It was the summer of fall of 2008 when I played PES 2009 for the first time. Now the actual game but the demo version which was included in the Digit magazine. There were just 6 playable teams about whom I knew nothing. Two years before that I had watched Zidane headbutt Materazzi in the FIFA 2006 final and I knew I liked the game of football. But one disadvantage of pre-internet Cricket crazy nation was that you wouldn’t know what’s happening in the world of sports. PES 2009 changed that for me. Since then I became a fan of Manchester United, because I played best with them among the 6 teams, and started following this lovely enthralling game which would be over in just 90 minutes. Twelve years, FIFA 11-20, countless elations, heartbreaks and nail-biting moments later, football remains one of my constants. And now I am merging my other two constants, analytics and marketing, to create a blog about the Top 5 Football Leagues Predictions.
Note: This is not a python tutorial for simulating football fixtures. The code (with proper comments including sources of data) will be made available only on request.
What does Football League Predictor do?
The simulator acts as a predictive engine to forecast home team win percentage, draw percentage and away team win percentage of a match up between two football sides ready to face-off in the next 7 days (the number of days can be altered as per requirements)
In the simplest of terms, as on 15 December 2020 (IST) the model predicted that Barcelona has only a 22% chance of winning the game at Camp Nou against Real Sociedad. This doesn’t bring confidence to the predictive model being used. Barcelona average player salary is 7.675 time higher than Real Sociedad’s. How can we expect a team like Barcelona to perform worse than Real Sociedad? Especially when on 17th December 2020 (IST), Barcelona defeated Real Sociedad by 2 goals to 1. The prediction is totally off the base here. Or is it?
If we look at the data of the matchup between Barcelona and Real Sociedad, both teams were evenly matched in terms of possession, shots, corners and passing accuracy. Moreover Real Sociedad was better in terms of shots accuracy and tackle accuracy. On paper Real Sociedad would have won. And this brings confidence to the predictive model. Also the fact that Real Sociedad had lost only 1 of their 13 games while Barcelona had lost 4 out of their 11 games till 15th December 2020 gave the result in favour of the former. Also since the predictive engine is trained only on the league matches of the current season, the result do not take former glory into account.
Elements of Football League Match Prediction Model
But these predicted results shows the same team twice. If the whole predictive model for football league fixtures is based on the past performance of teams in the league decided by the following factors:
- Actual Goals Scored (G)
- Actual Goals Conceded (GA)
- Expected Goals Scored (XG)
- Expected Goals Conceded (XGA)
- Home Advantage (HA) : HA for each individual team was estimated as the percentage of goals scored in home matches by that team. For example, if a team scored 50 goals in their home matches and conceded 30, then their unadjusted HA would be 50/(50 + 30) × 100% = 62.5%.
- Away Disadvantage (AD) : Correspondingly, AD for each team was estimated as the percentage of goals conceded in away matches. HA greater than 50% represents superior performance in home matches, whereas AD greater than 50% represents inferior performance in away matches.
- Poisson Distribution: Simulate home and away goals by drawing from a random poisson distribution ‘10,000’ times. This is because goals in football follow poisson distribution
then doesn’t it mean that the prediction of next to next game as shown in the League Predictor is not valid?
To answer that question, yes! you are totally correct. This week of 15th December 2020 to 21st December is a cramped week for league games in absence of national duty and cup fixtures.
That is why I have created a Football Top 5 Leagues Match Prediction page where I will be updating the league-wise weekend prediction on a weekly or bi-weekly basis with a later update comparing with the actual results. Hop on to Football Top 5 Leagues Weekend Prediction to see the latest winning predictions of your favourite team for this weekend.
I wanted to share my insights on how I have made a 18.53% profit till date by gaming on football using the Top 5 Football Leagues Match Predictions. This is notable since only 11% of gamblers actually end up with a profit and even the best of sports gamers is right only 55% of the times.
Since gamiing is a whole new topic with minute imperative details, I am allocating a second post as continuation of this topic. Click here to read Part 2/2. In the next blog I will be discussing how to make a profit in football gaming via Top 5 Football Leagues Match Predictions.