Art of analysis: Global broodmare sire evaluation

Ahead of the 2023 Magic Millions National Broodmare Sale, Tom Wilson’s analysis highlights the best-performing broodmare sires, peaking into the catalogue to see which opportunities buyers can avail over the next three days.

Using statistics to evaluate breeding performance

A common practice across the racing and bloodstock world is to evaluate performance vs. total numbers of runners. One distinction that we’ll bring in here is a preference to evaluate statistics relating to breeding against named foals, rather than the conventional ‘to-runners’ metrics.

Across the dataset of the top 50 broodmare sires with lots in the catalogue this week, the average returns of Group 1-winners-to-foals is 0.54 per cent, Group-winners-to-foals is 2.04 per cent and stakes-winners-to-foals is 3.80 per cent. These will be taken as sample baselines, from which we compare over and under performance.

The top performers

We’ll start with the overall performance as a broodmare sire vs. total number of named foals in our dataset. On raw numbers, the great Galileo (Ire) was top-ranked on stakes-winners-to-runners, with 284 stakes winners globally from 4935 named foals as a broodmare sire - a return of 5.75 per cent. For buyers looking to maximise their chance of success, Galileo is the obvious starting point.

However, he is the sire of just one mare in the catalogue, with Eclipse Plus offering Lot 926, Hermione. The dam of three winners from four to race, the 18-year-old is offered with a pregnancy to Santos.

The broodmare sire of 13 Group 1 winners in Australia and New Zealand, one of those is also up for sale on Tuesday as Lot 645, the brilliant Snapdancer (Choisir).

Sample sizes present something of a challenge in racing analysis. A particular technique is that we like to employ is to normalise samples per 1000 foals. This gives a more reliable view and a chance to compare like with like; here Galileo again topped our performance table.

In the table below we’ve adjusted to a rate of Group winners and stakes winners as expected per 1000 runners. We call these metrics 'expected Group 1 winners', ‘expected Group winners’ and 'expected stakes winners', or simply, ‘xG1W’, ‘xGW’ and ‘xSW’.

To read how and why this makes for a more useful analysis, please see the appendix below.

As with Galileo, buyers are offered just one opportunity to buy a daughter of the elite second-ranked broodmare sire, O’Reilly (NZ), with Lot 1340 part of Edinburgh Park’s Unreserved Dispersal Sale. Twice a Listed winner and Group 3 placed, she was served last year by Glenfiddich. O’Reilly placed second on our xSW per 1000 foals, with a return of 58.

Who's available?

Removing the withdrawn lots (as of Monday morning), the most-represented sire in this year’s catalogue is Snitzel, with 29 race fillies or broodmares. As a broodmare sire, Snitzel placed 32nd out of our top 50 at the sale, with 28 expected stakes winners per 1000 foals.

Again, towards the top of the table, daughters of Zabeel (NZ) are of limited availability over the next few days. However, one of them is particularly notable; alongside Lot 1048 from Vinery, the dam of three winners including Group 2 placegetter Commander Harry (NZ) (Reliable Man {GB}), Zabeel is represented in Lot 751.

Consigned by Three Bridges Thoroughbreds, she is more famous as Zenaida (NZ), the dam of G1 Surround S. winner Sunshine In Paris (Invader) - herself heading through as Lot 700.

Fifth in the xSW ranking, five mares by Shamardal (USA) will go under the hammer this week, whilst Street Cry (Ire) provides three opportunities to access the upper end of the table. He’s the sire of: Lot 832, Widden Stud’s three-time winner who sells in foal to Russian Revolution; Lot 844 from Milburn Creek, selling in foal to St Mark’s Basilica (Fr); and Lot 869, a Listed-placed three-time winner from Godolphin who was served last year by Impending.

Far more opportunities lie with Darley’s now-pensioned Lonhro. He’s the sire of 12 in the catalogue, eight of those within Godolphin’s consignment, which includes Lot 830, a half-sister to dual Group 1 winner Cosmic Endeavour (Northern Meteor), and Lot 1130, the half-sister to Pierro served last year by Astern.

Also inside the top 10, there are ample opportunities for buyers to secure daughters of Redoute’s Choice (who has 15 in the sale).

Buyers are spoilt for choice with 19 daughters of Medaglia D’Oro (USA) in the sale, the Darley America resident sitting seventh as a broodmare sire by xSW with 47 expected stakes winners per 1000 foals. It’s interesting to note how high Iffraaj (GB) falls, also from limited runners. Another outcross influence for Australasian breeders, the son of Zafonic (USA) is 18th by xSW and has two in the sale: Lots 1171 and 1334.

Another American influence with plenty of representation is More Than Ready (USA). Thirteenth by xSW, we will see 13 mares by him through the ring this week.

His compatriot Hussonet (USA) also deserves a mention; his Group 1 winners-to-foals percentage (1.01) puts him up there with the best, and his mares have made waves, including the dams of Shoals (Fastnet Rock), Extreme Choice, King’s Legacy and Rothfire (Rothesay). Buyers will likely need deep pockets, however, with Hussonet’s only progeny in the sale being Hussy By Choice (Lot 932) from Glastonbury Farms, out of Little Flower (Redoute’s Choice), a half-sister to Castelvecchio.

Closer to home, and with the largest representation amongst the top 20, Coolmore Australia’s former shuttler Fastnet Rock has 22 mares in the sale. The broodmare sire of seven elite-level winners, all bar two of those came in Australia, most recently in the shape of Chris Waller’s G1 Flight S. winner Zougotcha (Zoustar).

Plenty of opportunities, then, for breeders to turn the statistics in their favour. Whether breeding to race or breeding to sell, a stakes winner is invaluable - and the numbers don’t lie.

Table: Broodmare sires ranked by xSW

Appendix - Using statistics per 1000 runners to evaluate broodmare sires

Managing and understanding variations in sample sizes is one of the more important aspects in meaningful statistical analysis. A jockey who goes three wins from five in their first runs at a track isn't going to convert that over a larger sample to 60 wins out of 100. The law of averages, or regression to the mean, always comes to bite in the end.

In racing and bloodstock statistics we observe a similar phenomenon. We get excited by the first-season sire who has winners from two of his first three runners, yet sadly that doesn't mean that 60 per cent of all progeny that make the racetrack are going to be winners. Be aware of making inferences from small samples.

In order to counter these limitations, we normalise samples to a consistent sample size. This gives a solid basis from which to do reliable analysis - it's ensuring you compare apples with apples.

When we evaluate performance metrics across bloodstock, we typically adjust to a standard figure per 100 or per 1000 runners. We approach this in two ways based on the samples available; for sires with >1000 runners in the sample we adjusted back to their performance level achieved per 1000, using: ((Current Stakes Winners / Current Sample)*1000).

For sires yet to achieve a 1000-runner sample, we take their current sample and then extrapolate out to 1000 runners, using the population average as an expected performance level for the remaining level of runners. The further away a sample is from that 1000-runner threshold, the more likely that the law of averages, or regression to the mean, will impact the overall performance returns.

That's why we leverage the population average to represent future expectations. Across our dataset, the average return of stakes winners to runners per broodmare sire was 2.05 per cent stakes-winners-to-runners.