Are you trying to accurately predict football matches?
I recently wrote the post ‘The Basics Of Creating A Football Prediction Betting Model‘ where I described various data-crunching methods i’d applied to the game, with moderate success. I touched on some of the difficulties one faces when trying to achieve accuracy in their football predictions.
Football is a complex sport to model. It’s difficult to earn from football betting. In this post i explain why that’s the case.
Consider Alternative Factors In Your Football Predictions
By using relatively simple statistics you’ll be able to compile your own probabilities for football matches and [fairly] accurately predict outcomes over the course of (let’s say) a season. However, the snag is that it’s difficult to consistently find value. Most stats-based approaches aren’t sophisticated enough to model individual fixtures with great precision.
In a nutshell, it’s not too difficult to create a system that breaks even over a lot of fixtures. But it is difficult to generate a longterm profit — because most football betting models are completely blind to a whole bunch of ‘alternative’ key factors.
► Ok… What are these ‘alternative’ Key Factors?
Consider the following:
- Squad changes
- First 11 team players
- Managers & Coaches (and their ethos)
- Club buyouts (and their impact)
- Changes to salary structures
- Stadiums (including expansions/rebuilds)
- Injuries & Player suspensions
- Relegations & Promotions
- Transfers (including loan deals)
- Psychological aspects (pressure, scandals, bad press etc.)
- Distance travelled to an away fixture
- Luck (or lack of) in previous fixtures
I believe that all of these factors, and many others, are critical for calculating accurate odds for the top flight leagues. Unfortunately most of them are incredibly hard to quantify. For example, how do we accurately measure the impact of pressure on a football team?…
Even if we attempt to measure the impact of less abstract factors, such as “distance travelled to an away fixture” we still may not learn anything relevant to specific teams right now. After all, delving back too far into a team’s history means entire squads were different, and therefore so was their overall “tolerance” for travelling long distances to games. Hence the difficulty in achieving accuracy.
Most of the football betting systems I’ve developed in the past didn’t include these factors at all. Nor did they have any way of interpreting all of these factors. So my advice is to be sure that you recognise the limitations in any stats-based approaches you attempt yourself.
► The betting exchange is the best marker we have
Collectively the betting markets bring together all of these ‘alternative’ factors, as well as all other variables. One person may overcompensate on one of the above factors by strongly believing that it will impact a match (e.g. the introduction of a new manager). Someone else may hugely under-compensate by believing it has little-to-no baring on the match. But with enough participants, and enough diversity, the exchange pools together everyone’s ideas and meets in the middle. What this achieves is an ‘average’ price — and this is the most accurate estimate that we have on offer.
The public is collectively wise, and as individuals we struggle to outsmart that. Therefore, we need to look for something influential that the public has neglected.
Some of my ‘alternative’ suggestions are worth keeping in mind. You might also find my Top Tips for Premier League Football Betting post helpful.
► Remember that every fixture is unique
In the past I noticed that my betting models could identify value in specific cases. This underlines the importance of being “selective” with your bets. If you’ve found an influence on a game (e.g. a strong attack vs a weak defence), then you should capitalise on it. But if this doesn’t apply to another game then don’t simply bet on the fixture for the sake of it.
Each game is influenced by different factors. One strategy alone isn’t all encompassing. So ensure that you make every effort to assess fixtures as standalone events. Remember that not one single football match is the same as another.
► Old-fashioned and new-age methods combined
New-age advanced/detailed data analysis techniques are likely to improve accuracy and reduce bias in making football predictions. However, recently i’ve started to think there’s potential merits to developing strategies incorporating human intuition…
I spoke to someone a few years back who worked for Arsenal and then Spurs (ironic I know), as a match ‘analyser’. As i understood it, his job was to closely watch the games and to offer a breakdown of what he felt worked well and what didn’t. Sounds like an amazing job. His opinions were then combined with statistical reports in order to provide a well-rounded analysis of the match, showing what was observed by an expert vs what was determined mathematically. So even top flight football clubs aren’t relying entirely on advanced data analysis.
The logic behind this approach is that data collection doesn’t account everything that a human does. For example, when we watch a football game we’ll often attribute success to good luck — like an unjust decision for a penalty kick, a red card etc. In many cases we’re better equipped to say “that was a jammy result” than a computer is.
Undoubtedly some people are naturally more suited, or otherwise more qualified, to offer their opinion than others. The question is, if a level-headed, unbiased expert is capable of scrutinising professional football teams to offer a valuable opinion, then wouldn’t those skills be transferable to value betting?
I think they are. Unfortunately it’s impossible to back-test.
► Testing my own intuition
The idea of human intuition interests me, and lead onto my Mug Betting Experiment. I’ve made absolutely no mathematical analysis in this. I simply try to compile all of the thoughts rolling around in the back of my mind into rational and unbiased selections on every Premier League fixture. I just wanted to test myself. After 90 games I have achieved over 25.98% ROI — which is most likely complete luck!
…But it’s not bad going either. Perhaps it serves as a reminder for us to keep an open mind and to avoid becoming too fixated on the historical results — you might miss a trick.