{ "license": "https://creativecommons.org/licenses/by/4.0/", "licenseName": "CC BY 4.0", "credit": "The Tennis Desk, https://thetennisdesk.com/research/does-putting-your-best-pair-on-line-1-win-more-matches", "provenance": { "statements": { "derived": { "from": "docs/research/records/lineup-stacking.json: each statement's own SQL, executed read-only on the research login by scripts/research-weekly-run.mjs, which re-runs every statement and refuses to write this file if any n, hits or value disagrees with the record" } } }, "slug": "does-putting-your-best-pair-on-line-1-win-more-matches", "question": "Does putting your best pair on line 1 win more matches?", "verdict": "null-but-tellable", "measuredOn": "2026-10-15", "statements": { "top-line-check": { "n": 13551, "hits": null, "value": [ 0, 0 ], "claim": "Check, not printed (required check): scorecards whose top doubles line is undefined, because a doubles line has no line number or more than one doubles line holds the lowest number. Then the same count inside the matches measured, which must be zero: 0, 0, over 13,551 matches measured.", "unit": "scorecards" }, "unplayed-check": { "n": 57984, "hits": null, "value": [ 0, 0 ], "claim": "Check, not printed (required check): lines whose score is missing, blank or has no games set, or whose winner is neither Home nor Visitor, and the scorecards they sit on. Each such scorecard is excluded whole as its own bucket: 0, 0, of 57,984 lines held.", "unit": "lines, scorecards" }, "line-lag-check": { "n": 138589, "hits": null, "value": [ -0.8, 1.4, 3, -1.2, -0.1, 0.2, -1.2, -0.6, 0, -0.6, -0.4, 0.1, -3.9, -1.4, -1.1, -0.7, 0 ], "claim": "Check, not printed (required check): on doubles lines played by players with at least six earlier rated lines, the mean change in the shadow model's level across the line (after minus before), in thousandths of a point, by roster NTRP and by doubles line rank (top, second, third or lower). That is twelve cells, null under fifty rows. Then, for each NTRP, the top line's mean change minus the larger of the lower lines' mean changes, in thousandths of a point. Last, one if the check fired and zero if not. The check compares the top line with each lower line separately: it fires if, for any NTRP, the top line's mean change exceeds either lower line's by twenty thousandths of a point or more, or if no NTRP can be read. So a printed difference under twenty can sit beside a fired flag: -0.8, 1.4, 3, -1.2, -0.1, 0.2, -1.2, -0.6, 0, -0.6, -0.4, 0.1, -3.9, -1.4, -1.1, -0.7, 0, over 138,589 player rows on doubles lines, from players with at least six earlier rated lines.", "unit": "thousandths of a level point; flag" }, "start-level-check": { "n": 10095, "hits": null, "value": [ 2.89, 2.89, 2.89, 3.37, 3.37, 3.37, 3.85, 3.85, 3.85, 4.33, 4.32, 4.33, 1.3, 2.3, 1, 4.2, 0 ], "claim": "Check, not printed: for a player's first rated line, when it was a doubles line, the mean shadow-model level before the line, in points to two decimals, by roster NTRP and by doubles line rank. That is twelve cells, null under fifty rows. Then each NTRP's spread across its cells that clear the floor, in thousandths of a point. Last, one if the check fired and zero if not: 2.89, 2.89, 2.89, 3.37, 3.37, 3.37, 3.85, 3.85, 3.85, 4.33, 4.32, 4.33, 1.3, 2.3, 1, 4.2, 0, of 10,095 first rated lines.", "unit": "level points; thousandths of a level point; flag" }, "start-level-counts": { "n": 10095, "hits": null, "value": [ 621, 752, 897, 901, 1049, 1078, 806, 935, 909, 408, 438, 404 ], "claim": "Check, not printed: first rated lines in each of the twelve cells of the start-level check, in the same order: 621, 752, 897, 901, 1,049, 1,078, 806, 935, 909, 408, 438, 404, of 10,095 first rated lines.", "unit": "lines" }, "date-format": { "n": 200201, "hits": null, "value": 0, "claim": "Check, not printed: rows in the pinned run whose match date is missing or not in two-digit month, day and year form. It must be zero: 0, of 200,201 rows.", "unit": "rows" }, "date-year": { "n": 200201, "hits": null, "value": [ 40, 3, 35 ], "claim": "Check, not printed: rows in the pinned run whose match date is well formed but whose year falls outside the seasons measured, and the scorecards and people those rows touch. These rows get no date key, so they count toward nobody's earlier lines, and their players count as inexperienced on those lines: 40, 3, 35, of 200,201 rows.", "unit": "rows, scorecards, people" }, "side-labels": { "n": 200201, "hits": null, "value": 0, "claim": "Check, not printed: rows in the pinned run whose side maps to neither home nor away. It must be zero: 0, of 200,201 rows.", "unit": "rows" }, "duplicates": { "n": 200201, "hits": null, "value": [ 0, 0 ], "claim": "Check, not printed: person-line keys in the pinned run with more than one row, and how many of those disagree on the level before the line. The second must be zero: 0, 0, of 200,201 keys.", "unit": "keys" }, "run-coverage": { "n": 15811, "hits": null, "value": [ 99.6, 99.6, 99.6, 99.6, 99.7 ], "claim": "Check, not printed: of the decided scorecards with both team ids in each season measured, how many in a hundred carry rows from the pinned run. Each must be at least ninety-five: 99.6, 99.6, 99.6, 99.6, 99.7, of 15,811 decided scorecards.", "unit": "per hundred, by season" }, "line-types": { "n": 57984, "hits": null, "value": [ 46041, 11943 ], "claim": "Check, not printed: lines labelled doubles and lines labelled singles, in any case: 46,041, 11,943, of 57,984 lines.", "unit": "lines" }, "line-types-unknown": { "n": 57984, "hits": null, "value": 0, "claim": "Check, not printed: lines whose label is neither doubles nor singles. It must be zero: 0, of 57,984 lines.", "unit": "lines" }, "excluded": { "n": 16065, "hits": null, "value": [ 0, 251, 3, 0, 1944, 0, 0, 20, 0, 0, 1, 0, 295, 13551, 16065 ], "claim": "Scorecards held, each counted once in the first bucket it falls into, in this order: no score; tied; a team id missing; season outside those measured; a line with no players named; a score text recording a default, walkover or retirement; a line with no played score or no Home or Visitor winner; fewer than two doubles lines; a player row on no side; an undefined top doubles line; a doubles pair without two rated players; a named slot that does not match a pinned-run row (a named player with no row, a row on a line or side the scorecard does not hold, or named and rated counts that differ); a tie for the best pair; and kept. The kept count must equal the funnel's matches measured. The last figure is the distinct scorecard ids held, which must equal the scorecards held: 0, 251, 3, 0, 1,944, 0, 0, 20, 0, 0, 1, 0, 295, 13,551, 16,065, of 16,065 scorecards held.", "unit": "scorecards per bucket" }, "funnel": { "n": 13551, "hits": null, "value": [ 7401, 3536, 240, 1268, 1106, 1106 ], "claim": "Of the 13,551 matches measured, each counted once in the first bucket it falls into: both or neither team put its best pair on top; a strength gap of a tenth or more; even overall but a doubles gap of a tenth or more; a best pair less than a tenth ahead of the second on either side; and the full cell. Then, as a check, the full cell counted directly, which must equal the last bucket: 7,401, 3,536, 240, 1,268, 1,106, 1,106.", "unit": "matches per bucket" }, "fielded-coverage": { "n": 13551, "hits": null, "value": [ 176182, 176182 ], "claim": "Check, not printed: across the matches measured, the player slots named on every line, singles included, and the pinned-run rating rows on those scorecards. The two must be equal: 176,182, 176,182, over 13,551 matches.", "unit": "slots, rows" }, "head-population": { "n": 1106, "hits": null, "value": [ 0, 1106, 488, 1106 ], "claim": "Check: the figures in 0, 1,106, 488, 1,106 are, in order: one if the start-level check fired and the headline is the experienced cut, zero if not; the headline matches counted directly; the experienced cut; and the full cell. 1,106 headline matches.", "unit": "flag, matches" }, "headline": { "n": 1106, "hits": 567, "value": 51.3, "claim": "Take the 1,106 team matches between teams of similar strength before the match, where each team's best pair was clearly ahead of its second pair, and only one team put the pair the Desk's shadow rating model rated highest on the top doubles line. In those matches, that team won 51.3 out of a hundred.", "unit": "per hundred" }, "headline-interval": { "n": 1106, "hits": null, "value": [ 48.3, 54.2 ], "claim": "The ninety-five in a hundred interval around that share runs from the first to the second of 48.3, 54.2, over 1,106 team matches.", "unit": "per hundred" }, "full-cell": { "n": 1106, "hits": 567, "value": 51.3, "claim": "Across every qualifying match, whether or not the players had earlier rated lines, the team that put its best pair on top won 51.3 out of a hundred, over 1,106 team matches.", "unit": "per hundred" }, "headline-experienced": { "n": 488, "hits": 241, "value": 49.4, "claim": "Keeping only the matches where every player in both teams' top two pairs had at least six rated lines before that date, the team that put its best pair on top won 49.4 out of a hundred, over 488 team matches.", "unit": "per hundred" }, "headline-experienced-interval": { "n": 488, "hits": null, "value": [ 45, 53.8 ], "claim": "The ninety-five in a hundred interval around that share runs from the first to the second of 45, 53.8, over 488 team matches.", "unit": "per hundred" }, "headline-pair-history": { "n": 789, "hits": 396, "value": 50.2, "claim": "Keeping only the headline matches where neither team's top-line pair had played more lines together before that date than the team's best pair, the team that put its best pair on top won 50.2 out of a hundred, over 789 team matches.", "unit": "per hundred" }, "headline-pair-history-interval": { "n": 789, "hits": null, "value": [ 46.7, 53.7 ], "claim": "The ninety-five in a hundred interval around that share runs from the first to the second of 46.7, 53.7, over 789 team matches.", "unit": "per hundred" }, "pair-history-dropped": { "n": 1106, "hits": null, "value": [ 317, 789 ], "claim": "Check: of 1,106 headline matches, the first of 317, 789 were left out because a team's top-line pair had more lines together before that date than its best pair. The second figure is the matches kept.", "unit": "matches" }, "headline-home-share": { "n": 1106, "hits": 555, "value": 50.2, "claim": "In the headline matches, the team that put its best pair on top was the home team 50.2 times out of a hundred, over 1,106 team matches.", "unit": "per hundred" }, "venue-baseline": { "n": 1549, "hits": null, "value": [ 53.5, 50 ], "claim": "Check; the second figure is the headline's venue-only bar. Take the 1,549 similar-strength matches with a clear best pair on both sides where both teams or neither team put that pair on top. The home team won the first of 53.5, 50 out of a hundred of them. Given how often placing teams were at home in the headline, venue alone would have them win the second.", "unit": "per hundred" }, "home-split": { "n": 1106, "hits": null, "value": [ 54.2, 48.3 ], "claim": "In the headline matches, the share won by the team that put its best pair on top when it was at home, and when it was away: 54.2, 48.3, over 1,106 team matches.", "unit": "per hundred" }, "by-margin": { "n": 2374, "hits": null, "value": [ 48.9, 50.3, 53.5 ], "claim": "Among similar-strength matches where one team put its best pair on top, the share that team won, by the smaller of the two best-pair leads: under a tenth, a tenth to a quarter, and a quarter or more: 48.9, 50.3, 53.5, over 2,374 team matches.", "unit": "per hundred, by margin band" }, "by-margin-counts": { "n": 2374, "hits": null, "value": [ 1268, 781, 325 ], "claim": "Check, not printed: the matches in each by-margin band, in the same order. Bands two and three must sum to the headline n: 1,268, 781, 325, over 2,374 team matches.", "unit": "matches per band" }, "by-margin-experienced": { "n": 907, "hits": null, "value": [ 49.9, 49.5, 49 ], "claim": "The by-margin bands again, keeping only matches where every top-two-pair player on both sides had at least six earlier rated lines, for leads under a tenth, a tenth to a quarter, and a quarter or more: 49.9, 49.5, 49, over 907 team matches.", "unit": "per hundred, by margin band" }, "by-margin-experienced-counts": { "n": 907, "hits": null, "value": [ 419, 331, 157 ], "claim": "Check, not printed: the matches in each band of the experienced by-margin statement, in the same order. Bands two and three must sum to the n of the experienced cut: 419, 331, 157, over 907 team matches.", "unit": "matches per band" }, "by-gap": { "n": 2995, "hits": null, "value": [ 16.7, 51.3, 58.5, 86.8 ], "claim": "With a clear best pair on both sides, the share won by the team that put that pair on top, when that team was weaker before the match, even on both measures, even overall but uneven in doubles, and stronger: 16.7, 51.3, 58.5, 86.8, over 2,995 team matches.", "unit": "per hundred, by strength band" }, "by-format": { "n": 1106, "hits": null, "value": [ 58, 50.8, null ], "claim": "The same share for team matches with two, three, and four or more doubles lines: 58, 50.8, withheld, over 1,106 team matches.", "unit": "per hundred, by format" }, "by-league": { "n": 1106, "hits": null, "value": [ 55.4, 51.1, 48.1, 48.7, null ], "claim": "The same share in adult eighteen-and-over and forty-and-over leagues, combo, mixed, fifty-five and over, and other leagues: 55.4, 51.1, 48.1, 48.7, withheld, over 1,106 team matches.", "unit": "per hundred, by league type" }, "by-season": { "n": 1106, "hits": null, "value": [ 51.3, 56, 52.4, 46.4, 50.8 ], "claim": "The same share measured alone for each season measured, oldest first: 51.3, 56, 52.4, 46.4, 50.8, over 1,106 team matches.", "unit": "per hundred, by season" }, "by-phase": { "n": 1106, "hits": null, "value": [ 51.9, 46.9 ], "claim": "The same share in the regular season, and in playoffs and sectionals: 51.9, 46.9, over 1,106 team matches.", "unit": "per hundred, by phase" }, "strength-gaps": { "n": 1106, "hits": null, "value": [ 1.5, 1.9 ], "claim": "Check: in the headline matches, the placing team's mean level gap over the other team before the match, in thousandths of a point, first over every player slot fielded and then over doubles players only: 1.5, 1.9, over 1,106 team matches.", "unit": "thousandths of a level point" }, "concentration": { "n": 1106, "hits": null, "value": [ 371, 279.1, 91.9 ], "claim": "In the headline matches, how far the best pair led the second pair before the match, on average, in thousandths of a point: for the team that put it on top, for the team that put it lower, and the difference between the two: 371, 279.1, 91.9, over 1,106 team matches.", "unit": "thousandths of a level point" }, "balanced-concentration": { "n": 369, "hits": 193, "value": 52.3, "claim": "Where the two teams' best pairs led their second pairs by amounts within a tenth of a point of each other, the team that put its best pair on top won 52.3 out of a hundred, over 369 team matches.", "unit": "per hundred" }, "cluster": { "n": 1106, "hits": null, "value": [ 875, 896, 3.3, 3.1 ], "claim": "Check: the figures in 875, 896, 3.3, 3.1 are, in order: the distinct team-seasons that put the best pair on top in the headline matches; the distinct team-seasons that faced them; and, for each of those two groups, the share of the matches held by its ten largest team-seasons, in a hundred. Over 1,106 team matches.", "unit": "team-seasons; per hundred" }, "one-per-team-season": { "n": 875, "hits": 441, "value": 50.4, "claim": "Keeping one headline match for each team-season that put its best pair on top, chosen without looking at the result, that team won 50.4 out of a hundred, over 875 team matches.", "unit": "per hundred" }, "replication-within-team": { "n": 852, "hits": 1025, "value": [ 1025, 1015.1, 9.9, 0.66 ], "claim": "This counts 852 team-seasons. Each played similar-strength matches with a clear best pair, put that pair on top in some of them and lower in others, and both won and lost. The figures in 1,025, 1,015.1, 9.9, 0.66 are, in order: the on-top wins observed, the on-top wins expected if order made no difference, observed minus expected, and that difference in standard errors.", "unit": "wins; standard errors" }, "replication-counts": { "n": 2669, "hits": null, "value": [ 1106, 852, 2014, 1510 ], "claim": "Check, not printed: the figures in 1,106, 852, 2,014, 1,510 are, in order: the team-seasons with matches of both kinds; those that are also informative, meaning they won some and lost some; and the on-top and the lower matches, each summed over the informative team-seasons. Of 2,669 team-seasons with any qualifying match.", "unit": "team-seasons; matches" }, "base-rate": { "n": 27102, "hits": 16836, "value": 62.1, "claim": "Of 27,102 team lineups in the matches measured, 62.1 in a hundred put the pair the Desk's shadow rating model rated highest on the top doubles line.", "unit": "per hundred" }, "lines-won": { "n": 3650, "hits": null, "value": [ 69.2, 69.1, 3640 ], "claim": "This answers a different question, about lines rather than team matches, and it is unadjusted for who the pair met. In similar-strength matches, take the pair the Desk's shadow rating model rated highest, when it was clearly ahead of its team's second pair. How often it won its own line, out of a hundred, first on the top doubles line and then on a lower line, is the first two of 69.2, 69.1, 3,640. The last figure counts best-pair line-sides where both teams' best pairs met, which are left out. A line-side is one team's side of one line, so a line where the best pairs met counts once for each side whose best pair was clearly ahead, up to twice. Over 3,650 best-pair line-sides.", "unit": "per hundred; lines" }, "headline-count": { "n": 1106, "hits": null, "value": 1106, "claim": "The share in headline rests on 1,106 team matches.", "unit": "team matches" }, "headline-experienced-count": { "n": 488, "hits": null, "value": 488, "claim": "The share in headline-experienced rests on 488 team matches.", "unit": "team matches" }, "headline-pair-history-count": { "n": 789, "hits": null, "value": 789, "claim": "The share in headline-pair-history rests on 789 team matches.", "unit": "team matches" }, "balanced-concentration-count": { "n": 369, "hits": null, "value": 369, "claim": "The share in balanced-concentration rests on 369 team matches.", "unit": "team matches" }, "one-per-team-season-count": { "n": 875, "hits": null, "value": 875, "claim": "The share in one-per-team-season rests on 875 team matches.", "unit": "team matches" }, "base-rate-count": { "n": 27102, "hits": null, "value": 27102, "claim": "The share in base-rate rests on 27,102 lineups.", "unit": "lineups" }, "headline-interval-low": { "n": 1106, "hits": null, "value": 48.3, "claim": "The low end of the headline's ninety-five in a hundred interval is 48.3 out of a hundred.", "unit": "per hundred" }, "headline-interval-high": { "n": 1106, "hits": null, "value": 54.2, "claim": "The high end of the headline's ninety-five in a hundred interval is 54.2 out of a hundred.", "unit": "per hundred" } } }