What Is a Good Dressage Score? How to Actually Read Your Percentage

You ride down centerline, salute, and forty minutes later a number appears next to your name: 64.231%. Whether that number is good depends on what it actually measures, what level you rode, who was sitting at C, and what you were trying to accomplish that day. Most riders learn to read their percentage the way most people read a thermometer — higher is better — and stop there. There is a lot more information in a scoresheet than that.

What the percentage literally is

A dressage test is a fixed sequence of movements — the scored units on the sheet, each with a directive telling the judge what to evaluate. The judge marks every movement from 0 to 10, in half-point increments. Some movements carry a coefficient, a ×2 multiplier applied to work the test writers consider diagnostic for the level. After the final halt, the judge awards collective marks — holistic scores for the ride as a whole.

Your percentage is simply: all those marks, multiplied by their coefficients, added up, divided by the maximum possible points.

Take a concrete example. The current USEF Training Level Test 1 (the 2023 test cycle, in effect through November 30, 2026) has 13 scored movements, six of which carry a ×2 coefficient, plus five collective marks: Gaits (×1), Impulsion (×2), Submission (×2), Rider's Position and Seat (×1), and Rider's Correct and Effective Use of Aids (×1). Maximum: 260 points. Score 165 points and you rode a 63.462%.

Errors of course come off the total, not off a single movement — under the current USEF national tests, 2 points for the first error, 4 for the second, elimination for the third. That is why one wrong turn can quietly cost you nearly a full percentage point on a Training Level sheet.

The important consequence: your percentage is an average of judgments, not a single judgment. A 65% ride is not a ride the judge thought was "65% good." It is a ride whose movements averaged, roughly, 6.5 out of 10.

What the 0–10 scale actually says

The scale itself has fixed semantics, shared by USEF and FEI judging:

Read your percentage back through this table and it becomes interpretable. A 60% ride averaged 6s: the work was satisfactory — the movements happened, correctly enough, without much to criticize or much to praise. A 70% ride averaged 7s: fairly good, consistently. An 80% ride averaged 8s, which is why scores in that range are rare and newsworthy even at the top of the sport.

Notice the cliff between 5 ("sufficient") and 4 ("insufficient"). That single point is the line between "this movement met the standard" and "it did not." A test full of 5s — a 50% ride — is a test that barely cleared the bar on every movement. More scale terminology is in the glossary.

The bands, and why 60% is the benchmark

Here is the honest framing: the interpretation bands riders use are community convention reinforced by qualification thresholds, not physics. Nothing in the arithmetic makes 60% special. Two things make it special in practice.

First, the semantics above. Averaging "satisfactory" across an entire test is the traditional definition of confirmed work — the horse and rider can reliably perform everything the level asks. That is why "scoring 60s" is the conventional signal that a combination is established at a level and can start schooling the next one.

Second, the sport's gatekeeping is built around scores in this neighborhood. USDF Regional Championships require a set number of qualifying scores, with minimum percentages that vary by level and by division (open, adult amateur, junior/young rider). FEI-side, championship eligibility runs through MERs — Minimum Eligibility Requirements — which are likewise minimum-score rules. The exact percentages change by year, level, and division, so don't take a number from an article (including this one); take it from the current USDF and USEF published requirements. But the structure is stable: the sport formally defines "good enough to advance" as a minimum percentage, and those minimums cluster in the low-to-mid 60s territory that convention already treats as confirmed work.

So the working bands most of the American dressage community carries around look like this:

Treat that table as a dialect, not a law. It is how scores are read, and it is calibrated by the qualification thresholds — which is exactly why it holds up reasonably well in practice.

The bands, measured

Those bands are convention. Here is what they look like against data. The figures below come from the DPI index — a database of US national (USEF/USDF) results, restricted to scores on the standard 0–100 scale (freestyles, young-horse, and equitation classes, whose marks aren't comparable, are excluded). As of mid-2026 it holds about 170,000 comparable scored rides, and the overall median across all of them is 63.95%. By level:

Three things fall out of that table, and each sharpens the convention rather than overturning it.

The low-60s is the real center of gravity. Every level's median sits in the low-to-mid 60s — the whole sport clusters right where convention puts "confirmed." A 63% isn't a mediocre score fishing for sympathy; it is, almost exactly, the middle of the sport.

70% is genuinely rare. At most levels only about one ride in twenty breaks 70%; even at the levels where more do — Training, First, Grand Prix — it's roughly one in eight. When the convention calls 70%+ "exceptional," the data agrees: it is the top few percent of rides, at every level.

The middle is the hardest place to score. Notice the sag — medians are highest at Training and First, dip through Second, Third, and Fourth, then rise again at Grand Prix. That is not because Grand Prix is easier; it's survivorship. The middle levels are where combinations are learning collection and getting marked hard for the gaps, while the riders who reach Grand Prix are the ones who closed those gaps — so the pool at the top is pre-filtered for quality. Read honestly, the toughest percentage to earn is a strong one at Third or Fourth Level, not at the top of the sport.

None of this is a law of nature either — it is one index's coverage, US national only, and it will shift as more results land. But it is measured, with every score tied to the competition it came from. That is the difference between a benchmark and a vibe.

The same number is not the same number

A percentage only carries meaning inside its context. Three contexts matter most.

Level. A 65% at Training Level and a 65% at Prix St. Georges are both "fairly good on average," but the standard being judged against is vastly different. The FEI levels — Prix St. Georges and Intermediate I (the Small Tour) up through the Big Tour of Grand Prix, Grand Prix Special, and the Kür — demand collection, and the whole distribution of scores compresses and shifts as difficulty rises. Comparing your Third Level percentage to a friend's First Level percentage tells you almost nothing. This is why any credible ranking has to compare within level, not across it — see how DPI handles this on the national leaderboard and in our methodology.

Class size and company. A 66% that wins a class of four is a different fact than a 66% that places eighth of thirty at a championship qualifier. Placing is relative; the percentage is the closest thing dressage has to an absolute measure — but "closest thing" is doing real work in that sentence, because of the third context.

The judge. Judge-to-judge variance is real and normal. Judges are humans applying a standard, and standards get applied with different severity. Some sit systematically lower or higher; the same ride can plausibly score two or three percentage points apart under two competent, licensed judges. This is not scandal — it is why championship classes use a Ground Jury of multiple judges and average their scores, and why USDF's qualifying structures typically require scores from more than one judge. For your own reading: a score earned under a judge known for low marks (your scribe friends will tell you) means more than the same score from a generous one, and no single-judge score should be over-interpreted in either direction.

Read past the headline number

The percentage is the least informative line on your scoresheet. If you want to actually improve, read the sheet bottom-up.

Movement-level scores. Where are your 5s? A test that averages 6.4 with marks ranging 6–7 is a different training picture than one that averages 6.4 with 8s on the trot work and 4s on every canter transition. The first horse needs overall polish; the second has a specific hole.

Coefficient movements. The test writers already told you what matters: the ×2 movements are the diagnostic ones for the level. A 5 on a coefficient movement costs you double, which is the test's way of saying this is the thing to fix first.

Collective marks. Gaits, Impulsion, Submission, and the rider scores are the judge's summary of the whole ride, and they map loosely onto the Training Scale — rhythm, suppleness, contact, impulsion, straightness, collection. A submission collective that consistently trails your gaits collective is a judge telling you, across many movements, that the horse's quality exceeds its current rideability. That is a training direction, delivered for the price of your entry fee.

The trend, not the test. Any single test is one judge, one day, one warm-up. Five tests are a signal. The question worth asking is not "was 63% good on Saturday?" but "is my median moving?" A combination scoring 61, 62, 64, 63, 65 across a season is confirmed and climbing, whatever any individual Saturday said. Rider score histories on DPI rider pages are built around exactly this view — the trajectory, with each score's source attached.

When a "low" score is progress

Some of the best scores on a rider's record are the ugly ones.

Move up a level and your percentage should usually drop — often meaningfully. A 58% in your first Fourth Level test after a season of 66% at Third is not regression; it is the expected cost of being judged against a harder standard, and it is the score of a rider actually developing. The same logic applies to a young horse's first season out, a horse returning from time off, or a first test under a notably demanding judge. In all of these, the honest comparison is against your own baseline in the new context, not against your peak in the old one.

The failure mode to avoid is the opposite one: camping at a level to farm 68s. The percentage exists to serve the training, not the other way around.

A good dressage score, then, is a percentage read in full context — level, division, judging, and trend — against the fixed meaning of the 0–10 scale. If you want your own scores assembled that way, with provenance attached, DPI's index and qualification planner do exactly that.