Anthony Fox

Facing break point, servers barely change their aim

November 2026

Key points

On 1 February 2026 Novak Djokovic served at 5-6 in the fourth set of the Australian Open final, 15-40 down to Carlos Alcaraz, who needed one of the next two points to win the title. Volunteers for the Match Charting Project, who log matches point by point, recorded the point as played in the deuce court, the side on which every game opens. Djokovic's first serve, aimed wide towards the sideline, went into the net. To the returner, break points such as 15-40, one point from taking the server's game, look like the moment for a serve kept in reserve.

Game theory, by contrast, expects no serve to be kept in reserve. A returner soon learns to lean towards whichever serve wins more often, so a server does best to mix his aims until the wide serve and the T win equally often, whatever the score. Two economists, Mark Walker and John Wooders, found that balance among top servers in 2001. Across 5,569 charted men's matches, servers facing 15-40 in the deuce court aim wide almost as often as at 30-30, just as the theory predicts. A returner who waits for break point to read the server therefore learns nothing beyond what 30-30 had already shown.

Do servers aim differently on break point?

Break point changes servers' aim only a little, as game theory expects. In the deuce court the wide share is 44.5% on other points, against 47.2% at break point. Most break points are played in the advantage court, the other side, where the wide share barely moves from 50.5% to 51.1%. So the one change worth explaining is a few more wide serves to the deuce court.

Break-point serves land almost where other first serves land

Men's first serves by aim, deuce and advantage courts, % aimed wide, at the body and down the T

Wide means towards the sideline, body at the returner and down the T along the centre line. One ball per first serve, hit from the left baseline; depth is not charted, so each ball's depth is drawn at random.

When do the extra wide serves begin?

The extra wide serves begin once the returner is within two points of the game, well before break point. In the deuce court the wide share climbs from 42.7% at 40-15, with the server comfortably ahead, through 30-30 to 46.7% at deuce. Deuce sits within half a percentage point of break point. In the advantage court, by contrast, the share holds steady from 15-30 through all three break-point scores played there. In the deuce court, break point adds little to the aim servers have settled on by 30-30.

The deuce-court rise comes before break point

Share of first serves aimed wide by score, %, from the server well ahead to the returner closest to winning the game

Top, the deuce court; below, the advantage court; filled dots are break points.

Is it the same server, or weaker servers reaching break point?

The same server follows that pattern within a single match, so weaker servers facing more break points cannot account for the rise. A server's wide share to the deuce court rises from 43.5% on the game's first point to 47.4% on break point. Against 30-30 and deuce, on the other hand, the same share barely moves on break point, from 46.6% to 46.5%. In this sense each server carries to break point the mix of aims he chose at 30-30, as the theory asks.

The same server aims alike at 30-30, deuce and break point

Deuce court, % of first serves aimed wide, each server compared with himself in the same match

Hollow dot the other point, filled dot break point; top pair the game's first point, bottom pair 30-30 and deuce. Each pair counts only matches in which the server played both scores, so each pair has its own set of matches.

Which serve gives way to the wide one?

The extra wide serves replace serves down the T, the aim along the centre line that ends more points than any other. Servers aim down the T with 48.5% of deuce-court first serves at 40-15 but only 44.4% at 15-40. The wide serve, on the other hand, lands 66.8% of the time in the deuce court against 59.5% for the T. Giving up the T costs almost nothing, because once the rally is counted the T wins no more often than the wide serve, the balance the theory predicts.

The T ends the most points yet wins no more of them

Deuce court, men's first serves by aim, %

Aces and unreturned serves are counted among first serves that land; points won include points finished on a second serve.

Does the balance hold on break point?

The balance holds on break points too, as a rerun of Walker and Wooders' test shows. Their test asks, for each server in each court, whether first serves aimed wide and down the T win points equally often. In ten Grand Slam and Masters matches, all but two of them finals, the two economists found serves to either side winning about equally often for 39 of 40 servers and courts.1

Rerun on all charted points, the test fails for 5 of 70 servers and courts, close to the 3.5 failures chance alone would produce. On break points alone, the test passes for 64 of 65 servers and courts, with one failure where chance would produce about three. Because each aim wins about as often as the other even on break point, the score gives a server nothing to gain by changing his aim.

Wide and T serves win about as often, on break points too

Points won by the server after a first serve aimed wide (across) and down the T (up), %, one circle per server and court, servers with 1,200+ first serves to each side on all points and 100+ on break points

Rust is the deuce court and teal the advantage court; on the dashed line the two aims win equally often.

Djokovic's wide serve at 15-40 in Melbourne, on the point that gave Alcaraz the title, was no serve kept in reserve. Although the serve went into the net, the wide aim was as good a choice as the T, since across the charted record Djokovic's deuce-court first serves aimed wide and down the T win 68.9% and 68.1% of points. Facing break point, servers across the men's tour keep the aim they had settled on by 30-30, because break point leaves the wide serve and the T winning about equally often.

1. Mark Walker and John Wooders, "Minimax Play at Wimbledon", American Economic Review 91(5), December 2001, pp. 1521-1538. Ten matches between highly ranked professionals at the four Grand Slams and the year-end Masters, all but two of them finals; first serves only, with serves to the centre (6%) left out. Serves to either side won at equal rates in 39 of 40 server-and-court cases at the 5% level (the exception Sampras serving to Agassi in the deuce court, 1995; joint test p = 0.852), though servers switched sides too often to be truly random (p = 0.001).

Method and sources. Data from the Match Charting Project (Jeff Sackmann and volunteers), men's point-by-point files, 5,569 matches from 8 January 2010 to 21 May 2026, licensed CC BY-NC-SA 4.0; the charts on this page may be reused under the same licence. Tie-breaks are excluded and lets removed. Aim uses first serves with a charted direction (wide, body, down the T); serve depth is not charted. A break point is any point at which the returner is one point from winning the game: 15-40, 0-40, 30-40 or advantage to the returner. In the deuce court the only break-point score is 15-40, so most break points are played in the advantage court. Free points are aces and unreturned serves among first serves that land; points won include points finished on a second serve after a first-serve fault. Same-match comparisons set each server's break-point serves against his own serves at the other score in the same match, among servers with at least ten such points, weighted by the other score's serves. The rerun of Walker and Wooders' test applies Pearson's chi-square to each server in each court (points won after first serves aimed wide against down the T, body serves left out; 1,200+ first serves to each side on all points, 100+ on break points), then their joint test, which gives p = 0.136 on all points (5 of 70 cases fail at the 5% level, about 3.5 by chance; mean gap 1.2 percentage points) and p = 0.367 on break points (1 of 65, about 3.2 by chance; mean gap 4.0 points, from smaller samples). The Djokovic serve in the opening is point 224 of the 2026 Australian Open final in the charting file, coded wide and into the net. Charted matches lean toward big events and well-known players.