, then simply asking an AI to run millions of calculations to determine the optimum fantasy selection. It might help you win, but it takes away much of the romance – and fun – of playing. It might even be considered cheating.
Data analysis will also change sports betting
Obviously, fantasy sports won’t be alone in being impacted by AI in this manner. Sports betting, too, could be transformed. At the very least, it could lessen the advantage of the bookmaker over the punter. However, betting on virtual sports, which is becoming more popular given the better graphics, gameplay and wider betting options of the games, would not be impacted given its software-based predetermined outcomes and use of random number generators (RNGs).
It’s difficult to stress how impactful AI and machine learning will be on fantasy sports and sports betting. The nature of sport obviously means no model can be 100% correct in predicting an outcome. It’s impossible – or almost impossible – to anticipate a player unfairly getting sent off after an incorrect referring/VAR decision.
But it is certainly possible for a computer to analyse millions of pieces of data and determine, for example, that a player’s performance levels always dip on Boxing Day (perhaps because they overindulge at Christmas) or that they tend to score 7% more goals when it is raining.
That’s the bottom line. At the moment, computers are much better at analysing structured data. As things stand – and this is just an opinion – humans are better at using unstructured data. A player could be scoring goals or keeping clean sheets, but you can often see more than what the statistics say. But as computers get better at picking out useful unstructured data, the whole complexion of fantasy sports might change.