4SCORE is a unique tool. Forecasts are based on machine learning algorithms. In this article, we'll explain how to access them, and you can read about the calculation method itself in our guide, as we use recognized algorithms and aren't just some "black box" solution.
This tool is available as part of the MAXIMUS subscription, learn more about paid subscriptions
What are AI forecasts
These are forecasts made by machine-learning algorithms (artificial intelligence, or simply AI). We don't source them from anywhere—all algorithms are customized by our specialists and run on 4SCORE servers.
!
Interesting
As with many 4SCORE tools, we try to calculate the actual probability of an outcome, compare it to the bookmaker's odds, and, if the calculated probability is lower than the bookmaker's odds, tell you about it (since in this case the bookmaker's odds are higher than the calculated odds, which we consider to be value).
!
The essence of the tool
Every day, our AI machines analyze the day's matches and bookmakers' odds. We look for matches where the machine considers the probability of a particular outcome to be higher than the bookmaker's.
Machines analyze all matches of the day and calculate the probabilities of the outcomes. Once everything is calculated, the machines translate the probabilities into odds
The obtained odds are compared with the bookmaker's. This allows you to find matches that are worth considering, or at least paying attention to, since the bookmaker's odds can be value bets
Let's look at the example above (synthetic data). In it, we found a forecast “Total corners under 10.5”, The probability of which, according to the AI machine, is higher than the bookmaker's. Here's what you can deduce from this screenshot:
A forecast has been found for the Freiburg - Bologna match. Yes, it's pretty obvious, but you have to admit, what's a prediction without a match? In any case, we'll show it clearly
Model confidence (С) - 80%. Note the "C" (Confidence) parameter in the upper right corner, which is equal to 80%. This means that the model used to obtain the prediction is 80% confident that it has correctly predicted the probability. How the model "understands" whether it is confident or not can be found in our article
Probability by model (P) - 52%. That is, the model believes that there is a 52% probability that the total number of goals in the match will be less than 2 (note that we found exactly this outcome above, but it is not in the screenshot)
Coefficient by model - 1.89. Мы рассчитали его из вероятности, которую дала модель по простой формуле k = 1/p, где p - как раз вероятность 52%
Bookmaker odds - 2.74. We take it from the bookmaker's line, so it's not calculated as is. We try to update the odds every 10 minutes and every minute in LIVE.
ROI is positive, 68%. That is, the bookmaker's odds are 68% higher than the calculated ones, the bet can be very valuable if the ML machine is not mistaken (and it thinks it is 80% sure)
Let's sum it up. Below the forecast, we briefly describe why we consider it worthy of your attention, as well as where the parameters were derived. Finally, we write whether we consider the forecast to be value-oriented
To summarize, the essence of the tool is to use machine learning to calculate the value coefficient and find one in the line that is higher than it, thereby “highlighting” the forecast for you to consider
How to access ML forecasts
On the main page of the site, in the navigation bar, go to the “AI forecasts” section.
1
Specify filter values
Filters help you customize your report to find the matches you need. First, select the market and bookmaker where you want to search for statistically relevant value odds, then use the filters to refine your criteria:
Leagues. You can select a list of leagues, the system will include predictions for matches from which in the report, so as not to be distracted by championships and matches that are of no interest to you
Date. Choose which day to generate a report for. Reports are available for two days in advance. The filter includes the seemingly useless "yesterday" option—this is required to retrospectively assess the completeness of the data without purchasing a subscription
Confidence. Forecasts with either high confidence (80 or higher) or within the specified range will be selected. The more confident the model is in the accuracy of its prediction, the better
Probability. Forecasts will be selected where the model's probability lies within the specified range. Probability is the model's estimate of the likelihood of the predicted outcome occurring in the match (if 100, the model believes it will definitely happen)
Coefficient. Filter by bookmaker odds. This is useful, as placing multiple predictions at odds of 1.1 is clearly not the best strategy
Bookmakers. Which bookmakers' lines should be compared with the probabilities obtained from ML models? That is, the report will only present predictions for the selected bookmakers
ROI. The percentage by which the ML model's odds are higher than the bookmaker's. The greater the difference, the more valuable the forecast. Convenient if you're looking for the juiciest ones.
2
Get the report
The system will analyze the entire game day and generate a report. It's worth noting that our ML machines calculate match probabilities every day at 8 a.m. Moscow time.