soccerpet algorithm is an automated, self-learning system which predicts outcomes of football games with high accuracy. It crunches tons of numbers and applies advanced statistical analysis (Poisson distributions, Dixon and Coles model, Dixon and Robinson model) on a scale, which is not achievable by a human. As a result it generates the most accurate predictions and helps users to significantly improve their ROI.
Initial predictions are based on factors such as:
Statistics - points, goals, attack and defense, goalkeeper, midfield struggles, wing struggles, league standings and many more Team Dynamics - team progression or regression based on current performance Football News - transfer news, injuries, suspensions, etc. Data is processed using methodical models (i.e. modified Dixon & Coles, method of similar opponents). We are also using past accuracy and machine learning to make our predictions better with each game.
Algorithm Recommendation indicates the most probable event for each game picked by our algorithm in second stage of analysis. During this stage we are analysing initial predictions, bookmaker probability and most important factor - past performance of particular recommendations in particular league. After that we use Supervised Machine Learning (two class classification) to assess each of generated recommendations and mark them as “true positive” (1) or “false positive” (0). In this way riskier and potentially unprofitable picks are filtered out.
In the end our PRO users get clear recommendation for each game: Home Win, Home Win or Draw, Away Win, Away Win or Draw, Over 2.5, Under 2.5, BTTS – Yes (both team to score – yes) and BTTS – No (both team to score – no). When game is very tricky to predict, we mark it as No Bet.
To generate predictions for tournaments we use a modified version of standard soccerpet algorithm.
Group Stage Predictions
Our system generates probability for each team to win the group and probability for each team to qualify for the next stage of the competition (playoffs). Analysis consists of two main parts:
Calculation of Team Index - for each team, the system calculates the Index. It is a number which describes the current strength of a team in the context of particular tournament. Index is adjusted according to factors which have additional influence on team’s performance such as: stability (consistency of team’s results), players and squad (assessment of each player individually and team squad as one unit) or coach (experience and results in international tournaments). Simulation of Matches - simulation of all games in each group. Eventually, we have predictions for every game and, therefore, final points for each team. These points will be combined with the Index to get final probabilities to win the group and to qualify to the playoff. Tournament Winner Next step, knowing the results in each group, is the simulation of the playoff stage. In this part algorithm additionally takes into consideration past tournament performance and burnout rate for each team. This allows us to simulate different possible results for each of playoff rounds and sum up to final probabilities.
All probabilities are updated after each round of the tournament.