We all have that one guy in our social media feed who claims himself to be a football ‘guru’ and funnily enough his predictions do come true at times. On the other hand, some people make predictions based on various factors like analyzing data on players, teams, form, and history and still could not be 100% correct on their football betting predictions.
Determining the accuracy of football predictions
To determine the accuracy of any football prediction, you have to measure the ‘distance’ between the probability of the outcome and the outcome itself. There is a simple formula for measuring the distance.
Distance = 1 – The probability of the outcome
The shorter the distance the higher the accuracy. A score of 0 indicates 100% certainty of the outcome and the only way you can get a zero is if you are God. If your prediction goes completely wrong, you get a maximum score of 1. Let’s try to understand this with the help of some examples.
Example 1: Brighton plays Arsenal and says the probability of them winning is around 30%. If this outcome occurs then the distance is 1-0.3 = 0.7
Example 2: Liverpool plays Everton at home and says the probability of them either winning or getting a draw is around 80%. If this outcome occurs, the distance is 1-0.8 =0.2. As I mentioned earlier, the shorter the distance the better the accuracy.
How do predict football outcomes?
- Previous records
Probably the easiest way to forecast football outcomes. This method uses time-series data to focus on the previous head-to-head encounter between two football teams. The probability of the same result happening again is taken as 1. Certain teams have strong records against certain oppositions and on certain grounds.
For example, Everton last won a game against Liverpool at Anfield in 1999. The record was broken recently when the Toffees won 2-0 against Liverpool on February 2021.
- Home-Away form & Draws
This method of prediction involves the use of cross-sectional data. If you are a football fan you’ll know the home-away advantage can make a big difference to results. The home fans can create an intimidating atmosphere for the away team at times which plays a psychological role in the players’ minds. A great example is Liverpool. The Reds went over three and a half years without losing a single Premier League game at Anfield until January 2021 when they lost to Burnley. It had to be Sean Dyche, isn’t it?
If we look at the Premier League itself, in the 2021/22 season, there were 163 home wins, 88 draws and 129 away wins, further proving the home advantage factor.
Some Advanced methods
• Dixon and Coles
Up until the 1990s, there was a strong belief among people that predicting the outcomes of football matches was pointless because it was down to chance. It was then Mark Dixon and Stuart Coles dropped a masterclass. The duo used a ‘Poisson Process’ named after physicist Simeon Poisson.
Initially, their model worked on the assumption that goals were scored at a consistent rate during the match. It was also assumed that the goal total varied depending on the teams too.
To calculate the expected goals for each side to score they divided things into two factors namely the attack and defence.
Expected goals scored by the home side were determined by:
Attacking ability (Home) * Defensive weakness (Away)
Expected goals by the away side:
Attacking ability (Away) * Defensive weakness (Home)
Dixon and Coles collected data from the English football league’s four divisions i.e 92 clubs across several years and it equated to 185 factors to analyze. Certain rules like promotion and relegation made things complicated. Therefore, they decided to use more computational methods.
As you might have guessed already, such a complicated method doesn’t come with perfection. Although what Dixon and Coles did was brilliant, the Poisson Process wasn’t a perfect fit for football.
The major flaws include underestimation of draws. There are more draws than it predicts in reality. Research done on Bundesliga games for around four decades revealed teams taking fewer risks at 0-0 in the final 10 minutes, on average.
Another flaw worth noting is that goals are not scored at a fixed rate. There are more goals scored in the final 15 minutes of each half than in other periods. Dixon and Coles also didn’t consider the fact that players get exhausted during games and hence the fixed attacking and defensive ratings were not correct towards the end of matches.
• Bayesian Model
Talk about making the best football prediction, the Bayesian model is based on making the best use of the information we know so that we can be wrong less often. The Bayesian model uses relevant updated data to change its calculations. It was named after Thomas Bayes, an English minister who lived in the 18th century. Although it was Bayes’ idea, the formula was introduced by a French mathematician named Pierre-Simon Laplace which we use today. The formula is an algebraic equation with four variables.
P(AB) = P(A) * P(BA)/P(B)
The probability of event A given B is equal to the probability of event A multiplied by the probability of event B given A divided by the probability of event B.
Let’s try to understand the Bayes model with the help of an example.
Say, Chelsea is playing at home against West Ham United on a Tuesday night. You know the following stats:
• Chelsea have won 6 out of the last 10 at Stamford Bridge
• West Ham have won 2 of the last 10 games at Stamford Bridge with 2 draws
You can say Chelsea has a 60% chance of winning, a 20% chance for West Ham victory and a 20% chance of a draw. If the odds are greater than 1.67, you’d quite comfortably back a Chelsea win.
But, what if you find out they have played four times on a Tuesday night and West Ham has won 2 games with 1 draw and 1 Chelsea win? Here’s where the Bayes model comes into play, taking into account the new conditions that could change the odds of the outcomes.
So, West Ham has 2 wins at Stamford Bridge on a Tuesday night and not any other day of the week. There have been a total of 4 Tuesday night games. Applying this information to the formula:
P(A) (West Ham winning Tuesday night games) = 2/4 = 0.5
P(BA) (West Ham winning in all meetings against Chelsea) = 2/10 = 0.2
P(B) (Tuesday night games) = 4/10 = 0.4
Therefore, P(AB) = P(A) * P(BA)/P(B)
P(AB) = 0.5 * 0.2 / 0.4
0.1/0.4 = 0.25 = 25%
So, the game is on a Tuesday night increasing the chances of a West Ham win from 20% to 25%. Although this may not be a significant increase now you get the deal. Would you use the Bayes model now to make football predictions for tonight or football predictions weekend?
However, the best free football prediction site remains Betclever. Don’t forget to check us out for the best football prediction for the weekend.
Predicting football results is a skill and to master this skill you have to let go of your beliefs. You cannot assign a 100% probability to one situation and 0% to the other. The best example is the 2015/16 league winners Leicester City. There were at odds of 5000/1 and guess what happened? Similarly, once you learn to master this skill you will be able to make more accurate predictions in today’s football.