Keeneland September 2026 · Market Structure
What Really Changed at the Bottom of Keeneland September?
Examining BloodHorse's finding that the strongest percentage growth at Keeneland September came near the bottom of the market, and testing Tony Lacy's explanation against the underlying supply, shows that far fewer inexpensive horses reached the ring.
One of the more interesting pieces of post-sale analysis came from BloodHorse. As it regularly does at the conclusion of major sales, BloodHorse divided the 2026 Keeneland September results into ten equal price groups to see where the market actually strengthened.
This year's pattern stood out. The eighth decile increased 16.0%, the ninth rose 17.9% and the bottom decile rose 22.3%. Growth at the top continued as well, with 70 yearlings bringing at least $1 million compared with 36 two years ago, but the largest percentage gains came much farther down the market.
Tony Lacy, Keeneland's vice president of sales, offered an explanation that immediately caught my attention. The North American foal crop has been shrinking for two decades, and the mares and stallions most likely to leave production along the way tend to carry the least commercial value. Breeders have also become more selective about which families they carry, which mares they breed and which stallions they use.
The argument is intuitive. Keep removing horses from the bottom of the breeding population and the average horse left behind should improve.
The decile table tells us where prices moved. A fixed-price look at the sale tells us something more useful about why.
The bottom got smaller
A decile is a rank bucket. It always holds roughly 10% of the sold population, so its average can rise simply because the composition of the bucket changes.
The cleaner test is to hold the price constant.
Here is the number of Keeneland September yearlings that sold for $10,000 or less over the last five years:
| Year | Sold for $10,000 or less |
|---|---|
| 2022 | 326 |
| 2023 | 338 |
| 2024 | 366 |
| 2025 | 352 |
| 2026 | 296 |
Figure 1 · The bottom got smaller
Fifty-six fewer yearlings sold for $10,000 or less than a year ago.
Show the data
| Sale | Sold for $10,000 or less | Share of all sales |
|---|---|---|
| 2022 | 326 | 11.0% |
| 2023 | 338 | 11.7% |
| 2024 | 366 | 12.6% |
| 2025 | 352 | 11.4% |
| 2026 | 296 | 9.8% |
There were 56 fewer horses at $10,000 or below than last year, and 2026 produced the smallest group in the five-year period.
The same thing happened farther up the market. The number of horses selling for $25,000 or less fell from 807 in 2025 to 698 in 2026. As a share of all sales, the sub-$10,000 population dropped from 11.4% to 9.8%.
That changes how I read the BloodHorse decile result.
In 2025, the bottom decile contained 306 horses and topped out at exactly $10,000. This year only 296 horses in the entire sale brought $10,000 or less. To fill a 307-horse decile, the bucket had to reach into the $11,000 horses.
The average therefore rose from $5,358 to $6,552, a 22.3% increase, because there were fewer very inexpensive horses available to occupy the bottom of the distribution.
The 22.3% gain was driven in large part by composition: there were fewer very inexpensive horses available to fill the bottom decile.
The cheap horses disappeared before the hammer
Clearance helps tell us where the supply went.
Among horses in this price range, clearance improved from 91.7% to 93.7%, the strongest clearance of any price tier in the sale. Meanwhile, offerings in the $10,000-and-under band fell from 384 to 316, a 17.7% decline.
The sale as a whole barely changed by comparison. The number cataloged fell 1.1%, offered fell 1.5% and sold fell 1.5%. The supply of horses in the $10,000-and-under band contracted roughly an order of magnitude faster.
That matters because it separates the bottom-end change from the slight contraction in the overall sale. Keeneland had almost the same size marketplace, yet substantially fewer horses reached the ring at the very bottom.
The inexpensive horses that did arrive also cleared at a very high rate.
Far fewer horses reached the ring at the very bottom of the commercial market.
That brings us back to Lacy.
Lacy's explanation looks stronger through this lens
Lacy told BloodHorse that breeders have become increasingly selective about the mares and families they continue to carry. He also pointed to intense competition for placement at Keeneland, saying the sale received around 3,000 requests for roughly 1,600 spots in Books 3 and 4.
Those are two different filters operating before a horse sells.
The first happens in the breeding shed. Breeders decide which mares remain in production, which families justify continued investment and which stallions receive support.
The second happens at the sale company. Keeneland selects from a much larger pool of available yearlings and decides which horses enter the catalog and where they belong.
Both filters can reduce the number of commercially weak horses reaching the ring.
The five-year fixed-threshold numbers fit that process remarkably well. Fewer horses sold for $10,000 and below. Fewer sold for $25,000 and below. Clearance among the inexpensive horses that remained improved.
That gives Lacy's explanation considerably more support than the decile averages alone.
It also leaves room for a second mechanism working alongside the shrinking foal crop.
Keeneland is selecting harder too
The Keeneland catalog represents a selected portion of the North American yearling population.
If 3,000 horses are competing for roughly 1,600 places in Books 3 and 4, the standard for making those books can rise. A horse that once fit comfortably in Book 3 can land in Book 4. A previous Book 4 type can move deeper. Other horses move outside Keeneland altogether.
Commercial quality gets compressed down through the catalog.
We saw versions of this throughout September, and the Fasig-Tipton October catalog gives us a look at the downstream population. More than 500 October yearlings were cataloged at Keeneland September, including 423 withdrawals and 92 that went through the ring and RNA'd.
So the lower end of Keeneland can improve through two related processes: a stronger population arriving at the sale and a tougher selection process determining which horses make the catalog.
The 296 horses at $10,000 and below are probably telling us something about both.
Other forces can amplify the effect
Once the supply change is established, several other parts of the market can help explain why the remaining horses found support.
Sire concentration is one. The breeding population has become increasingly concentrated around stallions the market already supports. In our Fasig-Tipton October model, sire identity alone accounted for about 45% of the variation in log price among 1,037 sold yearlings. That figure carries far more than genetics, including mare quality, stud fee, breeder quality and commercial fashion, but it shows how much information travels with the sire.
Buyer behavior matters too. Lacy pointed out that major owners and trainers were shopping the later books. Counting Stars gives buyers an easy recent reason to keep looking. She sold for $13,000 during the tenth session of Keeneland September in 2024 and became a multiple Grade 1 winner. One horse is an anecdote, but successful horses emerging from the back of the sale encourage serious buyers to keep working those sessions.
Capital cascades downward. A buyer who once found plenty of horses around $250,000 may now encounter similar commercial profiles at $400,000. That pushes the buyer into another part of the market, where the same process repeats. Our September work has already shown many of the same buyers spending materially more than they did a few years ago.
Purses support the middle and lower market. A useful racehorse has substantial earning opportunities in Kentucky today. Better economics for ordinary racehorses give buyers room to pay more for them. Healthy purses raise the value of competence.
Pinhookers add another bid. At Fasig-Tipton October last year, 439 of 1,103 sold yearlings later appeared in a major juvenile-sale catalog. In the $50,000 to $100,000 range, the figure was 54%. Those buyers are thinking about physical development, athleticism, breeze potential and resale value, creating another source of demand beneath the elite end of the yearling market.
All of those forces can strengthen prices. The fixed-threshold data suggest they were operating on top of a more basic shift in supply.
There were simply fewer very inexpensive horses to buy.
Commercial quality and racing quality
This is the distinction I keep coming back to.
The 2026 sale gives us good evidence that the commercial population at Keeneland became stronger at the bottom. Fewer horses sold below fixed low-price thresholds, the inexpensive horses that reached the ring cleared at a very high rate, and Keeneland was selecting from an enormous pool of candidates for the middle books.
That is meaningful evidence in favor of Lacy's argument.
The racing question takes longer.
A more selective breeding population should eventually produce better outcomes on the track if the improvement extends beyond commercial appeal. Sale prices give us the market's judgment today. Racing records tell us what those horses became.
Follow the horses
Over 15 or 20 years, we can track each Keeneland book and ask what percentage became starters, winners, black-type horses and graded stakes winners. We can then see whether racing production from the later books improved as the North American foal crop contracted, whether the gap between the front and back of the sale narrowed, and how much of that movement reflects the changing sire population.
That gets us closer to the question Lacy's explanation ultimately raises: whether fewer weak commercial horses at the bottom of Keeneland also means better racehorses coming out of the population.
Methodology. Decile figures are BloodHorse's, from its MarketWatch analysis of the 2026 Keeneland September Yearling Sale. Everything else uses Thoroughbred Intelligence's normalized Keeneland September exports for 2022 through 2026, covering 22,092 yearling records and including identifiable post-sale transactions.
The fixed-price counts use final sale price for horses that sold, so they describe the sold population only. For the clearance-by-price-tier analysis, unsold horses have no final sale price, so they are assigned to a price band using their reported RNA bid. Withdrawals have no price of any kind and are excluded throughout. Clearance within a band is horses sold divided by horses sold plus horses that failed to meet reserve. The same method is applied to 2025 and 2026.
Catalogued, offered and sold counts describe the whole sale: 4,692 catalogued, 3,738 offered and 3,078 sold in 2025 against 4,642, 3,682 and 3,033 in 2026.