How Goals, Sharp and Brain work
Each algorithm looks at matches from a different angle. Here you can see what each one relies on, where it performs best and what to keep in mind. Numbers come from already verified predictions.
Reliability and odds always go together
A correct prediction is worth as much as the odds it is played at. A method that is right 80% of the time at odds of 1.20 earns less than one that is right 55% of the time at 2.00: 0.80 × 1.20 = 0.96 in the first case, 0.55 × 2.00 = 1.10 in the second. That is why, next to each algorithm's accuracy, you always find the range of its winning odds: compare the two before choosing.
Goals
Highest oddsPure statistics: rates each team's attack and defence from goals scored and conceded.
What it relies on
- Goals scored and conceded by each team in the league
- Attacking and defensive strength, split between home and away
- The league average, so a few matches are not overrated
Strengths
- Simple and transparent: based on results only
- Stable: not swayed by moving odds
- Useful in leagues with many matches played
Keep in mind
- Ignores bookmaker odds and chances created
- Early in the season it has little data and becomes more cautious
Sharp
Most accurateThinks like a pro: combines expected goals (xG) with bookmaker odds.
What it relies on
- Expected goals (xG): the quality of chances created and conceded
- Bookmaker odds, with the margin removed
- More weight to recent matches
Strengths
- Combines what happens on the pitch with the market view
- Predictions usually close to real probabilities
- Reacts quickly to form changes and injuries reflected in the odds
Keep in mind
- Following the market too, it often backs favourites: winning odds are usually lower
- The match needs odds and xG data
Brain
Artificial intelligence: a model that learns from match data.
What it relies on
- An artificial intelligence model from our data partner
- Many variables at once: form, results, statistics and match context
- Updated independently from the other two algorithms
Strengths
- Spots relationships in the data that are hard to see
- An independent view to compare with Goals and Sharp
- Available for many smaller leagues too
Keep in mind
- Like any model it can be confidently wrong: always check the confidence
What confidence means
It is the probability the algorithm gives to the result it predicts. A "1 at 72%" prediction means the algorithm estimates a 72% chance of a home win. BigStake statistics focus on predictions above 70%: less frequent, but more reliable.
Using them together
When two or three algorithms point to the same result, different viewpoints reach the same conclusion. On the matches page we show it with the "agree" badge.
Algorithm combinations for higher odds
We are studying how to combine the predictions of the three algorithms to get good reliability at higher odds too. The method is still being researched: we will release it in the near future, as soon as the results are satisfactory.
Accuracy based on already verified full-time result predictions (1, X, 2), including historical tests. Past results do not guarantee future ones. Predictions do not guarantee results. Play responsibly. 18+