Both Teams to Score: How Tactiq Estimates BTTS

By Tactiq AI · 2026-05-05 · 6 min read · Methodology

Both Teams to Score (BTTS) is one of the most-asked secondary questions about a football fixture. After "who wins?" comes "will both score?". The two questions get answered by different machinery in Tactiq's analysis pipeline, and reading the BTTS output together with the over/under and win probabilities gives a clearer picture than any one of them alone.

This article walks through how Tactiq computes BTTS probability, where it adds value, and how it interacts with the other outputs.

The two questions BTTS answers

BTTS yes is the probability that the home side scores at least one goal AND the away side scores at least one goal. BTTS no is the probability that at least one of the two sides finishes scoreless.

This is a different question from total goals. Consider three fixture scorelines:

2-2. Total goals: 4 (over 2.5). BTTS: yes.
3-0. Total goals: 3 (over 2.5). BTTS: no.
0-0. Total goals: 0 (under 2.5). BTTS: no.

The first two have the same over/under outcome but opposite BTTS outcomes. The first and third have opposite over/under outcomes but same BTTS outcome (no, because at least one side blanked).

The BTTS output captures a question about who scored, not just how many goals were scored.

The math

Tactiq's BTTS probability calculation rests on each side's expected goals scored figure for the fixture. The home side has an expected goals scored against this specific opposition (call it lambda_home). The away side similarly has lambda_away.

If we treat each side's goal output as drawing from a Poisson distribution with the corresponding lambda, the probability that the home side blanks is e^(-lambda_home). The probability that the away side blanks is e^(-lambda_away). The probability that AT LEAST ONE blanks is the sum of these two minus the probability that both blank.

BTTS no probability is then 1 - the probability that both score, or equivalently:

P(BTTS no) = e^(-lambda_home) + e^(-lambda_away) - e^(-lambda_home) * e^(-lambda_away)

P(BTTS yes) = 1 - P(BTTS no)

The Poisson assumption is approximate. Real football scoring distributions are slightly over-dispersed, meaning the variance is slightly higher than the mean. Tactiq applies a small correction factor calibrated per-league to handle this. The headline math is the formula above; the production output incorporates the correction.

When BTTS probability moves the read

The BTTS output is most informative when it diverges from the over/under output. Three examples:

High over probability + low BTTS probability. A fixture where one side is heavily favoured and expected to score multiple goals while the other is expected to be shut out. A typical 3-0 or 4-0 favourite scenario. The fixture will likely have lots of goals, but most will come from one side. Useful for fans who want to read "this is a probable rout" rather than "this is a probable open game".

High over probability + high BTTS probability. A fixture between two attacking sides with leaky defenses. The expected output is open and end-to-end. Both sides expected to score, multiple times. The classic chaotic top-six clash or relegation six-pointer between two flammable sides.

Low over probability + high BTTS probability. A fixture expected to be tight and low-scoring but where neither side is expected to be completely shut out. A 1-1 or 2-1 result is the modal outcome. Common in defensive fixtures with two sides who do create some chances but rarely many.

Low over probability + low BTTS probability. A fixture expected to be tight and where one side is likely to be shut out. The classic 0-0 or 1-0 grind. Often top-of-the-table sides versus relegation strugglers, where the favourite wins by suppressing the opposition entirely.

Reading the two outputs together gives texture that reading either alone misses.

Where BTTS calibrates well

Tactiq's BTTS calibration is strongest in the leagues with the deepest historical data: Premier League, Bundesliga, La Liga, Serie A, Eredivisie. Brier scores in these leagues sit comfortably below 0.20 across recent seasons, indicating well-calibrated probabilities.

Calibration is moderately strong in Ligue 1, Primeira Liga, Süper Lig, Championship, and the Belgian Pro League. Brier scores in the 0.20 to 0.22 range. The output is reliable enough for fan use, with occasional larger misses on fixtures featuring unusual squad disruption.

Calibration is weaker in cup fixtures, very early season fixtures (first five matchdays before in-season form is reliable), and fixtures involving sides on extended cold streaks where the model's expected goals scored figure may be lagging the side's true current level.

The relationship to lineup overrides

BTTS output is sensitive to lineup absences in much the same way as the win probability. A side missing its first-choice striker will have a lower expected goals scored figure, which lowers their contribution to BTTS yes. A side missing its first-choice goalkeeper will allow more goals, raising the opposition's contribution to BTTS yes.

When using the simulator's lineup-out override, the BTTS output updates alongside the win probability. The delta block shows both. A typical pattern: marking the away side's first-choice striker out drops away win probability by 3 points and drops BTTS yes probability by 4 to 6 points (because the side expected to score is now diminished).

Reading BTTS in context

The BTTS output is most useful when read as part of the full analysis, not as a standalone signal. The win probability tells you who is favoured and by how much. The over/under tells you whether the fixture is likely to be open or tight. The BTTS tells you whether the goals are concentrated on one side or distributed.

Together, the three outputs paint a picture of fixture texture that any one alone misses. A 60-percent home favourite with 65 percent over and 70 percent BTTS yes is a different read than a 60-percent home favourite with 40 percent over and 35 percent BTTS yes. Both have the same headline winner; the texture is wildly different.

BTTS modeling is included in Basic and Premium tiers across iOS, Mac, and Android, with the same calibration on every platform. Free tier shows win probability only.

Frequently Asked Questions

What is BTTS in football analysis?

BTTS stands for 'Both Teams to Score'. The output is a probability that both the home and away sides score at least one goal during the fixture. It is a separate prediction from total goals: a 1-1 fixture is BTTS yes and over 1.5 but under 2.5. A 4-0 fixture is BTTS no and over 2.5.

How does Tactiq compute BTTS probability?

Each side has an expected goals scored figure for the fixture. The model treats each side's goals scored as drawing from a Poisson-like distribution. BTTS yes is the probability that both distributions produce at least one goal. The math factors out the probability that either side blanks.

Why is BTTS modeled separately from over/under 2.5?

The two outputs answer different questions. Over/under is about total goals. BTTS is about who scored them. A 4-0 result is over 2.5 but BTTS no. A 1-1 result is BTTS yes but under 2.5. The two outputs can move in opposite directions in the same fixture.

Which leagues have the best BTTS calibration?

BTTS calibrates well in leagues with stable fixture distributions and reliable squad data: Premier League, Bundesliga, La Liga, Serie A, Eredivisie. Calibration is weaker in leagues where fixture variance is high (cup competitions, end-of-season dead rubbers in lower divisions).

When does BTTS probability change my read of a fixture?

Most when one side has a clearly weak attack or a clearly leaky defense. A fixture between a 1-goal-per-game attack and a 0-conceding-per-game defense will show low BTTS yes probability. A fixture between two open sides will show high BTTS yes. The output is most informative when it diverges from over/under: high over with low BTTS suggests a probable rout.

Is BTTS available on Free tier?

No. Free tier shows win probabilities only. BTTS and over/under outputs are included in Basic tier across iOS, Mac, and Android, with the same calibration on all platforms.