Tennis HQ — projections, power rankings & backtest

🚨 Backtest verdict: model is anti-correlated with skill.
Across 1217 graded picks (5/16–6/21), shadow-grade win rate is 32.1% with $-73.71 hypothetical P&L at $2 stakes. Mean CLV vs Pinnacle close is 16.16pp — looks elite on its face — but the signal is INVERTED: win rate at positive CLV = 42.9%; win rate at negative CLV = 65.6%. When we "beat" the close, we lose. When we lose the close, we win.

Calibration: ECE 4.36pp · Brier 0.2417 · barely better than random (0.25). 🔒 Real money stays gated. The dashboard's "0 picks today" isn't a bug — the gates work.

Shadow grade P&L

$-73.71

on 1217 graded $2 picks

Inverted CLV

42.9% / 65.6%

pos / neg CLV win rate

Calibration ECE

4.36pp

target < 5pp · current 4.36

Pinnacle close history

370

picks matched (of 1802)

Three play types — what works, what doesn't

Different markets = different math. The backtest says NONE of these are real-money safe yet. Paper-only.

Kalshi match winner (ML)

CURRENT — paper only

Sim hallucinates +30pp edges on longshots. Backtest: 30pp+ "edges" lost −$145. Gates correctly SKIP these. Don't unblock until sim improves.

PrizePicks props

CURRENT — paper

Aces, total games, BPs Won, sets. Total Games OVER for heavy favorites is the most stable signal. Demon lines too aggressive — fade them.

Parlays (2-3 leg)

PAPER · cross-tournament

Same-match correlated combos: ML + Total Games OVER for heavy favorites (ρ ≈ −0.20, suppresses EV). Cross-tournament 2-leg fav parlays only at ≥62% ML probability per leg.

Surface map — what each surface rewards

Static priors from Sackmann match history. Use these to validate sim outputs against intuition.

Clay	+ Heavy spin	Long rallies, return-game leverage, fewer aces. Higher total games. Specialist tilt is biggest here.
Grass	+ Big serve	Short points, ace rate 1.25× hard. Lower total games. Returners suffer.
Hard	~ Neutral	Balanced. Most ATP/WTA matches. Sim baseline assumption.
Indoor	+ Power serve	Faster than outdoor hard, +10% serve hold. Short careers' specialists thrive.
Aces — surface mult	G 1.25 · H 1.0 · C 0.85 · I 1.10	PP Aces baseline ATP 4.5/match, WTA 2.5
BPs Won conversion	ATP .41 · WTA .45	Strong returners on clay = the PP BPs OVER play

Tomorrow's projected winners · 156 matches simulated

10k Monte Carlo trials per match. Confidence tier from sim probability. Today's slate is mostly Wimbledon qualifying → most projections are coin-flips (correct behavior; the LOW_N regression caps confidence).

Fabian Marozsa v Alex MolcanAlex Molcan68%grass · exp ~23 gamesHIGH

Anton Matusevi v Rei SakamotoRei Sakamoto68%grass · exp ~38 gamesHIGH

Yu Hsiou Hsu v Stefanos SakelStefanos Sak67%grass · exp ~38 gamesHIGH

Dane Sweeny v Franco RoncadeFranco Ronca67%grass · exp ~38 gamesHIGH

Yu Hsiou Hsu v Stefanos SakelStefanos Sak65%grass · exp ~24 gamesHIGH

Stefano Travag v Luka MikrutLuka Mikrut65%grass · exp ~40 gamesHIGH

Anton Matusevi v Rei SakamotoRei Sakamoto64%grass · exp ~24 gamesMED

Otto Virtanen v Pedro MartinezOtto Virtane64%grass · exp ~40 gamesMED

Nikolas Sanche v Soon Woo KwonNikolas Sanc64%grass · exp ~40 gamesMED

Dane Sweeny v Franco RoncadeFranco Ronca64%grass · exp ~24 gamesMED

Vilius Gaubas v Michael MmohMichael Mmoh64%grass · exp ~39 gamesMED

Maya Joint v Emiliana ArangMaya Joint62%grass · exp ~23 gamesMED

Power rankings · composite (surface ELO + form + top-10 bonus)

Current surface: grass. Top 25 each side. Composite = 0.65·overall ELO + 0.35·surface ELO + form bonus + top-10 rank bonus. Restricted to official top 300 to avoid small-sample artifacts.

ATP top 25

Rk	Player	Cty	Comp	ELO	GRASS	Form	Off rk
1	Jannik Sinner	ITA	2051.6	2167	1695	WARM	1
2	Carlos Alcaraz	ESP	1923.6	2070	1717	RETURNING	2
3	Rafael Jodar	ESP	1913.0	1893	1893	WARM	23
4	Alexander Zverev	GER	1898.3	1994	1594	WARM	3
5	Daniil Medvedev	RUS	1808.0	1884	1585	WARM	8
6	Arthur Fils	FRA	1791.0	1892	1546	WARM	21
7	Novak Djokovic	SRB	1788.6	1929	1636	RETURNING	7
8	Tommy Paul	USA	1778.5	1851	1587	WARM	28
9	Felix Auger Aliassime	CAN	1777.3	1854	1517	WARM	4
10	Adolfo Daniel Vallejo	PAR	1775.3	1755	1755	WARM	72
11	Alex De Minaur	AUS	1775.0	1851	1590	NEUTRAL	6
12	Daniel Merida Aguilar	ESP	1767.7	1748	1748	WARM	82
13	Yibing Wu	CHN	1766.8	1747	1747	WARM	101
14	Ben Shelton	USA	1764.8	1816	1561	WARM	5
15	Flavio Cobolli	ITA	1759.5	1833	1557	WARM	10
16	Casper Ruud	NOR	1758.5	1866	1502	WARM	14
17	Lorenzo Musetti	ITA	1753.5	1832	1607	NEUTRAL	16
18	Dino Prizmic	CRO	1753.2	1813	1499	HOT	70
19	Jiri Lehecka	CZE	1750.1	1814	1574	WARM	12
20	Alejandro Tabilo	CHI	1739.8	1798	1575	WARM	31
21	Joao Fonseca	BRA	1737.3	1837	1495	WARM	25
22	Alexander Blockx	BEL	1730.5	1820	1508	WARM	37
23	Jakub Mensik	CZE	1730.1	1825	1497	WARM	17
24	Toby Samuel	GBR	1728.8	1709	1709	WARM	150
25	Learner Tien	USA	1727.2	1847	1506	NEUTRAL	19

WTA top 25

Rk	Player	Cty	Comp	ELO	GRASS	Form	Off rk
1	Aryna Sabalenka	BLR	1963.9	2089	1588	WARM	1
2	Elena Rybakina	KAZ	1925.9	2040	1580	WARM	2
3	Alexandra Shubladze	RUS	1898.6	1879	1879	WARM	184
4	Jeline Vandromme	BEL	1883.0	1863	1863	WARM	176
5	Jessica Pegula	USA	1881.7	1979	1583	WARM	4
6	Iga Swiatek	POL	1876.4	1957	1658	NEUTRAL	3
7	Elina Svitolina	UKR	1869.3	1996	1550	WARM	8
8	Lisa Pigato	ITA	1866.5	1846	1846	WARM	132
9	Mirra Andreeva	RUS	1861.0	1995	1513	WARM	6
10	Coco Gauff	USA	1850.8	1980	1519	WARM	7
11	Akasha Urhobo	USA	1843.4	1823	1823	WARM	180
12	Tyra Caterina Grant	USA	1840.5	1820	1820	WARM	157
13	Dayeon Back	KOR	1825.7	1806	1806	WARM	279
14	Kaitlin Quevedo	USA	1824.7	1825	1825	NEUTRAL	107
15	Victoria Mboko	CAN	1815.1	1937	1515	WARM	9
16	Laura Samson	CZE	1810.3	1810	1810	NEUTRAL	137
17	Alice Tubello	FRA	1810.1	1790	1790	WARM	222
18	Belinda Bencic	SUI	1804.3	1904	1561	WARM	11
19	Carol Young Suh Lee	POC	1801.3	1781	1781	WARM	192
20	Jennifer Ruggeri	ITA	1801.1	1781	1781	WARM	213
21	Marta Kostyuk	UKR	1800.2	1959	1449	WARM	12
22	Anastasia Zolotareva	RUS	1798.0	1778	1778	WARM	233
23	Kristina Liutova	RUS	1796.5	1777	1777	WARM	229
24	Luisina Giovannini	ARG	1793.4	1843	1843	RETURNING	173
25	Angela Fita Boluda	ESP	1793.2	1773	1773	WARM	195

Surface specialists · Δ ELO (surface − overall, top 5 per surface)

Positive Δ = plays meaningfully better than overall game on this surface. Min 10 matches on the surface, top 200 rank.

CLAY

ATP

Zdenek Kolar+28

rk 162 · form NEUTRAL

Gustavo Heide+25

rk 182 · form RETURNING

Jan Choinski+20

rk 105 · form COLD

Juan Carlos Prado Angelo+19

rk 161 · form RETURNING

Gonzalo Bueno+18

rk 180 · form RETURNING

WTA

Katarzyna Kawa+26

rk 143 · form HOT

Despina Papamichail+8

rk 168 · form NEUTRAL

Teodora Kostovic+5

rk 182 · form NEUTRAL

Anna Bondar+5

rk 75 · form WARM

Simona Waltert-3

rk 90 · form WARM

GRASS

ATP

Shintaro Mochizuki+55

rk 129 · form RETURNING

Nicolas Jarry+29

rk 189 · form RETURNING

Billy Harris+26

rk 144 · form COLD

Adrian Mannarino+19

rk 46 · form COLD

Roberto Bautista Agut-25

rk 116 · form COLD

WTA

Tatjana Maria-18

rk 52 · form NEUTRAL

Beatriz Haddad Maia-64

rk 108 · form COLD

Viktoriya Tomova-107

rk 167 · form NEUTRAL

Ajla Tomljanovic-111

rk 109 · form NEUTRAL

Xin Yu Wang-127

rk 31 · form NEUTRAL

HARD

ATP

Hugo Gaston+46

rk 119 · form WARM

Luca Van Assche+36

rk 98 · form WARM

Toby Samuel+30

rk 150 · form WARM

Gauthier Onclin+6

rk 186 · form HOT

Clement Chidekh+5

rk 200 · form NEUTRAL

WTA

Talia Gibson+54

rk 63 · form HOT

Hanne Vandewinkel+51

rk 97 · form HOT

Xiaodi You+1

rk 170 · form WARM

Kimberly Birrell-6

rk 74 · form WARM

Victoria Jimenez Kasintseva-7

rk 124 · form WARM

INDOOR

WTA

Linda Klimovicova-124

rk 165 · form NEUTRAL

Kimberly Birrell-177

rk 74 · form NEUTRAL

Ye Xin Ma-178

rk 189 · form WARM

Carole Monnet-194

rk 191 · form NEUTRAL

Elvina Kalieva-210

rk 135 · form WARM

Cheat sheet — today's filtered top edges

Filters applied: no model_claim flag, no longshot_insufficient_edge, ask 5–95¢, |edge| ≤ 30pp. These are the "cleanest" candidates the system would surface if gates were loosened.

Top Kalshi edges

1Marco CecchinatoELITE_LOW_N

Marco Cecchinato vs Marcelo Tomas Barrios Vera on hard — model 45% vs market 15% (+30.0pp). ELITE_LOW_N, ask 15¢.

2Andres AndradeELITE_LOW_N

Andres Andrade vs Colton Smith on hard — model 49% vs market 20% (+29.4pp). ELITE_LOW_N, ask 20¢. Form: Andres Andra=COLD | Colton Smith=WARM.

3Gonzalo BuenoSTRONG_LOW_N

Gonzalo Bueno vs Jerome Kym on hard — model 45% vs market 16% (+29.0pp). STRONG_LOW_N, ask 16¢.

4Adam WaltonELITE_LOW_N

Adam Walton vs Nick Kyrgios on hard — model 64% vs market 93% (+-28.5pp). ELITE_LOW_N, ask 93¢.

5Florent BaxELITE_LOW_N

Florent Bax vs Chris Rodesch on hard — model 45% vs market 17% (+28.0pp). ELITE_LOW_N, ask 17¢.

Top PrizePicks edges

1Martin LandaluceELITE_LOW_N

Martin Landaluce LESS 0.5 Total Tie Breaks vs Jan-Lennard Struff — proj 0.2 (+50.0% edge). Surface: clay.

2Rei SakamotoELITE_LOW_N

Rei Sakamoto LESS 0.5 Total Tie Breaks vs Anton Matusevich — proj 0.2 (+49.9% edge). Surface: clay.

3Jack DraperSTRONG_LOW_N

Jack Draper LESS 0.5 Total Tie Breaks vs Brandon Nakashima — proj 0.2 (+49.9% edge). Surface: clay.

Fade signals

Clara TausonRETURNINGRETURNING form tag — fade or fade-against.

Gabriel DialloCOLDCOLD form tag — fade or fade-against.

Raphael CollignonRETURNINGRETURNING form tag — fade or fade-against.

Maya JointRETURNINGRETURNING form tag — fade or fade-against.

Antonia RuzicCOLDCOLD form tag — fade or fade-against.

Francisco ComesanaCOLDCOLD form tag — fade or fade-against.

Calibration — predicted vs actual (10pp buckets)

If the model is well-calibrated, gap is ≈ 0 in every bucket. Green = within 5pp, amber within 10pp, red > 10pp.

Bucket	n	Predicted mean	Actual rate	Gap (pp)
0.2–0.30	20	0.283	0.350	+6.68
0.3–0.40	154	0.358	0.279	-7.93
0.4–0.50	521	0.451	0.432	-1.95
0.5–0.60	487	0.545	0.495	-4.99
0.6–0.70	124	0.640	0.581	-5.98
0.7–0.80	14	0.718	0.857	+13.93
0.8–0.90	1	0.823	1.000	+17.68

Where it bleeds — market price bucket (FLB signal)

The favorite-longshot bias hits hardest in the 30-50% market price bucket. Shadow-graded picks at $2 stakes.

Mkt P	n	Win rate	P&L @ $2	Mean CLV (pp)
<10%	92	0.100	$107.22	32.53
10-30%	609	0.223	$32.44	21.47
30-50%	754	0.334	$-129.83	11.79
50-70%	192	0.541	$-6.91	6.21
70-90%	55	0.472	$-29.21	3.48
>90%	100	0.585	$-47.42	1.92

Where it bleeds — edge bucket

Edge	n	Win rate	P&L @ $2
0-2pp	143	0.330	$-59.68
2-4pp	113	0.384	$-31.07
4-8pp	222	0.434	$-8.84
8-15pp	331	0.362	$-7.95
15-30pp	635	0.287	$178.84
30pp+	358	0.258	$-145.00

All data CA-legal (Kalshi + PrizePicks). Predictions are research-only, paper-trade until CLV vs Pinnacle proves real edge.
Built from tennis_* JSON artifacts. Live tactical dashboard →