We Have the Data. Who’s Keeping Score?
Why Thoroughbred racing needs a shared understanding of its health, greater accountability for results and a more ambitious approach to working together.
In this article
- The Industry Is Running at Two Speeds
- We Keep Studying the Decline Without Keeping Score
- The Whole Picture Is Still Fragmented
- An Encouraging Start. An Opportunity to Think Bigger.
- The Numbers That Matter Most Have No Common Public Source
- A Shared Scorecard Would Cover the Business From the Breeding Shed to the Television Audience
- Where It Could Start
On Tuesday I posted a prototype dashboard of the North American Thoroughbred industry on X. I built it for fun, by hand, from public sources, without collaboration or outside input, so it is far from perfect. It was never meant to be finished. Sometimes the best way to share an idea is to build a rough version of it. This was my shot at it.
I have a real stake in this business. I am involved in breeding, buying, racing and selling Thoroughbreds, and I read the same industry reports and social media threads as everyone else. The threads are full of complaining and mud thrown in every direction. The organizations that run the sport publish a steady stream of press releases. What remains difficult to find is a common, public record of which industry initiatives are working, which are falling short and whether our collective efforts are changing the industry’s trajectory.
I knew going in that the data would be hard to find. Each series took its own hunt through The Jockey Club’s Fact Book, Equibase’s monthly wagering releases, the Equine Injury Database, HISA’s metrics reports and a stack of press releases, each on its own calendar and with its own definitions. Several of the numbers I wanted most have no consistent public source at all.
So I wrote my version of the Jerry Maguire memo, a few ideas nobody asked for, put down in the hope of starting a conversation and perhaps getting something moving. The idea at the center of it is one shared, public scorecard for the health of the whole industry, built from data its organizations already hold.
The Industry Is Running at Two Speeds
Start at the top of the sport. The 2026 Kentucky Derby averaged 19.3 million viewers on NBC and Peacock by NBC’s updated count, peaking at 24 million, the most-watched Derby on record. Churchill Downs handled a record $487 million over Derby Week. Keeneland September sold 70 yearlings for $1 million or more, against 3 in 2010. Japanese buyers’ share of the money spent on mares at Fasig-Tipton November has doubled since 2020, and those mares are exported to Japan.
Then look at the rest of the year. At The Jockey Club’s August Round Table, Charlotte Clement de Bona said more than 80% of the people who watched the Derby never watched another horse race all year. Underneath that audience, the everyday card keeps shrinking. The US ran 47% fewer races in 2025 than in 2000. The North American foal crop has gone from 51,296 in 1986 to an estimated 17,000 for 2026. After inflation, betting on US racing is down 59% since 2000. And FanDuel is phasing out FanDuel TV, the network built around that everyday card, though live racing stays on the channel through 2027 and on FanDuel’s apps.
Figure 1 · Two speeds
Million-dollar yearlings at Keeneland September went from 3 to 70 while US races fell 47%
Show the data
| Year | Keeneland September yearlings at $1M+ | US races |
|---|---|---|
| 2000 | 55,486 | |
| 2001 | 55,128 | |
| 2002 | 54,304 | |
| 2003 | 53,500 | |
| 2004 | 53,595 | |
| 2005 | 52,257 | |
| 2006 | 51,668 | |
| 2007 | 51,304 | |
| 2008 | 50,120 | |
| 2009 | 49,368 | |
| 2010 | 3 | 46,379 |
| 2011 | 6 | 45,417 |
| 2012 | 7 | 45,086 |
| 2013 | 18 | 43,139 |
| 2014 | 13 | 41,276 |
| 2015 | 11 | 38,941 |
| 2016 | 9 | 38,284 |
| 2017 | 13 | 37,628 |
| 2018 | 27 | 36,586 |
| 2019 | 22 | 36,207 |
| 2020 | 15 | 27,700 |
| 2021 | 15 | 33,567 |
| 2022 | 30 | 33,452 |
| 2023 | 30 | 31,746 |
| 2024 | 36 | 30,852 |
| 2025 | 56 | 29,401 |
| 2026 | 70 |
California shows what that looks like in one state. Golden Gate Fields ran its last race in June 2024, ending major Thoroughbred racing in Northern California. The state raced 738 days in 2005 and 246 in 2025.
Two speeds. Each row stands on its own; the years and populations differ from row to row.
| Measure | Earlier | Latest | Note |
|---|---|---|---|
| Top of the sport | |||
| Kentucky Derby TV audience, average | 17.7 million (2025) | 19.3 million (2026) | NBC and Peacock, NBC’s updated count; 2026 peak 24.0 million; most-watched Derby on record |
| Derby Week handle, all sources | $473.9 million (2025) | $487 million (2026) | Churchill Downs Incorporated; a record |
| Keeneland September yearlings sold for $1 million or more | 3 (2010) | 70 (2026) | Keeneland results |
| Japanese buyers’ share of mare gross, Fasig-Tipton November | 13.7% to 18.4% a year (2017 to 2020) | 27.7% to 32.2% a year (2023 to 2026) | Our count: mares in foal and racing and broodmare prospects, post-sales included, RNAs excluded |
| Everyday card | |||
| US races | 55,486 (2000) | 29,401 (2025) | 2025 was the first year under 30,000 since the mid-1950s |
| North American race days | 7,015 (2005) | 4,115 (2025) | |
| North American registered foal crop | 51,296 (1986) | 17,000 (2026) | 2026 is The Jockey Club’s estimate |
| US handle, before inflation | $15.18 billion (2003, the peak) | $11.04 billion (2025) | Down a further 2.7% through October 2026 |
| US handle, after inflation | 2000 | 59% lower (2025) | In 2025 dollars, CPI-U |
| California race days and starters | 738 days, 7,877 starters (2005) | 246 days, 3,504 starters (2025) | |
| California foal crop | 1,856 (2015) | 1,065 (2024) | |
| Major track closures | Arlington Park (after 2021), Golden Gate Fields (2024), Aqueduct (2026) | Aqueduct’s racing moved to Belmont Park and Saratoga; the NTRA counts roughly 75 Thoroughbred tracks | |
The trajectory is not up and to the right, and the big days are hiding it.
Then there is safety. There are two things we don’t talk about at the dinner table: politics and vet scratches. Safety is the most hotly debated subject in the sport, and the arguments never settle: medication, surfaces, how often a horse should run, and whether tracks inside HISA’s rules are safer than the ones outside them. It also offers one of the clearest examples of what sustained industry-wide measurement can accomplish. The Equine Injury Database has published a fatality rate every year since 2009, and that rate fell from 2.00 per 1,000 starts in 2009 to 1.07 in 2025, a provisional figure and the lowest since the database began. Whoever deserves the credit, the industry can show the progress because it committed to counting and reporting.
We Keep Studying the Decline Without Keeping Score
Ask around this business why the sport shrank and you will hear a long list: casino gaming, sports betting, computer-assisted wagering, HISA, the cost of keeping a horse in training, consolidation in the stallion barns. Each argument arrives with the two or three numbers that support it, but we lack a common scorecard for judging the results of our collective efforts.
The industry has invested heavily in studies, programs, new organizations and regulatory reforms. Some efforts have produced measurable improvements, particularly in safety. Yet the broader business continues to contract. Fewer horses, fewer races, fewer racing days and substantially less inflation-adjusted wagering raise a reasonable question: Which initiatives are improving the industry’s long-term health, and how do we know?
We have been here before. In 2011 The Jockey Club commissioned McKinsey to study the industry, and the Round Table heard its forecast: without new growth strategies, handle would fall 25% in the next decade and the foal crop at least 9%. The Jockey Club committed about $10 million over five years to the report’s recommendations. By 2020 the North American crop had fallen 30% from 2010, from 28,420 foals to 19,916, a crop conceived before the pandemic. Those numbers alone cannot say whether the recommendations worked or what caused the decline, and I have not found a public accounting of how each recommendation performed against its original goal.
Which initiatives worked? Which fell short? Where has a state or a track made progress that others could copy? Who is tracking results against the goals that were set when each initiative launched? Panels and symposiums are not a scorecard. Announcing an initiative is encouraging. It’s an acknowledgement. Measuring it in public is the test.
Some of the competition can be measured from public data. Americans bet $166.94 billion with legal sportsbooks in 2025, according to the American Gaming Association, about 15 times the $11.04 billion bet on US horse races that year. That is the scale of the competition for the gambling dollar.
The Whole Picture Is Still Fragmented
Part of the problem is the structure of the industry itself. Racing associations, track operators, wagering companies, regulators, breeders and horsemen have different constituencies, financial interests and priorities. Each measures success by its own responsibilities, and most of the work happens separately. In some cases the organization running an initiative also controls most of the information needed to judge whether it worked. Those overlapping roles create potential conflicts of interest, and they are why independent measurement matters.
At my last startup, EssentialDx, we helped management teams work through organizational problems by establishing a shared understanding of what was happening through data. It was often remarkable how quickly longstanding disagreements could be resolved once everyone was looking at the same facts. But I also learned that some leaders were reluctant to see the findings at all. Once a problem was documented and understood, there was an expectation that they would do something about it. That experience has stayed with me as I’ve worked on this industry dashboard.
An Encouraging Start. An Opportunity to Think Bigger.
After I built the prototype, I went back through this year’s announcements and found that The Jockey Club has started down this road. At the August Round Table, Clement de Bona, its chief strategy and transformation officer, who becomes chief operating officer on January 1, proposed shared measurement of participation across the industry and said, “That’s a specific buildable dashboard and I want us building it together.” She committed to publish every key industry metric and the results of every Jockey Club initiative, including the ones that do not work, and set a standard that no experiment runs without measurement built in before it starts. Chairman Everett Dobson’s June letter promised measurable outcomes for every initiative and introduced the Thoroughbred Data Hub with Equibase. His October letter reported traceability connections for about 60% of more than 376,000 horses analyzed and promised regular research reports and dashboards to set baselines and measure progress against them. That is the kind of transparency and accountability the sport needs.
I also find it refreshing to see someone in a senior industry leadership position who brings a modern understanding of technology, data analytics and experimentation to the conversation. Racing has no shortage of institutional experience. It has far fewer people who have built products, run experiments and been held to the numbers.
The dashboard she described starts with participation: how many horses we breed, how many reach the races, how long they stay active and whether they are creating new fans. She pointed to safety as her model, and she asked tracks to share what their own experiments teach. I would like to see it succeed, and I would challenge all of us to take it further.
Where I part ways with her is the diagnosis. Clement de Bona argued that racing has more of a supply problem than a demand problem, pointing to a 32% increase in handle per starter since 2015. She was careful to call it nominal. Consumer prices rose nearly 36% over the same period, leaving inflation-adjusted handle per starter slightly below its 2015 level. Our own data show real handle per race flat over the decade, $372,358 in 2015 against $375,395 in 2025, while total real wagering declined 24%.
Figure 2 · Per race and in total
Since 2015, real betting per race went from $372,358 to $375,395 while total real betting fell 24%
Show the data
| Year | US handle, $ millions (2025 dollars) | Handle per race (2025 dollars) | US races |
|---|---|---|---|
| 2000 | $26,774 | $482,543 | 55,486 |
| 2001 | $26,539 | $481,406 | 55,128 |
| 2002 | $26,954 | $496,362 | 54,304 |
| 2003 | $26,560 | $496,454 | 53,500 |
| 2004 | $25,733 | $480,143 | 53,595 |
| 2005 | $24,003 | $459,329 | 52,257 |
| 2006 | $23,611 | $456,970 | 51,668 |
| 2007 | $22,864 | $445,652 | 51,304 |
| 2008 | $20,429 | $407,598 | 50,120 |
| 2009 | $18,480 | $374,340 | 49,368 |
| 2010 | $16,859 | $363,511 | 46,379 |
| 2011 | $15,415 | $339,400 | 45,417 |
| 2012 | $15,259 | $338,443 | 45,086 |
| 2013 | $15,032 | $348,451 | 43,139 |
| 2014 | $14,350 | $347,658 | 41,276 |
| 2015 | $14,500 | $372,358 | 38,941 |
| 2016 | $14,412 | $376,447 | 38,284 |
| 2017 | $14,328 | $380,780 | 37,628 |
| 2018 | $14,445 | $394,833 | 36,586 |
| 2019 | $13,899 | $383,866 | 36,207 |
| 2020 | $13,596 | $490,836 | 27,700 |
| 2021 | $14,514 | $432,389 | 33,567 |
| 2022 | $13,315 | $398,043 | 33,452 |
| 2023 | $12,313 | $387,873 | 31,746 |
| 2024 | $11,561 | $374,738 | 30,852 |
| 2025 | $11,037 | $375,395 | 29,401 |
A shortage of horses limits the number of races we can offer. I still read the diagnosis the other way around. Breeders breed for buyers, and the buyers are leaving: Joe Appelbaum showed at the same Round Table that ownership is down 46% over 20 years. Bettors keep betting the races we card, but wagering per race has only held its own after inflation, and most of the Derby audience never comes back. My read is that demand deserves at least as much attention as supply. A sport that cannot keep the viewers it already has will not be saved by carding more races. She may be right and I may be wrong, which is why we should agree on the measures and let them settle it.
Organizations tend to innovate within the boundaries of their existing responsibilities and incentives. The Jockey Club starts with registration and racing participation because that is what it knows best, just as tracks watch wagering and attendance, sales companies watch the auction market and regulators watch safety. None has the entire picture.
We already know much of what has happened to the industry’s breeding and racing populations. The harder questions are why owners and horseplayers leave, how the economics of breeding, selling and owning horses are changing, how dependent purses have become on gaming and other support, which racing markets are succeeding and why, and whether the initiatives we have paid for are producing results. Answering them takes decision-support information: the costs, objectives and results of what we try, set next to the trends they were meant to change.
I would love to see The Jockey Club’s work become a catalyst for something broader: one shared, publicly accessible scorecard that racetracks, wagering companies, horsemen, breeders, sales companies, regulators and independent builders all contribute to, all can use and all can measure progress against. It has to be credible to people who disagree about what is wrong with the business, because they are the ones who will use it.
The Numbers That Matter Most Have No Common Public Source
The hottest arguments in racing run on estimates. Computer-assisted wagering is the clearest case. The best public figure for the CAW share of US handle is a range of trade-press estimates, 20% to 30%. NYRA showed the number exists when it reported that CAW fell from 20% to 22% of its handle to 12% to 13% after its February 2026 rule changes. There is no figure for the industry as a whole.
The same is true of the questions underneath the decline. How many people bet on racing each year, and how many are new? How many horses leave racing, and where do they go? How much of each state’s purse money comes from outside the betting windows? Owner turnover, everyday attendance and the number of top mares exported each year are in the same position. Each of those numbers sits in somebody’s system: an ADW’s account records, a track’s gate count, a registry, a state commission’s licensing file. Few of them reach a public table, and none in a form that is consistent across the industry.
When we laid out every metric a complete scorecard would need in our working specification, 50 in all, 14 were already on our prototype and 16 more are public and only need building. The other 20 exist only in part, as an estimate or not at all, and 11 of those sit in three places: horseplayers, money and labor, and audience.
Figure 3 · What a full scorecard needs
Twenty of the 50 measures depend on organizations sharing data, and 11 of those sit in three modules
Show the data
| Module | On our prototype | Public, not yet built | Needs cooperation | Total |
|---|---|---|---|---|
| Horseplayers | 0 | 0 | 4 | 4 |
| Money and labor | 1 | 0 | 4 | 5 |
| Audience | 0 | 1 | 3 | 4 |
| Breeding and supply | 4 | 1 | 2 | 7 |
| Safety and welfare | 2 | 2 | 2 | 6 |
| Wagering | 2 | 2 | 2 | 6 |
| Market | 3 | 1 | 1 | 5 |
| Racing | 2 | 2 | 1 | 5 |
| Competing gambling | 0 | 3 | 1 | 4 |
| Venues | 0 | 2 | 0 | 2 |
| Governance | 0 | 2 | 0 | 2 |
| All modules | 14 | 16 | 20 | 50 |
A Shared Scorecard Would Cover the Business From the Breeding Shed to the Television Audience
The scorecard I am proposing has eleven modules. Each answers a question someone in the industry already asks, and each names the organizations that hold the data today.
| Module | The question it answers | Who holds the data |
|---|---|---|
| Breeding and supply | How big is the crop, how concentrated is the breeding, and how many of the best mares are exported? | The Jockey Club, stallion farms |
| Market | How deep is demand below the top of each sale? | Keeneland, Fasig-Tipton, OBS |
| Racing | How much racing is there, and how good is the everyday card? | Equibase, state commissions |
| Safety and welfare | Is the sport getting safer for horses and riders, and what happens to horses after racing? | Equine Injury Database, HISA, aftercare groups |
| Wagering | Where does the betting dollar come from, and what does the bet cost? | Equibase, tracks, tote companies |
| Horseplayers | How many people bet, how many are new, and who pays full price? | ADW operators, wagering hubs |
| Competing gambling | How does racing compare with sportsbooks, casinos and historical horse racing? | American Gaming Association, state regulators |
| Audience | Who watches and who attends, on the big days and on a Thursday? | Networks, streamers, tracks |
| Money and labor | Where do purses come from, how dependent are they on gaming, and are owners staying in the sport? | State commissions, horsemen’s associations, The Jockey Club |
| Venues | Where is racing concentrating, and who controls the tracks? | Equibase, state commissions |
| Governance | Which industry initiatives achieved their stated goals, at what cost and with what measurable impact? What does oversight cost, and where do the rules apply? | HISA, state commissions |
A few rules would keep it credible enough that people on opposite sides of an argument cite the same chart: every number carries its source, its population and its date, each measure stands on its own with no weighted index, and every published forecast gets scored against what happened.
Where It Could Start
Much of the information already exists within the industry’s organizations. Bringing it together would require cooperation, common definitions and a willingness to share. Privacy and commercially sensitive information can be protected. The greater challenge is getting the organizations involved to agree that an independent, shared view of the industry’s health is worth pursuing.
The British Horseracing Authority has published targets for betting turnover, attendance, horses in training and field competitiveness since 2023, and reports against them. Racing Australia publishes a national fact book with wagering turnover state by state. Here, the Thoroughbred Data Hub and its Innovation Incubator, Equibase’s free data releases and NYRA’s publication of its own CAW share all point toward more openness.
It could start small, with one track, one wagering provider and one state racing authority contributing to a shared pilot. A first version that proves its value will attract the next contributors, and the scope can grow from there.
A common scorecard gives everyone the same facts, and a basis for comparing results and holding each initiative to the goals it set. The structural problems still need people with the authority to fix them.
If we can agree on what to measure, make the information available and hold ourselves accountable for the results, we will have a better foundation for deciding what to do next. Every year without a shared scorecard is another year of the same arguments with the same missing numbers. If you think a metric is missing, or you hold data that belongs in the picture, you can reach me on X at @TB_Intel or through the form at tbintel.com.
About the analysis. Figures come from our Thoroughbred Industry Indicators page (The Jockey Club Fact Book, Equibase, the Equine Injury Database and HISA) unless noted. Other sources: NBC Sports and Churchill Downs Incorporated for the 2025 and 2026 Derby audiences and Derby Week handle; the American Gaming Association for sports betting; the 2011 Round Table exhibits for McKinsey’s forecast; the transcript of Charlotte Clement de Bona’s remarks at the August 6, 2026 Round Table; Joe Appelbaum’s ownership figures from the same Round Table, as reported by the Thoroughbred Daily News; Everett Dobson’s June 3 and October 23, 2026 letters; the Fact Book as reported by the Thoroughbred Daily News for race days and California; BloodHorse, from The Jockey Club’s records, for the mid-1950s comparison; the Associated Press for the NTRA track count; and Daily Racing Form and TrueNicks for FanDuel TV. Dollar figures are nominal unless marked as after inflation, which means 2025 dollars on the CPI-U annual average (2025 prices are 35.8% above 2015’s). Real handle per race is US handle divided by US races. Racing handle includes worldwide commingled wagering on US races; sportsbook handle is legal US sportsbook wagering. Japanese buyers’ share at Fasig-Tipton November is our count of mares in foal, racing and broodmare prospects sold, post-sales included, RNAs excluded, with a buyer counted as Japanese only when the buyer of record or named principal is Japan-based. Fatality rates for 2024 and 2025 are provisional. The CAW range is a trade-press estimate; NYRA’s figures are its own. Data Hub terms are from hub.jockeyclub.com as published on November 18, 2026. The 50 metrics come from our working specification for the scorecard.
Corrections. Every figure in this piece was checked against its source, and an error is still possible. If you find one, let me know through our contact form or @TB_Intel on X, and I’ll correct it quickly.