I'm always looking for something new to plant my teeth into. This morning I discovered my latest pet project; real money Fantasy Football (Soccer) Betting. In particular, I'm looking to explore whether I can use data and statistics to develop an algorithm that will help successfully predict the outcome of football matches & player performance.
This summer FanDuel launched an English Premier League (ELP) version of its platform in the UK. FanDuel describes itself as a "one-day fantasy football" platform. The site allows users to "go head-to-head with other Premier League fans for real money."
I signed up today and took a look around the site. There were only fifteen or so "contests" set up for users to enter. It seems like early days for the platform which will probably make me an early adopter on the UK version.
The contest that has caught my eye is the £7500 Fan Favourite. There are just under 1500 entries, with Prizes for the top 400. Entry is £5. It looks like each user is allowed up to 50 entries, so I imagine I'm up against far less than 1500 individuals. I have no idea how savvy the fellow contestants are, and I can't see any details about fellow competitors. I haven't deposited any money yet and I cannot work out how (or whether) I can set up my own contest.
I might have limited access due to me not depositing anything. Or it might be due to the newness of the site. I'm taking a punt with this platform, however I understand FanDuel has a solid reputation across the pond, so I think it's a safe enough bet. I'll do some more investigating before depositing any cash.
Football Prediction Algorithm
The real money wager aspect, is only a small part of why I'm signing up to FanDuel. My main reason is to explore whether I can use data to help predict football results and player performance.
The key word is help.
I realise there are so many variables and moving parts that it is virtual impossible to accurately predict the result of a football match. However I am convinced that the use of data analysis can help improve the quality of my predictions.
Fantasy Football games (like FanDuel) focus primarily on predicting player performance. However my starting hypothesis is that team stats (the players team and their opponents) are far more informative on how a player will perform than just looking at players stats alone.
My second hypothesis is that teams defensive and attacking qualities are largely mutually exclusive, particular in modern day football and in the EPL. A combination of how much a team defends coupled with how well will help determine good defensive players to choose. A similar logic can be applied for attacking players. Thus it makes sense to split the analysis between attacking data and defensive data.
My third hypothesis is that whether a team is playing home or away is relevant to the team's (thus the players) performance.
My fourth hypothesis is that the team who the player is playing against (and has played against) is important. Sounds obvious but is worth stating. Analysing who the opponent is can be simple enough. However when you start trying to analyse who the team has played against, it can get a lot more complex. For example, I may easily conclude that Zlatan Ibrahimovic is a great pick upfront. However it gets more complex when I factor in whether their games were against teams as miserly as Chelsea, or as leaky (defensively) as Arsenal.
Nanzo's Indexes
Each week I will be publishing EPL Indexes. These will comprise of:
- Home Team Attacking Index (HTA points) - team rating for attacking play at home. Rating will help inform of the best teams from which to choose forwards and attacking midfielders, when playing at home.
- Home Team Defensive Index (HTD points) -- team rating for defensive play at home. Rating will help inform of the best teams from which to choose forwards and defensive midfielders, when playing at home.
- Away Team Attacking Index (ATA points) - as per HTA points only for away attacking teams.
- Away Team Defensive Index (ATD points) - as per HTD points only for away defending teams.
These Indexes will help shape my match predictions. My match predictions will give me pointers as to the best players to consider for my fantasy football team. I will be publishing my prediction (and preview) of each match during the week. And I will publish my fantasy football team once the next round of matches are underway.
Along with submitting fantasy teams on Fan Duel, I will be placing modest bets (on match outcomes) on Betfair. I will also be entering "correct score predictions" on Sky's Super Six. There is nothing that focuses the mind, like staking some money on the outcome! It will also be fun to see how I do.
Just The Beginning...
I'm sure there are lots of approaches and football algorithms out there. I do intend the explore these at some point. However I like to start off with a blank sheet and crystallise my own thoughts before exploring the ideas of anyone else. I wouldn't classify myself as either a football or a data science expert. However I have an armchair interest in the former and a healthy curiosity about the latter.
If you're interested in Football, Prediction algorithms and/ or big data, I hope you follow me down this rabbit hole...