EACH year, hundreds of millions of euros, pounds and guineas are spent at public auctions in Britain and Ireland for the purpose of acquiring thoroughbred yearlings.
Premier auctions, such as Tattersalls October Book 1 and Goffs Orby Book 1, offer for sale horses with fashionable pedigrees and strong veterinary assessments.
These are the sales where the most desired bloodlines are auctioned, and which attract bids whose scale bears along the prestige and esteem conferred through the belief that future champions are to be sourced there. Buyers direct their resources towards these auctions because the horses on offer are closely related to past champions and, as their related bloodlines indicate, are presumed to reproduce that success on the racecourse.
In practice, however, this expectation frequently proves misplaced. Many of the most expensive purchases fail to deliver elite performances.
The data shows that higher-priced horses, in the main, tend to perform better, though the price-performance gradient flattens at the top.
Strongest bloodlines
Premier auctions, such as Tattersalls Book 1, attract the strongest bloodlines who are therefore bought for the highest prices, as evidenced by the average yearling price in 2022 that exceeded £270,000. The graduates from this sale are shown to be among the highest-rated horses, on average, in the sample.
Of all the yearlings sold at public auction that year, only 6.8% were rated high enough to compete at the elite blacktype level.
The figure, in fact, understates the shortfall: many costly purchases are exported to race abroad, while others never reach the track at all. These outcomes form a major theme of the industry.
A central empirical regularity, therefore, emerges: while auction prices respond strongly to observable signals of quality, they also tend to be weak predictors of outcomes, and substantial variation persists among horses purchased for similar prices.
In bloodstock markets, the most important post-sale input is training. Trainer quality, stable resources, race-placement decisions, along with a host of other relevant factors, are neither fully observable nor fully contractible at the time of auction, yet plausibly exert first-order effects on realised racing performance.
If these inputs are not fully capitalised into prices, then auction prices should be informative but insufficient predictors of performance, and systematic differences in outcomes may therefore persist among horses purchased for similar prices, depending on post-sale allocation decisions. This article evaluates these implications directly.
The results suggest auction prices are informative but incomplete predictors of performance: doubling price is associated with an increase of roughly 3-4 rating points once controls are included, and the unconditional relationship is concave, consistent with diminishing marginal returns at the top end.
Trainer quality
Second, trainer quality has a large and economically meaningful effect. Horses trained by top-tier operations are rated about 16 points higher than those trained elsewhere; instrumenting for trainer assignment reduces the estimate to roughly 10 points, implying that selective placement explains part, but not most, of the observed gap.
Subsample analyses show that the estimated premium is not driven by a few exceptional horses. Excluding horses rated above 100, for example, leaves the premium mostly unchanged. Sex-specific estimates show the premium is concentrated among colts, consistent with industry allocation practices and the structure of the instrument. When horses by high-fee sires are excluded, the similar point estimates confirm that the results are not confined to the elite stud-fee segment.
Finally, comparisons across price bands highlight the interaction between trainer quality and expenditure.
Horses purchased for less than £200,000 but sent to top-tier trainers tend, on average, to outperform horses purchased for more than £350,000 and trained by unranked operations by 9-10 rating points.
In plain terms, the horses purchased for lower sums, sent to top-tier trainers, are rated higher than those purchased for more than £350,000, and placed with unranked trainers.
Sales uncertainty
Yearlings are traded under substantial uncertainty, buyers observe rich but noisy signals such as pedigree, in-person inspections, veterinary assessments, and post-sale outcomes depend on training choices and organisational capability (i.e. training operations and personnel).
There is a widespread expectation among market participants that the most promising stock of future racehorses are most likely to be found at a small number of premier yearling auctions, most notably Tattersalls October Book 1 and the Goffs Orby Sale.
In 2022, the yearlings purchased at Tattersalls Book 1 achieved the highest average official ratings, at approximately 79, followed by the Craven Breeze Up at 76, while Orby and Tattersalls Book 2 horses averaged around 73, compared with ratings near 61 at lower-tier sales. Auction prices follow a similarly steep hierarchy, exceeding 270,000gns on average at Book 1, compared with roughly €131,000 at Orby, 112,000gns at the Craven Breeze Up, and under €11,000 at the bottom of the market. Direct evidence on ex-post performance and returns in Britain and Ireland is provided by Butler and Butler (2021), who evaluate auction purchases against earnings and show that most racehorses generate negative net financial returns, with losses becoming more severe as auction prices rise.
This opens up a natural role for post-sale choices formed in this article - in particular, the allocation of horses to trainers - to act as an additional margin through which performance is produced, and it motivates an empirical design that treats price as a pre-treatment signal and therefore estimates the marginal contribution of trainer assignment.
Translated into horseracing terms, trainers can be viewed as managers of biological assets - outcomes are partly endogenous to training and placement decisions, stable resources, jockey selection, and the broader organisational structure. The empirical challenge, however, is that high-endowment horses may be selectively allocated to high-quality trainers. A credible interpretation of a ‘trainer premium’ therefore requires separating sorting on latent ability from genuine value added.
If training is productive - rather than merely a label attached to superior horses - then one should observe persistent trainer effects after conditioning on pre-treatment indicators of expected ability, and ideally after addressing non-random assignment.
This study draws on a newly-constructed dataset that matches yearling auction prices to their racecourse performance for thoroughbreds sold at public sales auctions in Britain and Ireland. The sample consists of all yearlings that were sold in 2022 across major auction houses that went on to race in Britain and Ireland.
The design is structured such that a complete cohort of horses is allocated sufficient time to realise peak ratings, while avoiding the survivor biases inherent in using older sales cohorts.
Datasets used
The dataset combines three main variables of interest. Firstly, auction data were collected from Tattersalls (UK and Ireland) and Goffs (Ireland), covering all yearling sales in 2022.
Secondly, official handicap ratings were collected from Horse Racing Ireland (HRI) and the British Horseracing Authority (BHA), which provide data on the most up-to-date published ratings as of May 2025.
Finally, trainers are classified into tiers based on their international performance ranking. Specifically, trainers are ranked according to the number of blacktype winners they have trained. The top-50 trainers are designated as top-tier, while all others are categorised as unranked.
Among horses acquired for under £50,000, there is a large concentration in the lower performance bands: some 46% fall between 50 and 69, while fewer than 3% reach a rating of 100 or higher.
Moving up the price ladder, namely, the £50,000-£100,000 bracket, the lower share declines to 30.6%, accompanied by a rise in the 70–89 range; this pattern strengthens further among horses bought for £100,000 to £250,000. In the top expenditure group - those sold for more than £350,000 - only 13% comprise the lowest band. Larger proportions are found in the intermediate ranges: 28% in 70-79, 24% in 80-89, and 15% in 90-99.
Almost 15% exceed the 100-rating threshold, roughly six times the share observed among the least expensive purchases.
Among horses trained outside the top-tier, the bulk cluster at the lower end of the scale: 42% comprise the 50-69 band, 23.3% fall between 70 and 79, and thereafter the proportions taper off - 22.3% in 80-89, 11.8% in 90-99, 2% in 100-109, and a mere 0.3% in the 110-120 range.
The pattern among top-tier trainers is altogether different. Their distribution is spread more evenly across the spectrum, with a relatively smaller share (13.9%) in the lowest band and a larger presence in the middle and upper ranges: 19.8% in 70-79, 21.2% in both 80-89 and 90-99, 15.8% received ratings between 100 and 109, and 5.5% between 110-120 band.
Whether this characteristic is expressed vis-á-vis top-tier trainers’ operations, possessive of greater capital infrastructure, or through having access directly to elite jockeys, requires analysis beyond the scope of this article.
But these illustrations clearly point to a deep concentration of high-performing horses within the top echelon of the training hierarchy.
Overview
The central question is whether assignment to a top-tier trainer causally enhances a racehorse’s performance, conditional on indicators of quality observed prior to training. Trainer assignment is not random. Owners may place horses with particular trainers based on information unobserved, but correlated with both price and performance. A simple comparison of outcomes would therefore conflate trainer effects with selection on unobservables.
My research makes clear that less than 20 points separate average ratings between the most and least expensive horses.
At each price point, horses trained by top-tier trainers are rated higher, ranging from around 10 to 20 points.
Horses that are placed with top-tier trainers receive ratings about 16 points higher than those trained by unranked trainers, and this gap remains broadly unchanged as controls are added. Fillies are rated roughly four points below colts, which is consistent with, but slightly above, the usual sex allowance. Geldings show little difference.
Key findings
The main message of the results is straightforward. Horses sent to top-tier trainers are rated higher, and this difference persists after allowing for price and other characteristics. Price remains a meaningful predictor of performance, though part of its effect passes through trainer and sire quality.
In some instances, the advantage of elite trainers may consist merely in extracting marginal gains from already expensive horses. This, for example, may be the case at the very top of the market, where highly-priced horses are clustered in elite stables, and differences relative to expensive horses in unranked operations are small.
The results show that horses purchased for less than £100,000 and trained by top-tier trainers do not, on average, outperform the most expensive horses trained elsewhere. Horses acquired for less than £150,000 but trained by top-tier trainers exceed expensive horses in unranked stables by roughly seven rating points.
As the threshold rises to £200,000 and £250,000, the gap widens to nine and 10 points. These are large magnitudes in the context of official ratings and correspond closely to the trainer premium estimated through instrumental variables. Thus, trainer quality is an important productive asset of the racing industry, operating alongside price, pedigree, and other structural factors.
The allocation of horses to leading trainers at moderate price levels brings about marked gains in performance.
The evidence plainly points out that high expenditure, by itself, fails to carry the conditions requisite for success. Large sums spent on horses that are placed with unranked trainers do not, on average, guarantee better performance. Conversely, access to top-tier training confers advantages well beyond what the upper reaches of the market alone can bear.
Even at moderate price levels, assignment to a leading trainer raises performance significantly.
This finding suggests that the emphasis on purchase price tends to obscure the independent role of training quality in how racing outcomes are determined.
Conclusion and discussion
The results establish four empirical facts. First, auction prices are ‘informative’, but far from ‘sufficient’ measures of performance. The unconditional price-performance relationship is concave; average ratings differ by less than 20 points between the cheapest and most expensive horses, and fewer than 7% of purchases ever reach elite performance levels, even at the top of the market.
Second, trainer assignment is associated with large, heterogeneous outcomes. Across all price bands, horses trained by top-tier trainers are rated roughly 16 points higher than those trained by unranked trainers.
Third, while part of this premium reflects selection, the statistics imply a causal trainer effect of around 10 rating points.
Fourth, managerial capital operates as a substitute for expenditure: horses purchased at moderate prices and placed with elite trainers routinely match or exceed the performance of the most expensive horses trained elsewhere.
By contrast, purchaser spending bands and country effects add little explanatory power once price, pedigree, and trainer-tier are taken into account, and returns to sire quality flatten at the very top of the stud-fee distribution.
In the broadest sense, the evidence ultimately supports a production-function view of racing performance, in which realised ratings reflect the interaction of an initial endowment (summarised imperfectly by price and pedigree) and post-sale organisational capability (proxied here by trainer tier).
It also yields a general lesson that extends beyond bloodstock. In markets where assets are bought under uncertainty, but outputs are produced through subsequent, heterogeneous management, auction prices can aggregate information efficiently about what is observed at sale without fully internalising the value of complements chosen thereafter.
Conor Banahan joined the Economic and Social Research Institute in Ireland as a researcher in September 2024 until August 2026. He holds a BSc in Business Studies from DCU and an MSc in Economics from Trinity College Dublin. His research focuses on applied microeconomics, including labour markets, housing, energy pricing, and auction markets.
Scan the code to read the full research paper