Tennis Trading on Betfair: What the Pre-match Prices Do
Our 38-day exchange dataset holds 3,346 tennis markets: what pre-match prices do, why a median volume of 40 EUR makes most of them untradeable.
Our dataset is, by accident of coverage, a tennis dataset first and a football dataset second. Over 38 consecutive days from 31 July we recorded 6,027 distinct pre-match markets on the exchange. Of those, 3,346 were tennis and 2,681 were football. That ratio is not a preference. It is simply what the calendar produces: a busy football weekend might offer a few dozen fixtures across the leagues we track, while a single ATP 250 plus a WTA event plus the Challenger and ITF layers can produce more matches in one afternoon than football manages in two days.
So when the subject is tennis trading Betfair markets pre-match, we have more raw material than for anything else. What follows is what that material shows, and — more importantly — what it does not.
Two outcomes, not three
The structural difference between a tennis match odds market and a football match odds market is not cosmetic. It changes the arithmetic of every position.
Football gives you three runners: home, draw, away. That third runner is a working part. It gives a trader something to lay when they want exposure to “not this outcome” without buying the opposite team. It gives a hedging leg. It absorbs a chunk of the overround, and it is usually the least efficiently priced selection in the book, because the draw is the outcome fewest people have an opinion about.
Tennis has two runners. Player A, Player B. Backing A at 1.80 and laying A at 1.70 is the whole trade. There is no third price to lean on, no way to construct a position that profits from stalemate, because stalemate does not exist. The book is cleaner and, in liquid markets, tighter. It is also less forgiving: with two selections, any price you get is directly the inverse of the price on the other side, so there is nowhere to hide a bad entry.
The practical consequence is that pre-match tennis trading reduces to one question — did the price move in the direction of the position — with none of the multi-leg construction that football permits.
The retirement problem
A tennis market can end without a result in a way football almost never does. A player retires in the second set. A medical timeout turns into a walkover. On the exchange these situations are handled by the rules of the specific market, and a match that never reaches a completed result can leave a pre-match position settled in ways that had nothing to do with the price analysis that justified it.
This is not a marginal risk in the lower tiers. Challenger and ITF events, played by athletes managing schedules, travel and injury on thin budgets, produce retirements at a rate that would be a scandal in top-flight football. Any statistical work on tennis prices that does not account for retirements is measuring something slightly different from what it thinks it is measuring. Ours records the recorded outcome, whatever produced it, and we flag that as a known imprecision.
Ranking gaps produce prices football never sees
The most extreme pre-match football price in a normal domestic fixture might be 1.10 against 25.00. In tennis, a first-round draw that pairs a top-ten seed against a qualifier ranked outside the top 300 routinely produces 1.02 against 30.00 or worse. Individual sport, no substitutes, no eleven players to dilute a talent gap.
Those extreme prices matter for two reasons. First, the tick structure at short odds is brutal: between 1.01 and 1.10 the increments are 0.01, and a single tick at 1.02 is a large proportional move. Second, implied probabilities above 95% are exactly where the overround distorts most, and where a naive reading of “the market says 97%” is least reliable. Normalising for overround, which we do, is not optional in tennis the way it can be fudged in football.
The liquidity figure that reframes everything
Here is the number that should come before any discussion of strategy. Across all 6,027 pre-match markets we recorded, the median traded volume was 40 EUR.
Not 40,000. Forty.
The distribution:
| Traded volume | Share of markets |
|---|---|
| Over 1,000 EUR | 13.6% |
| Over 3,000 EUR | 6.0% |
| Over 10,000 EUR | 2.4% |
| Over 50,000 EUR | 0.7% |
The single busiest market we observed traded 369,000 EUR. The gap between that and the median tells the story: money on the exchange concentrates in a small minority of matches, and everything else is a price display with almost nothing behind it.
For tennis specifically, this is the dominant fact. The sport generates volume — more markets than football in our sample — but the volume per market is spread across a long tail of ITF and Challenger fixtures that nobody is trading. A market that has seen 40 EUR matched has no meaningful spread, no depth, and a price that can be moved several ticks by a bet the size of a restaurant bill.
That reframes the usual question. People ask which pre-match tennis strategy works. The prior question is which tennis matches are tradeable at all, and the honest answer from our data is: roughly one in sixteen, if 3,000 EUR is the threshold. Everything below that is not a market being analysed; it is noise being interpreted.
This is why any serious analysis has to filter by volume before it looks at a single price. Running a study across all 6,027 markets would produce numbers dominated by matches where the “price movement” is one person changing their mind.
What the filtered sample showed
Applying that filter to the first month of data left 247 markets with more than 3,000 EUR traded. For each selection we compared the first and last matched price before the start, and compared the actual outcome against the probability implied by the final price, normalised for overround.
- Selections that shortened by more than 3% won 85 times against the 99 their own closing price implied. z = −2.03, p = 0.04.
- Selections that drifted by more than 3% won 75 times against the 61 implied. z = +2.20, p = 0.03.
- The control group — prices that moved less than 3% either way — came in at z = +0.10. No deviation at all.
The reading is straightforward: over this sample, the market appears to have overreacted to pre-match money flow. Prices that were pushed in moved too far in; prices that drifted drifted too far out. The control group behaving perfectly is what makes the pattern interesting rather than an artefact of the method — if the method were broken, the flat prices would be wrong too.
Why that result is a hint and not a finding
Five things have to be said plainly.
One: it is a single month of data. Thirty-eight days is not a season.
Two: the month is August. In football that means early season, squads unsettled, promoted teams unpriced. In tennis it means the North American hard-court swing, with its own particular mix of players skipping events, arriving underprepared, or peaking for the last Slam. Neither sport is at its most representative in August.
Three: we tested several groups. Testing multiple thresholds and then reporting the ones that cleared significance is exactly how a p = 0.04 stops being worth 0.04. The true significance of these figures, adjusted for the number of comparisons made, is weaker than the raw numbers suggest.
Four: the calculations do not include exchange commission. A statistical edge measured before costs is not an edge after them, and on the exchange the commission is applied to net market winnings, which makes the correction non-trivial rather than a flat subtraction.
Five: the 247 markets are football and tennis combined. We have not split the result cleanly by sport, and with a sample that size, splitting it would leave both halves too small to say anything. So the overreaction pattern is a cross-sport observation that includes tennis. It is not a tennis-specific finding.
What this dataset structurally cannot tell you
Our capture runs pre-match only, four snapshots a day. That is a hard boundary, and it excludes the thing most people mean when they talk about tennis trading.
Most tennis trading on the exchange happens in-play. That is where the sport’s defining feature lives: the scoring system is a series of discrete, resettable units, so a single break of serve can move a price from 1.60 to 1.25 and back inside twenty minutes. Momentum, holds, break points, the third-set shift — none of it appears anywhere in our data.
The limit exists for a practical reason. Capturing in-play tennis properly means point-by-point resolution, tens of thousands of price updates per match, and a data pipeline an order of magnitude heavier than four daily snapshots across 6,000 markets. We chose breadth over depth. That was a deliberate trade-off, and it means anyone reading this for in-play insight is reading the wrong dataset.
What the data licenses
It licenses three statements.
That tennis dominates exchange market count without dominating exchange money, and that the median pre-match tennis market is far too thin to be traded meaningfully.
That among the small minority of markets with real volume, pre-match price movement over one month showed a measurable tendency to overshoot, with stable prices showing no bias at all.
That nothing here describes in-play behaviour, and nothing here has been tested over enough time, across enough conditions, or after enough costs to be treated as a rule.
Anything beyond those three statements would be reading more into 38 days than 38 days contains.
This content is informational and analytical in nature. It is not advice to gamble and does not constitute a recommendation to place any bet. Gambling carries financial risk and can be addictive. 18+.