Can a player move down without finishing below their range?

Yes. Of the 185 player-seasons that moved down, 48.6% finished at or above the bottom of their rating. Finishing low inside it was followed by a move down 4.8% of the time when the rating was computer-issued, and 17.0% when it was self-rated.

Why are computer-rated and self-rated players counted separately?

Because the two groups move at different rates from the same place. One combined figure would be wrong for both of them. A self-rated player who finished low inside the range moved down 17.0% of the time. A computer-rated player, in the same place, 4.8%. Some of that difference is a self-rating being corrected.

Tennis Desk Research · Marin & San Francisco USTA NorCal · Adult league

At the top of the range 17.0% of players moved up, and three tenths past it 76.0%

Q: Where do you have to finish to get bumped?

There is no score to reach. Players who finished above the rating they were given held a higher rating the next year 32.4% of the time. Players who finished high inside it, 3.1%. And 13.4% of every move up came from a player who never finished above their rating.

Q: Is there a minimum number of matches before a player can move?

Nothing here shows one. The Desk holds the lines played in two areas, not a player's whole USTA record, so a player seen twice here may have played a full season somewhere else. The ten-line rule used here is about how much of a season the Desk saw, not about how much USTA asks for.

By Thomas Chung

The further a player's Desk level finished past the top of their rating, the likelier they were to move up. Among 4,541 seasons of computer-rated players in Marin and San Francisco, the share who moved up rose with every tenth of a point:

Where the level finished	For a 3.5	Moved up
A tenth under the top	3.40 to 3.49	5.3%
At the top, up to a tenth past it	3.50 to 3.59	17.0%
Three tenths past it	3.80 to 3.89	76.0%

The rates pool every rating. For a 4.0, add half a point to the middle column.

A note on the outcome

A move here means a different published NTRP rating the next year, whatever the reason: a promotion, a corrected self-rating, an appeal or a reclassification. So no figure here is a forecast for any one player.

A rating names a range, and its top is the number on the roster

An NTRP rating covers a range of playing strength, and the number on the roster is its top, so a 3.5 covers every level above 3.00 up to 3.50. The Desk works out each player's Desk level on the same scale, to two decimals, which is why a tenth of a point can be measured at all.

A level against the top of its rating

moved up the next year ↑

2.2%

5.3%

17.0%

30.1%

53.9%

76.0%

−0.3↓

−0.3

−0.2

−0.1

+0.1

+0.2

+0.3

+0.4↑

10.6%

0.9%

moved down ↓

under the ratingover it

how many players finished at each tenth

Share of 4,541 computer-rated player-seasons that held a higher rating the next year, by how far the Desk level finished from the top of the range the published rating names. 0 is the top of the rating: an NTRP 4.0 covers everything above 3.50 up to 4.00, so 0 is the 4.00 edge and every step right of it is a tenth of a point past that. Marin-area and San Francisco USTA NorCal adult league, 2022–2025. The two end steps take in every level further under or over. A dot marks a step with too few players or too few moves to state a rate. Measured across every rating together, and for computer-rated players only, because at a tenth of a point no single rating carries enough players to publish a rate at every step, and at the two steps furthest past the top there are too few self-rated players or moves to state a rate.

The average past the top

Taken as one group, computer-rated players who finished past the top of their range moved up 32.4% of the time, over 1,301 seasons.

Players who moved without leaving their range

Some players moved up without passing the top of their rating: 13.4% of the 605 seasons that ended in a move up. And 48.6% of the 185 moves down came from players who never finished below their range.

Computer-rated players high inside the range moved up only 3.1% of the time, over 1,964 seasons. But 64.5% of the 5,286 seasons counted ended inside the range.

Moving down is a different shape

The further under the top a player finished, the likelier a move down. Self-rated players moved down far more often. Low inside the range, 17.0% of 159 self-rated seasons did, against 4.8% of 1,075 computer-rated ones. Below it, 46.0% of 87 self-rated seasons did, against 27.4% of 201 computer-rated ones.

Part of that gap is probably self-ratings being corrected, not worse tennis.

Read next: What happens after a bump?

Where did your own level finish?

Look up any player or team →

All Tennis Desk Research·Browse every league the Desk holds, by year and area

How this was measured

Scope. USTA NorCal adult league for Marin County and San Francisco, seasons 2022–2025, each followed into the next year's published rating. So these are not figures for NorCal or for the whole country.

Who is counted. A season counts only if the same player held a published NTRP rating in both years, because a move cannot be seen otherwise. The Desk must also hold ten or more of that player's lines, a rule about the Desk's coverage rather than about how much anyone played. The ten-line rule is not a USTA minimum, and nothing here shows that one exists. Only two areas are covered, so a player seen twice here may have played a full season somewhere else. These line counts also include Combo, which USTA excludes from year-end ratings entirely. In all, 5,286 player-seasons clear both rules.

Where a player finished. A player's finish is their Desk level after their last recorded line of the season, read against the range their published rating names. Seasons whose rating carried an appeal or a disqualification are left out, because they behave like neither group and every one of their cells is too small to publish. I hope to measure them properly if coverage reaches the rest of NorCal or another section.

Why one curve. No rating alone has enough players for a curve at tenths. Even taken alone, no rating shows a rate at more than five of the nine steps. Self-rated players are fewer again, and the two steps furthest past the top have too few of them, or too few moves, to show a rate. So the curve pools every rating and counts only computer-rated players, which limits what it can say about any one rating or about a self-rated player.

Measured. Every figure above was checked against the live database on October 22.

Method. Every figure above was produced by the SQL recorded in docs/research/records/where-do-you-have-to-finish-to-get-bumped.json, checked before this page was built. A group under fifty player-seasons, or with fewer than five moves, publishes its size instead of a rate. Reproduce with npx tsx scripts/measure-research-record.ts.