A past-performance line that reads "trainer/jockey: 4 for 10, 40%" is counting exactly what it says: how many times that specific trainer and that specific jockey have won together, out of how many races they have started together, usually over a recent window such as the current year or meet. Some providers add the in-the-money rate (finishes in the top three) and the flat-bet return on investment, what a $2 bet on every one of those starts would have returned.

The number is real. What it is worth depends entirely on how many starts it is built on, and that is the part a 40% headline tends to hide.

Where the stat comes from and what it measures

Past-performance providers print a trainer-jockey combination line because trainers tend to use the same handful of riders repeatedly, for reasons that have nothing to do with any single race: the jockey rides regularly for that barn, knows how its horses run, and is available on the right days. The stat is simply that pairing's joint record, filtered down from each rider's and each trainer's much larger overall numbers.

It is a record of what already happened, not a forecast. A combination that has won 40% of its starts this meet has not been shown to win 40% of its future starts; it has been shown to have won 40% of a specific, usually small, set of past ones.

Why the sample size matters more than the percentage

Two trainer-jockey combinations, same win rate claim, different reliability

Combination A has started together 10 times this meet. Combination B has started together 300 times over the past three years.

Combination A4 wins in 10 starts = 40%
Combination B60 wins in 300 starts = 20%
Which number moves more on the next single resultA: one more win or loss changes it by 10 points. B: by about a third of one point

Combination A's 40% looks far more impressive printed next to Combination B's 20%, but A's figure is built on so few starts that a single result swings it dramatically, while B's lower rate rests on enough races that it is a far more stable description of how that pair actually performs. A 4-for-10 record is closer to a coin flip's worth of evidence than a proven trend.

The general rule: treat any combination stat under roughly 20 to 30 starts as too small a sample to lean on by itself, and read the raw win-loss count, not just the percentage, before deciding whether a number means anything at all. The return-on-investment figure some providers add, what a flat $2 bet on every one of that combination's starts would have returned, suffers from the same problem: a pairing that hit one long-priced winner in a 10-race sample can show a strong ROI that one different result would erase.

How much weight a combination stat deserves

Starts together What the win rate tells you
Under 20 Mostly noise. One result shifts the percentage by 5 points or more.
20 to 75 A loose signal. Worth noting, not worth betting on by itself.
75 and up A reasonably stable description of how that pairing performs, still just one factor among the horse's own form, class and trip.

How a jockey ends up on a particular horse

A trainer or owner chooses the jockey for a given race, and that choice is usually arranged through the jockey's agent rather than negotiated directly. An agent represents one jockey at a time and spends the days before a race card lining up mounts across several trainers' barns, trying to get their rider on the best available horse in as many races as the card allows. A jockey who rides regularly for one trainer is often simply the rider that trainer's agent relationship, and track habit, has put on that barn's horses race after race, which is the mechanical reason trainer-jockey combination stats exist at all.

What a winning jockey is actually paid

Riding fees are set by state regulation or by agreement, and they vary by state. California's rule is a useful, sourced example of the shape these fees take: absent a different written agreement, the jockey on the winning horse is paid 10% of the owner's win purse, and the jockey on the 2nd or 3rd place horse a minimum of 5% of that finish's purse share, plus a small flat addition. Losing, non-placing mounts are paid a fixed fee set by the race's total purse rather than a percentage.

A jockey's cut under California's rule (hypothetical purse)

Say a stakes race pays its winning owner $100,000, and the same jockey's mount finishes second in an earlier race on the card where the owner's place share is $20,000.

Win mount, 10% of $100,000$10,000
Place mount, 5% of $20,000$1,000, plus the flat addition the rule sets for that purse range

The jockey's fee tracks the purse directly: a win in a big stakes race is worth far more to the rider than a win in a low-level claiming race, which is part of why top jockeys compete hardest for mounts in the biggest races rather than riding every available horse on a card.

Using the stat without overreading it

A strong, well-sampled trainer-jockey combination is one piece of a handicapping read, alongside the horse's own form, class and trip, covered in handicapping horses for beginners, how to handicap class and trip handicapping. It describes a pairing that has worked well together before, not a guarantee that it will again in a single race with a different field, a different distance and a different trip. Read the start count before the percentage, and read the combination as context for a horse's chances, not as the answer on its own.