The Reverend Project: THE PREFACE

A few years back, I fell in love with machine learning. The idea of outsourcing the analysis of a game you want to place a bet on to the machine was appealing to me. Still is, to be honest.

My journey into the world of maths, stats, and the machine learning algorithms has began with big hopes, but the reality decided to show its ugly face and deflate my dreams of creating a piece of software that will make me rich while I eat, sleep, and fool around doing nothing but cashing out the checks… it nearly broke my heart.

For years now, I’ve been applying machine learning to the problem of the betting profitably. So far, nothing revolutionary came out of my efforts, but there are some useful bits and pieces. Today, it is a collection of scripts for scraping, transforming, analyzing data, and calculating the probabilities of the outcomes. In a word: a mess!

In the same fashion, the programming languages I’m using to keep this monstrosity alive are: R, Python, Go, my favorite Ruby, and one language I refuse to name for the reasons too personal to tell you in the public forum.

So, why I didn’t succeed in creating an automated system for profitable betting?

There are some practical consideration I failed to account for at the beginning of my journey:

1. The lack of smarts and effort – the effort part I’ll admit to be guilty of, but the smarts… I’ve read the most of the published papers on betting by the guys sporting PhD behind their names, I didn’t find myself out of depth there. So far, the scientific consensus is you can’t bet profitably on football.

2. Getting data is hard. Really hard. - You want as much data to train your models on in order for it to produce the satisfactory predictions. If I could, I’d collect data on the eating habits of a player, whether or not the referee had sex the last night, vocal prowesses of the club’s fans… unfortunately (or not) I can’t get my hands on such info.

3. 99% is still not 100% - or the ball is round and bounces in every direction. You have a bunch of guys on the pitch chasing a round object with purpose of putting it behind a goalkeeper, there are people cheering them on, people making annoying sounds, gaffers and referees,… it’s a complex dance and even a club superior in every aspect of the game, performing on a high level, can (and will) lose the game. No sure bet exists (outside of the arbitrage opportunities).

4. Bookies don’t like me – for one reason or the other, I’ve kept getting my accounts terminated with no clear explanation of what I did wrong. That wouldn’t be the problem, but for the fact that I wanted to build a fully automated system that would collect the games of the day, calculate the outcomes, and place the bets while I watch porn or something.

With the history and the reasons for the failure of my original dream out of the way, let me tell you a bit about my plans for the future of this project.

I’m no longer chasing the dream of automated riches, instead I’m pulling out the useful bits of this project and assembling them into a website that will be launched in the early August this year, in time for the new seasons of the English football leagues.