DeepNews collects match data for all teams in all leagues and up to 5 years into the past.
Using the sequence of match results, DeepNews calculates, even across leagues (cross-competition), a so-called Elo number for each team (see below) - separated into home and away Elo.
As a guideline, if team A has 100 Elo points more than team B, it will win a match with the probability of 67% to 33%.
DeepNews simulates the progress of a league (DeepNews self-development), i.e. all matches not yet played, based on the fixture list and the Elo numbers of the teams reached so far. Here, a result is determined or weighted 'diced out' for each future game based on the Elo numbers.
However, since 'rolling the dice' - even if it is Elo-weighted - sets a fixed, even sometimes improbable, value for a match, DeepNews simulates the league not 1x, but many thousands of times and averages the results at the end.
This kind of simulation (mathematically: 'Monte Carlo Simulation') gives a very realistic picture of which team will win with which probability, e.g. qualify for the Champions League, get relegated or get into the relegation.
The Elo number was originally developed for rating chess players by Arpad Elo.
It indicates the playing strength of a player or team on a logarithmic scale (comparable to the Richter scale for earthquakes), statistically validated.
In chess, Elo numbers range from 1,500 (beginners) to 2,500 (grandmasters).
In soccer the scale is a bit 'tighter' than in chess, i.e. an absolute top team is around 2,250 Elo points.
First and second Bundesliga are on average about 250 points apart.
The average home advantage is about 40 Elo points.
The Elo numbers for 3 selected teams - as of April 2023: