Both Teams To Score (BTTS): When It's Actually Value

Both Teams To Score is one of the most popular football markets in Europe — and one of the most narrative-driven. Pundits sell it with phrases like 'leaky defences on both sides' or 'this fixture always delivers goals.' None of that is a price. BTTS is a probability question with a clean mathematical answer: what is the chance that each team scores at least one goal? If you can estimate that number better than the bookmaker's implied price, you have value. If you can't, you are paying the margin for a story.

What the BTTS market actually is

BTTS (also written GG/NG, from the Italian 'gol/no gol') is a two-way market. Yes wins if both teams score at least once — 1-1, 2-1, 4-3 all count, and it doesn't matter who wins. No wins if at least one team fails to score — 0-0, 1-0, 3-0 all count. There is no draw outcome and no push: every match settles one side or the other.

That two-way structure makes BTTS easy to analyse. Convert both prices to implied probabilities, strip the margin, and you have the market's estimate. A typical soft-book price of 1.72 (Yes) / 2.05 (No) implies 58.1% + 48.8% = 106.9% — a chunky 6.9% overround, noticeably worse than what the same book charges on the main 1X2 market. If the margin concept is new to you, read devigging explained first — everything below builds on it.

How a model prices BTTS: from goal rates to probability

The standard approach — and the one our football model uses — starts from a Poisson model: estimate each team's expected goals in this specific fixture (call them λ_home and λ_away) from attacking and defensive strength, home advantage, schedule and lineup information. Once you have the two goal rates, BTTS follows directly, because the probability that a Poisson-distributed team scores zero goals is e^(−λ).

• P(home scores at least once) = 1 − e^(−λ_home)
• P(away scores at least once) = 1 − e^(−λ_away)
• P(BTTS Yes) ≈ the product of the two (if the scoring processes were independent)

The inputs matter more than the formula. λ estimates built on shots-on-target and expected goals are far more stable than ones built on raw recent scorelines — a team that won 3-0 from three shots is not a three-goal attack. Garbage λs in, garbage BTTS price out.

Worked example: a mid-table Bundesliga match

Suppose the model rates a match at λ_home = 1.6 and λ_away = 1.2 — a fairly open game between two mid-table sides. Then:

• P(home scores) = 1 − e^(−1.6) = 1 − 0.202 = 79.8%
• P(away scores) = 1 − e^(−1.2) = 1 − 0.301 = 69.9%
• P(BTTS Yes), independent case: 0.798 × 0.699 = 55.8% → fair odds ≈ 1.79
• P(BTTS No): 1 − 0.558 = 44.2% → fair odds ≈ 2.26

Now compare against the market. If a soft book offers BTTS Yes at 1.95, the expected value is (0.558 × 1.95) − 1 = +8.8% — comfortably above a sensible minimum threshold. If the book offers 1.72, the EV is (0.558 × 1.72) − 1 = −4.0%: same match, same model, no bet. The number decides, not the fixture's reputation. This is the same expected value logic that drives every market we price.

The correction that matters: goals are not fully independent

The clean product formula above assumes the two teams' scoring is independent. Real football disagrees slightly, and it disagrees exactly where BTTS is decided: in low-scoring games. Empirically, 0-0 and 1-1 occur a bit more often than independent Poisson predicts, and 1-0/0-1 slightly differently too — game state feeds back into how teams play. The standard fix is the Dixon-Coles correction, which adjusts the probabilities of those low-score outcomes; our own model fits this dependence from over 100,000 matches rather than borrowing a textbook constant.

For BTTS the direction is what you should remember: a naive independent Poisson tends to underprice BTTS No because it underrates the 0-0. The effect is small — typically a fraction of a percentage point to a point of probability — but BTTS edges are small too. A model that ignores low-score dependence will systematically lean a little too hard on Yes.

The classic BTTS traps

BTTS attracts more folklore than almost any other market. Three patterns cost bettors real money:

• Streak-chasing: 'BTTS landed in six of their last seven.' Seven matches tell you almost nothing — the sampling noise on a ~55% event over 7 games is enormous. Books know the streaks too, and shade the price before you arrive.
• Narrative fixtures: derbies and 'must-win' games get sold as guaranteed goals. Historically, high-stakes matches are, if anything, slightly more cautious — and the price already reflects the story.
• Ignoring the margin: at a 7% overround you need a genuinely large modelling edge just to reach break-even. Betting BTTS 'for fun' at soft-book prices is one of the fastest quiet leaks in a betting bankroll.

None of this means BTTS is unbeatable. It means BTTS is beatable the same way every market is: with a calibrated probability estimate compared against a de-vigged reference price — never with a hunch about a fixture's personality.

So when is BTTS actually value?

In our experience pricing BTTS across eight leagues with the Poisson model, genuine value shows up in a few repeatable spots: when the market over-reacts to a short scoring or clean-sheet streak that the underlying shot data doesn't support; when lineup news (a missing striker, a reshuffled defence) shifts a team's λ and slower books haven't moved; and when a soft book's BTTS price is simply out of line with what the same book's own totals and 1X2 prices imply. BTTS, over/under totals and the correct score grid are all views of one underlying score distribution — when a book's BTTS price disagrees with its own totals price, one of them is wrong.

We apply the same discipline here as everywhere else: a minimum EV threshold before a bet is surfaced, a sharp-market sanity check on the price, and honest tracking of results per market. BTTS is a normal market with slightly lazy pricing at the soft end — treat it with normal rigour and it can contribute; treat it as entertainment and it will quietly bill you for it.