Questions, answered.
How theories work, how to read a result, what the data covers and how plans and billing work. If your question is not here, we would like to hear it.
What is a theory?
The basic idea in four short answers.
What is a theory in Ballchemist?
A theory is a football belief written as a sentence the data can check. You choose one or more conditions and one outcome and Ballchemist looks at every past match where the conditions were true and counts how often the outcome followed.
Say you want to know how often a team scores when it has won its last two matches. The conditions are that the team won its previous match and also won the one before. The outcome is that the team scores.
What is a condition?
A condition is a rule a match must meet to be counted. Matches that fail it are discarded and the outcome is only measured on the matches that pass. You can use recent form, goals, chance quality, discipline and context such as rest days. Most of them also exist for the opponent.
What is an outcome?
An outcome is the thing the theory predicts. It can be a single event such as a team scoring or a red card or a broader result such as a win, over 2.5 goals or both teams scoring. The full list is on What we can test.
How do theories differ from ordinary predictions?
A prediction is an opinion about one match. A theory is a rule tested against many past matches, so it shows its work: how many matches qualified, how often the outcome happened and how that compares with the baseline. Nothing is hidden behind a score. Match forecasts live on the Predictions page and are not derived from theories.
Reading a result
What the four numbers mean and how far to trust them.
What do hit rate, baseline, sample size and edge mean?
Hit rate is the share of qualifying matches where the outcome happened. Baseline is how often the same outcome happens across the whole data set with no conditions. Sample size is the number of qualifying matches. Edge is the gap between hit rate and baseline in percentage points.
How many matches do I need before I trust a result?
As a rule of thumb aim for 30 or more qualifying matches and prefer 50 or more. A 100% hit rate on five matches says very little. An 87% hit rate on 86 matches says a lot more. The season by season view also helps, because a real pattern tends to show up in most seasons and not just one.
How accurate are theories?
A theory is only as good as its conditions and its sample. Results describe what happened in the past and give probabilities for the future, never certainties. Read the edge and the sample size together and treat one strong result as a lead to check and not as a conclusion.
Where can I read more about the method?
The Methodology page walks through the five steps from a written sentence to a verdict, including the traps to avoid.
Data sets
Choosing what a theory is tested against.
What is a data set?
A data set is the group of matches a theory runs against. It is defined by a league or a club and the seasons you pick. Examples are Premier League 2023/24 or every Galatasaray match from the last ten seasons. The same theory can be run on different data sets to see whether a pattern travels.
What is the difference between a team data set and a league data set?
A team data set looks only at one club, home and away, so conditions describe that club. A league data set looks at every team in a competition, which gives a much bigger sample and a baseline for the whole league.
Can I search and sort my data sets?
Yes. On the Data Sets page you can search by name, club or league, filter by leagues or clubs and sort by newest, most used or name.
What happens when I delete a data set?
Deleting a data set also removes every run made on it. The theories that ran on it stay, and their run counts and best results are recalculated from the runs that remain. Before anything is deleted, a window lists the theories and runs that will go and you confirm in two steps.
Building conditions
Combining rules without losing your sample.
Can I combine multiple conditions?
Yes. Conditions are always joined with AND, so a match only counts when every condition is true. For example, a team won its last two matches and scored at least two goals in each and its opponent lost last time. There is no OR and no nesting.
How many conditions should I use?
Start with two or three. One condition is often too general and five or more usually leaves too few matches to learn anything. The sample size next to your result tells you when you have gone too far.
Can I filter by home or away?
Yes. Venue is a condition like any other, so you can test home form and away form separately or combine venue with something else, such as a team playing at home after a midweek match.
Can I use conditions about the opponent?
Yes. Most conditions have an opponent version, so you can pair an in-form attack with a leaky defence in one theory. That is where many of the interesting questions sit.
Should I use recent form or season-long numbers?
Recent form over the last three to six matches usually says more about the next match than a season average. Combining both can work well, such as a team that won its last two and scores two goals a game this season.
Example theories
Four ideas to try. Run them to see the real numbers.
Hot streak
If a team has won its last three matches then it wins the next one. Use three result conditions for one, two and three matches ago, all set to win and measure a win.
High-scoring home team
If a team plays at home and scored three or more goals in each of its last two home matches then it scores at least twice again. This pairs venue with attacking form.
Struggling away side
If a team plays away and lost its last three away matches then it loses again. Add an opponent condition such as a strong home record to sharpen it.
Defensive strength
If a team kept a clean sheet in its last two matches and concedes under one goal a game over its last five then the match finishes under 1.5 total goals.
Leagues and seasons
What the data covers and how fresh it is.
Which leagues are supported?
The major European leagues, including the Premier League, La Liga, Bundesliga, Serie A, Ligue 1, Eredivisie, Primeira Liga and Süper Lig, along with second tiers such as the Championship and 2. Bundesliga. Coverage differs by league, so the data set picker is the full and current list.
How far back does the data go?
For most leagues it starts with the 2015/16 season. New leagues and seasons are added over time.
How often is match data updated?
Final scores and statistics are normally available within a few minutes of the final whistle. In-play scores refresh every couple of minutes during live matches.
Are international competitions included?
Club competitions are: the matches our clubs play in the Champions League, the Europa League and the Conference League count towards their record. National-team tournaments such as the World Cup are not part of the data.
Support
When something does not look right.
Why is my theory finding no matches?
Your conditions are probably too specific. Remove one at a time to see which one is limiting the sample and check that the league and season you picked actually contain those situations.
How long does a theory take to run?
Most runs finish within seconds. Large data sets, many conditions and a busy queue can make it longer. Testing a single league or using fewer conditions is the quickest way to speed it up.
How accurate is the match data?
It covers official results, goals, cards and match statistics. Occasional differences can appear when a data provider issues a correction. If you spot one, tell us which match it was.
How do I contact support?
Join the Discord server or write to info@ballchemist.com. We reply as soon as we can.
Where do I manage my account?
Open your Profile page to update personal details and interested leagues or the Subscription tab to review your plan.
Plans and billing
Subscriptions in plain terms.
Which plans are available?
Ballchemist offers several plans with different limits on theories, data sets and Elite content. The Subscription tab of your profile compares them side by side.
Can I cancel at any time?
Yes. Change or downgrade your plan from the Subscription tab of your profile, or write to info@ballchemist.com to cancel. You keep paid features until the end of the current billing period and there are no cancellation fees.
Is there a free way to try it?
Yes. You can build and run a theory on the home page without an account and a free account lets you try the Theory Builder on a limited data set before you choose a plan.
What payment methods do you accept?
Major credit and debit cards, processed securely by Stripe. You can update your payment method from your profile at any time.
Nothing matches “”. Try a shorter word or a different spelling. You can also read the Methodology or see What we can test.
Glossary.
The words used across Ballchemist, in plain language. Click a related term to jump to it.
Core concepts
Condition
A rule a match must meet to be counted in a theory. Several conditions are joined with AND.
Data set
The group of matches a theory is tested against, defined by league or club and seasons.
Match filter
A setting that narrows which matches a theory looks at, such as league, season or venue.
Outcome
The result the theory predicts, such as a win, both teams scoring or over 2.5 goals.
Theory
A football belief written as IF conditions THEN outcome and tested against past matches.
Statistics
Average goals per match
Total goals scored or conceded divided by the number of matches. A team averaging 2.5 is more likely to score twice than one averaging 1.0.
Baseline
How often the same outcome happens across the whole data set with no conditions. It is what the league already gives you.
Clean sheet rate
The share of matches in which a team conceded no goal. A steady marker of defensive strength.
Edge
Hit rate minus baseline, in percentage points. A theory only says something when the edge is clear of zero.
Goal difference
Goals scored minus goals conceded, in one match or over a run of matches.
Goals conceded
The number of goals a team has let in. Low numbers point to a strong defence.
Hit rate
The share of qualifying matches where the outcome happened. If 100 matches qualified and 58 produced the outcome, the hit rate is 58%.
Sample size
The number of matches where the conditions were true. More is better and 30 is a sensible minimum.
Statistical significance
Whether a result is likely a real pattern and not luck. It depends on both the sample size and how far the hit rate sits from the baseline.
Analysis concepts
Form
A team's recent results and goals, usually over the last three to six matches.
Head to head
The record between two specific teams. It can show patterns that general form misses.
Home advantage
The edge teams usually have in their own stadium. Across the major leagues home sides win more often than away sides.
Momentum
The effect of recent good results. Teams on a run often keep going, though a streak does not guarantee the next win.
Conditions
Away form
How a team performs on the road. Some sides look very different away from home.
Comparison operator
How a number is compared in a condition: more than, less than, exactly, at least or at most.
Home form
How a team performs in its own home matches, usually over the last few home games.
Streak
An unbroken run of the same result, such as three wins in a row.
Match context
Rest days
The number of days since a team's previous match, a simple way to measure fatigue and fixture congestion.
Time of season
Early, mid or late season. Motivation and congestion change over the year, so some patterns hold only at certain points.
Venue
Where a match is played, at home or away. Venue can be used as a condition.
No terms in this group match your search.
Still stuck?
Join the Discord server or write to info@ballchemist.com. We read every message.