Group Three
That's the answer. The question, teased at the end of the first piece in this series, was: at what class does the median horse stop paying for its own keep?
Group Three.
On a full campaign of six or more starts, the median horse does not net enough prizemoney to cover the cost of being in training at any class below Group 3. On all runners (including those with fewer starts), the line doesn't cross until Group 1.
I tested it against every cost assumption I could find. It survived all of them.
What the owner actually receives
The first thing this piece needed was a number the earlier pieces didn't carry: the gap between gross prizemoney (what the database measures) and net prizemoney (what lands in the owner's bank account).
In NSW, the deductions are published. Jockey welfare: 1%. Equine welfare: 1.5%. Then the connections' percentages under Local Rule 72(2): trainer 10%, jockey 5%, stablehand 2%. The owner nets 80.9% of gross. A horse that earns $100,000 in gross prizemoney delivers $80,900 to its owner before a single training bill is paid.
Victoria's splits are structured differently and the exact percentages need confirming from their Rules of Racing, so the class-level analysis below uses NSW costings only. The shape will be similar; the precise crossing point may differ by a class or two.
The cost anchors
Racing NSW's 2024 Strategic Plan surveyed trainers across all tiers. Annual training, racing, spelling, and veterinary costs for a horse in work eight months and spelling four:
Metro: $70,023. Provincial: $57,166. Country: $47,452. Hobbyist owner-trainer: $18,564.
These don't include the cost of buying or breeding the horse. They don't include transport to the races. And they don't include riding fees, which are a per-start cost on top. At $292.69 per ride in NSW, a horse that starts twelve times a season adds $3,512 before it has earned a cent.
Where the money runs out
The database holds every race, every runner, every dollar paid across two seasons. I grouped runners by the highest class they reached in the season, then summed all their earnings (not just earnings at that class). The median was computed net of the 19.1% deductions above.
A country maiden horse, the most common runner in the dataset, earns a median of roughly $750 net per start. Its season costs about $16,000 per start at the country training rate. For every dollar the median maiden earns, its owner spends more than twenty.
Moving up the class ladder, the median earnings climb, but the cost line barely moves (it steps from country to provincial to metro as the horse graduates). The crossover, where net median season earnings first match annual training costs, happens at Group 3 for horses with a full campaign. Below that, every class produces a median horse that loses money. The break-even share of runners covering costs sits below 50% at every level through Listed.
The hobbyist rate ($18,564) brings the crossing point down to roughly BM 70, which is Class 3-4 in NSW terms. That horse nets about $21,000 to $26,000 a season. If you train your own horse, feed it from your own paddock, and do the veterinary minimum, the median Class 3-4 horse roughly washes. If you pay a professional trainer, it doesn't.
The sex split that isn't
One finding I expected to matter and it didn't. Geldings and mares earn about the same at every class level. The difference is what happens after racing: a mare carries broodmare residual value that can partially offset the racetrack loss; a gelding's racetrack economics are the whole economics. Entire males are too few below Listed to quote reliably, but the handful at the top carry the lottery ticket of a stallion career.
The practical consequence: if you own a gelding (and most horses in training are geldings), the numbers above are the complete picture. There is no second act.
The reconciliation
Racing NSW's Strategic Plan states that NSW owners collectively pay $567 million and receive $295 million net, a deficit of $273 million. The database, using the same cost anchors and net prizemoney, implies a deficit of $149 to $166 million. The gap is the horses the data cannot see: foals that were bred, paid for, reared, and never started. Every one of them is pure cost with zero return, and they sit in RNSW's aggregate but not in the race results. The fact that the implied deficit is roughly 60% of the stated one is consistent with a 55-72% lifetime starter rate, which is the same range piece 3 used.
Racing NSW's own Strategic Plan reports the same deficit from the top down: $567 million in costs, $295 million returned, 51.9 cents on the dollar. The database arrives at the same picture from the bottom up, horse by horse.
State bonus schemes
Three schemes materially improve the returns for qualifying horses and deserve a mention: BOBS (NSW), VOBIS (VIC), and QTIS (QLD). These pay bonuses on top of prizemoney to horses by registered sires, and for eligible owners they shift the break-even line down by one or two classes. They don't change the structural finding, because the majority of horses at the lower end of the class ladder don't qualify, but for a horse that does, the bonus can be the difference between a net loss and a wash. Anyone modelling their own position should check eligibility before reading the unsubsidised numbers above as their personal P&L.
What piece one promised
The first piece in this series ended with a question: at what class does the median horse stop paying for its own keep? The answer survived every cost assumption I threw at it. The crossing point moved from Group 1 to Group 3 depending on whether you counted all runners or full campaigns, and from Group 3 to BM 70 if you used the hobbyist cost instead of the professional one. It never dropped below Class 3-4 on any combination. Below that line, the median horse is a consumption good, not an investment, and the industry's own numbers say so.
None of this is hidden. Racing NSW publishes the cost survey. The prizemoney is public. The deduction percentages are in the Rules of Racing. The only thing missing was someone putting them in the same spreadsheet and drawing the line.
Notes
NSW only, FY2024-25 costings from Racing NSW Strategic Plan 2024 (trainer survey: 8 months in work, 4 months spelling). Prizemoney net of jockey welfare (1%), equine welfare (1.5%), trainer (10%), jockey (5%), stablehand (2%) = owner receives 80.9% of gross. Riding fee of $292.69 per start (NSW Jockeys' Association published schedule) treated as an additional per-start owner cost. Earnings from the two-season database (2024-25 and 2025-26), grouped by highest class reached in the season, all starts included. Class-to-benchmark mapping follows NSW conventions: CL1 = BM 58, CL2 = BM 64, CL3-4 = BM 70, CL5-6 = BM 78. Break-even defined as net median season earnings >= annual training cost at the applicable venue tier. State bonus schemes (BOBS, VOBIS, QTIS) are not included in the prizemoney figures and would lower the break-even line for qualifying horses. The implied NSW owner deficit of $149-166M against RNSW's stated $273M is reconciled by the non-starting horse population (55-72% lifetime starter rate). All analysis is my own; queries and figures available to anyone who wants to check them.