Trifecta strategy, done without tips, means putting a chance on every finishing order from the win market and buying only the orders that look set to pay more than that chance deserves. The hypothetical 10-runner race below can finish in 10 x 9 x 8 = 720 orders. Even the likeliest one comes up only about once in 49 races on the raw Harville formula, or once in 71 when its known lean towards fancied runners is corrected.
A trifecta betting strategy built on prices rather than tips has four steps. The first two turn the win market into fair win chances, then into each order's chance through the Harville formula, corrected for its bias. The last two weigh anchoring the winner against spreading the places, and test each order's fair dividend against its approximate before the jump. Trifecta betting sets out what each layout costs, and horse racing betting covers the rest of race day.
Step 1: fair win chances for a hypothetical 10-runner race
The race below is made up, with illustrative win prices and no real runners. Each price implies a chance of 1 / price, and the ten implied chances add up to 110.09%. The extra 10.09 points are the margin, so scale every implied chance down by the same factor:
fair chance = (1 / price) / 1.1009
| Runner | Win price | Fair chance |
|---|---|---|
| 1 | $3.40 | 26.72% |
| 2 | $4.60 | 19.75% |
| 3 | $6.00 | 15.14% |
| 4 | $8.00 | 11.35% |
| 5 | $11.00 | 8.26% |
| 6 | $15.00 | 6.06% |
| 7 | $19.00 | 4.78% |
| 8 | $26.00 | 3.49% |
| 9 | $34.00 | 2.67% |
| 10 | $51.00 | 1.78% |
The fair chances are rounded to two decimals and add up to exactly 100%, and every figure below is worked from them. This proportional method takes the same share off every runner; other methods take more off the outsiders, as removing the bookmaker margin explains. The fair odds calculator runs it on a whole market as its multiplicative method.
Step 2: the chance of each trifecta order
Harville's assumption is simple: take the winner out, and each remaining runner's chance of second is its win chance scaled up to fill the gap; take the second out too, and the same scaling sets third. For runners i, j and k in that order:
chance = pi x pj / (1 - pi) x pk / (1 - pi - pj)
For the order 1-2-3: 0.2672 x 0.1975 / 0.7328 x 0.1514 / 0.5353 = 0.02037, or 2.037%. Its fair dividend is 1 / 0.02037 = $49 per $1. The chances of all 720 orders add up to 100%, and the Harville formula entry works the two-place version. In a spreadsheet, give each order a row and sum the column for a ticket's chance.
Where Harville goes wrong: second and third
David Harville published the formula in 1973. Benter's 1994 report tested it on Hong Kong races from September 1986 to June 1993. Fancied runners filled second and third less often than Harville predicted and outsiders more often, and the effect was stronger for third, because the contest for the minor places is more random than the formula assumes.
Benter's correction flattens the win chances before sharing out second and third, by raising each one to a power below 1: 0.81 for second and 0.65 for third, fitted to those Hong Kong races. They are used here only to show the size of the effect; a correction for Australian racing would need fitting to a large set of local results. For the order 1-2-3:
- Runner 2's share of second = 0.1975 to the power 0.81, over that power summed for Runners 2 to 10 = 0.2688 / 1.1356 = 0.2367
- Runner 3's share of third = 0.1514 to the power 0.65, over that power summed for Runners 3 to 10 = 0.2931 / 1.3149 = 0.2229
- corrected chance = 0.2672 x 0.2367 x 0.2229 = 0.01410, or 1.410%, a fair dividend of 1 / 0.01410 = $71
The correction moves chance away from orders with fancied runners placing and towards orders with outsiders placing. The order 1-2-10 rises from 0.2395% to 0.3507%.
Step 3: anchor the winner or spread the minor placings
The chance a ticket lands is the sum of the chances of the orders it covers:
| Ticket | Combinations | Raw Harville | Corrected |
|---|---|---|---|
| A: Runner 1 standout to win, 2 to 5 for second and third | 12 | 13.4% | 10.7% |
| B: Runner 1 to win, 2 to 5 second, 2 to 8 third | 24 | 18.3% | 16.3% |
| C: box of Runners 1, 2, 3 and 4 | 24 | 26.6% | 19.6% |
| D: Runner 1 as a roving banker, with 2 to 5 | 36 | 34.6% | 26.2% |
| E: Runner 1 to win, 2 to 8 for second and third | 42 | 23.2% | 21.6% |
Anchoring the winner turns a ticket into a bet on one runner's win chance. Tickets A, B and E all need Runner 1 to win, so none can land in more than 26.72% of races. If you rate Runner 1 at 30% instead, every order with Runner 1 winning is about 30 / 26.72 = 1.12 times as likely as these figures say, if the other runners give up that chance in proportion.
Spreading the minor placings buys the part of the race raw Harville gets wrong. Ticket B adds 12 orders to A, each with Runner 6, 7 or 8 third, and keeps 16.32 / 18.35 = 89% of its raw chance after correction. The box keeps 19.56 / 26.56 = 74%, because all 24 of its orders put fancied runners in the places.
The box still lands more often for the same 24 combinations, but its lead over B shrinks from 26.56 - 18.35 = 8.21 points to 19.56 - 16.32 = 3.24. Anchor when you rate one runner above its price; spread second and third when you do not.
Risk: Betting involves risk. A higher chance of landing is not a higher return, these chances are estimates, and even ticket D, which lands most often here, misses about three races in four on corrected chances. See responsible gambling for limits and support.
Big flexi boxes or full units on small ones
Spread $12 across a ticket and each order carries $12 divided by the number of orders, so a landed ticket breaks even at a dividend per $1 equal to its combination count and loses money below it. Three ways to spend $12 on the same race:
| $12 on | Combinations | Flexi | Lands (corrected) | Dividend needed to return $12 | Landings returning under $12 at fair dividends |
|---|---|---|---|---|---|
| Ticket A at $1 a combination | 12 | 100% | 10.7% | $12 | none |
| Box of Runners 1 to 6 | 120 | 10% | 49.9% | $120 | 27.5% |
| Box of Runners 1 to 8 | 336 | 3.57% | 78.6% | $336 | 60.0% |
The 8-runner box lands in nearly four races in five, but 84 of its 336 orders have a fair dividend under $336, and those orders account for 60% of its landings. Those wins lose money before commission takes its share. Ticket A lands far less often, but every order on it has a fair dividend above $70, so at fair dividends each landing returns more than five times its $12.
If every order paid exactly its fair dividend, each of the three tickets would return its $12 on average, and commission comes off all of them. So shape sets the rhythm of the returns, frequent and small or rare and large, while value comes only from the prices of the orders you hold. The flexi calculator gives the count and percentage for tickets A, C, D and E and both $12 boxes. It multiplies group sizes, so B, whose groups share Runners 2 to 5, comes out too high: count it by hand, 1 x 4 x 6 = 24.
Risk: Betting involves risk. Ticket A misses in about nine races in ten here, and a run of misses can last far longer than a budget expects. See responsible gambling for limits and support.
Step 4: fair dividend against the approximate before the jump
The tote pays a winning trifecta order the pool, after refunds and commission, divided by the units on that order: Queensland's Wagering Rule, in effect from 22 September 2026, sets it out that way (ss 100 and 149). An approximate runs the same sum on the pool as it stands, so it shows what an order would pay if betting closed now. Set it against your chance:
value ratio = chance x approximate
A ratio above 1 means the pool is offering more than your chance warrants, if the figure holds. With illustrative approximates five minutes before the jump:
| Order | Approximate | Raw chance | Raw ratio | Corrected chance | Fair dividend, corrected | Corrected ratio |
|---|---|---|---|---|---|---|
| 1-2-3 | $58.40 | 2.037% | 1.19 | 1.410% | $71 | 0.82 |
| 1-5-2 | $121.80 | 0.9149% | 1.11 | 0.7421% | $135 | 0.90 |
| 2-1-7 | $192.60 | 0.5872% | 1.13 | 0.5905% | $169 | 1.14 |
| 1-2-8 | $205.60 | 0.4695% | 0.97 | 0.5432% | $184 | 1.12 |
| 1-2-10 | $338.00 | 0.2395% | 0.81 | 0.3507% | $285 | 1.19 |
Raw Harville would buy 1-2-3 and 1-5-2, which the corrected chances rate at 0.82 and 0.90, and pass 1-2-8 and 1-2-10, the two orders with the longest prices third. The two agree only on 2-1-7. Taking only the orders that clear 1.10 on corrected chances, a margin for late money, leaves 2-1-7, 1-2-8 and 1-2-10: three straight trifectas, $6 at $2 each.
That ticket lands 0.5905% + 0.5432% + 0.3507% = 1.48% of the time, about one race in 67. It misses 50 races in a row (1 - 0.0148) to the power 50 = 47% of the time.
Its average ratio, (1.14 + 1.12 + 1.19) / 3 = 1.15, holds only if the corrected chances are right and the approximates do not move. They do: they change until betting closes, your own money lowers the dividend on the orders you back, and the final figure is declared after the result, as tote approximates explains. Operators differ in which exotic approximates they show.
Risk: Betting involves risk. A ratio above 1 is an estimate built on other estimates, a ticket of three long orders can miss for a long run of races, and there is no guarantee of profit. See responsible gambling for limits and support.
Trifecta pricing mistakes that cost money
- Pricing orders with raw Harville. At $58.40, the order 1-2-3 looks 19% above its raw fair dividend of $49 and sits 18% below its corrected $71.
- Feeding Harville the bookmaker's prices without removing the margin. The order 1-2-3 then comes out at 3.090%, a fair dividend of $32 instead of $49, and the 720 orders add up to 143.4%, so orders look better value than they are.
- Treating a box as neutral. Box C bets as much on Runner 4 winning as on Runner 1, though Runner 1 is more than twice as likely to win, so the box overstakes its weakest runner.
- Reading an approximate ten minutes out as the price. A ratio of 1.12 can fall under 1 after one late bet on the order.
When to pass on the race
Pass when no order clears your margin on corrected chances. Shape alone cannot rescue a race the pool has priced well, because commission comes off every ticket. Pass, too, when you cannot see approximates for the orders you would hold.
Price the race again after a late scratching, because the win market re-forms and every order's chance moves. On the tote, any combination with the scratched runner in it is refunded and the others stand, as exotic scratching rules sets out.
Where the win prices come from
The Terminal shows the win market step 1 starts from: racing and sports prices from 40+ bookmakers side by side, runner by runner, with Betfair back and lay beside them. Prices are read on a repeating cycle, not tick by tick, so confirm each one in your bookmaker account before you bet.
B337 gives no ratings or selections, and a tote dividend cannot be compared in advance, because it does not exist until it is declared. A free account opens a limited view of the Terminal with live odds, and Betfair prices start with Terminal View. Betfair is a trade mark of its owner, and B337 is not affiliated with it. For free and confidential support call 1800 858 858 or visit gamblinghelponline.org.au.
Risk: Betting involves risk. A fair chance taken from the market is an estimate, and three runners still have to finish in one exact order. See responsible gambling for limits and support.