Punt strategies for nine-category leagues

Up: 2026-27 Fantasy Basketball Encyclopedia

Related: 2026-27 category rankings · 2026-27 fantasy basketball tiers · 2026-27 fantasy basketball rankings · 2026-27 fantasy basketball player projections · 2026-27 points-league rankings · Robust punt builds in fantasy basketball · Percentage categories and volume · 2026-27 fantasy basketball projection methodology · 2026-27 sleepers, breakouts, and fades · 2026-27 fantasy basketball evidence and confidence policy

Board dependency Every player line quoted here comes from the Rome desk's frozen July 18 export, player-projections-2026-07-18.csv. These are projections, not facts. Re-run the correlation and supply tables after camp roles settle.

Checked against corrected data 2026-07-18 The July 18 export fabricated shot attempts for 429 of 549 players (Projection engine defects and corrections). This note's correlation matrix and all eight supply counts were re-run against the corrected export and reproduce exactly — they are computed from counting stats and raw FG%/FT% rates, none of which the defect touched. What did change is the rank of several anchors, because ranks depend on attempt-weighted impact. Where an anchor moved ten or more slots it is flagged inline; the punt-FT% section is materially rewritten. Board: 2026-27 volume-corrected nine-category board.

The threshold math

In Yahoo's head-to-head categories format with the default nine categories, you need five of nine. Model each category as an independent Bernoulli trial with win probability \(p_i\); matchup win probability is

\[ P(\text{win}) = \Pr\!\left[\sum_{i=1}^{9} X_i \ge 5\right], \qquad X_i \sim \text{Bernoulli}(p_i). \]

Independence is a lie — category outcomes are correlated within a matchup — but the comparative arithmetic survives it. Take a balanced roster at \(p_i = 0.55\) across all nine: \(P(\text{win}) = 0.6214\). Now punt one category to \(p_1 = 0\) and ask what the other eight must reach:

The single most useful number here is 0.6118. Conceding one category and lifting the other eight from 55% to 60% loses to simply being balanced. Punting is not free optionality; it is a bet that concentration buys you at least six points of win probability per surviving category. A double punt needs each of seven categories at ~69%. That is why most failed punt teams are not unlucky — they are arithmetically underwater from the draft.

Two corollaries the table makes obvious. First, a soft punt beats a hard one: keeping \(p_1 = 0.05\) instead of 0 is worth +0.011 win probability at no cost, because bad weeks for opponents are free. Never actively trade away a category you are merely bad at. Second, surplus above certainty is worthless. If you win rebounds 95% of the time, the 96th percentile rebounder adds nothing — that wasted surplus is precisely what punting recovers.

Category correlations on this board

Computed directly from the July 18 export: Pearson \(r\) across the 241 players projecting ≥40 GP and ≥16 MPG with complete category lines. FG% and FT% are raw rates here, not attempt-weighted impact — see Percentage categories and volume for why roster-level ratio value depends on attempts.

Two clusters fall out cleanly. The big-man cluster: FG%–REB (+0.62), REB–BLK (+0.67), FG%–BLK (+0.54). The guard cluster: FT%–3PM (+0.72), PTS–AST (+0.67), AST–STL (+0.53). The clusters are joined by strongly negative bridges: FG%–FT% (-0.59), FG%–3PM (-0.59), FT%–REB (-0.41), FT%–BLK (-0.38).

A punt is coherent when the category you concede sits on a negative bridge to the categories you want. Conceding FT% buys you free access to the entire big cluster. Conceding steals buys you nothing, because steals' only strong ties are to assists (+0.53) and points (+0.42) — categories you are trying to keep. Elite Fantasy Basketball's punt-blocks guide reaches the same structural conclusion from the opposite direction. Note that NBC's stat-correlation work measures a different object — category-to-overall-value correlation, where FG% (0.112) and FT% (0.168) are the weakest 9-cat drivers, which is a second reason the percentages are the cheapest things to concede.

Why punt FT% is the most popular build

Three reasons, in order of importance.

1. It is the only punt with a bounded, self-limiting cost. FT% is a ratio over attempts, and the players who wreck it barely shoot free throws. Worked example using the seven measured-attempt bigs on this board — Walker Kessler (.700 on 3.0 FTA), Daniel Gafford (.690/2.7), Zach Edey (.735/2.8), Mark Williams (.750/2.7), Kel'el Ware (.740/1.6), Isaiah Hartenstein (.650/2.2), Isaiah Stewart (.750/1.8):

\[ FT\% = \frac{\sum FTA_i \cdot FT\%_i}{\sum FTA_i} = \frac{11.94}{16.7} = .715 \]

A guard-heavy core of SGA (.880/8.9), Devin Booker (.875/7.5), Luka Dončić (.790/9.0), Kyrie Irving (.900/4.0), Tyrese Maxey (.890/5.4), Austin Reaves (.870/7.0) and Joel Embiid (.855/8.6) posts \(43.35/50.4 = .860\) on 50.4 attempts versus 16.7. You lose FT% every week with near-certainty — which is exactly what a punt wants, a reliable zero rather than a coin flip — but the low attempt volume means the concession never metastasises into a points or FG% problem. (Attempt figures are the corrected projections; the ratio itself is unchanged to three decimals.)

2. The bridges are the strongest on the board. No other single concession opens three categories at \(|r| \ge 0.38\).

3. Supply arrived in the middle rounds. Elite Fantasy Basketball notes that the build used to require spending early picks on Gobert/Drummond/Jordan types; now the FG%/REB/BLK sources are mid-round. This board agrees: 25 of 241 qualified players project FT% ≤ .720 alongside ≥1.0 BLK or ≥8.0 REB, and most sit outside the top 50.

The eight punts

Supply counts below are from the same 241-player pool.

Punt FT%

Anchors, with corrected rank and the FT% priced against real projected attempts rather than a points proxy — the two numbers that matter for this build are the percentage and the volume behind it:

Board correction: Jalen Duren projects .755 here on 6.4 attempts, close to the .780 reference — he is a rebound/FG% anchor, not an FT% discount. Do not pay the punt tax for him.

Correction 2026-07-18 — the two headline punt-FT anchors repriced This build's economics changed more than any other under the volume correction, because punt FT% is the one build whose entire thesis is free-throw volume, and volume is exactly what the July 18 export fabricated.

Giannis fell 21 → 46. The proxy assigned him 5.8 free-throw attempts; his real 2025-26 rate of 9.9 per game scales to 11.6 at his projected 34 minutes (StatMuse confirms 9.9 on 65.0%). His .665 was being priced at roughly half its true cost. The practical effect is the opposite of a downgrade for this build: Giannis is a worse player on a balanced roster and therefore cheaper to acquire, while remaining exactly as productive for a manager who has already conceded the category. He is now a round-4 price for a top-20 punt-FT asset, which is the single best value this build has.

Gobert fell 99 → 138 for the same reason — .560 on 3.8 real attempts rather than the ~2.1 the proxy implied (ESPN has him at 52.6% in 2025-26). He is no longer a mid-round anchor at all; he is a rounds-11-to-12 pick in a 12-team league and a free agent in a 10-team league. Do not spend a round-8 pick on him.

Net: the previous version of this section listed Gobert second among anchors on the implicit assumption he was a rounds-8-to-9 asset. He and Giannis have swapped roles — Giannis is now the build's centrepiece at a discount, Gobert is late-round filler. Recorded rather than silently overwritten because the earlier ordering was this desk's genuine belief on the defective board.

Declare: rounds 3–5, or round 2 if Giannis falls to you. Failure mode: center congestion. You bank FG%/REB/BLK early, then discover AST and 3PM are unrepairable because the perimeter shelf emptied while you were taking your fourth big. The two non-bigs above — Daniels (now 41) and Ausar Thompson (68) — exist precisely to relieve this, and Daniels rising 65 → 41 makes that relief more expensive than it was.

Punt FG%

Anchors: Trae Young (.438, rank 33, was 55), LaMelo Ball (.420, 29), James Harden (.430, 42), Damian Lillard (.430, 101, was 81), Derrick White (.421, 32), Fred VanVleet (.390, 112, was 102), Brandon Miller (.445, 31), Coby White (.445, 81), Immanuel Quickley (.443, 61, was 51), Reed Sheppard (.435, 69, was 54), plus stretch blockers Myles Turner (.450, 117, was 128), Kristaps Porziņģis (.452, 38), Jaren Jackson Jr (.475, 47). Supply is deep and the count is unchanged by the correction: 40 qualified players combine FG% ≤ .450 with ≥2.0 3PM.

Correction 2026-07-18 — the anchor got expensive. Trae Young rose 55 → 33 because the proxy priced his .875 free-throw shooting on ~4.6 attempts when his real projected volume is 8.1 (StatMuse has him at 7.3/g in 2025-26, scaled to 32 projected minutes). He is the rare punt-FG% anchor who is also a genuine FT% and AST asset, and he now costs a round-3 pick rather than a round-5 one. Everything else in this lane got cheaper: Lillard, VanVleet, Quickley and Sheppard all fell, because inefficient perimeter volume was the group the proxy most flattered. The build is still deep — you are simply buying its best piece earlier and its filler later.

Declare: rounds 2–4. Failure mode: accidental double punt on turnovers. High-usage inefficient guards carry TO at +0.82 correlation with points; you set out to punt one category and quietly punted two. The stretch bigs above are the fix — they preserve BLK and REB without the TO tax.

Punt turnovers

Anchors: Luka Dončić (4.1), Trae Young (4.2), Cade Cunningham (3.7), Nikola Jokić (3.6), Giannis Antetokounmpo (3.5), Josh Giddey (3.5), Jalen Johnson (3.3), Paolo Banchero (3.2), Alperen Şengün (3.0). Only 17 players project ≥3.0 TO — the scarcity here is on the punting side. Declare: it can be declared in round 1 without cost, the one punt for which this is true, because taking Dončić or Jokić first overall already commits you. Failure mode: none structurally — TO is the highest-leverage concession on the board (\(r = +0.83\) with AST, \(+0.82\) with PTS). The real risk is that it is too easy: everyone in your league wants the same usage monsters, so the reallocation premium gets bid away. Check ADP before assuming it is cheap (ADP and draft market desk).

Punt assists

Anchors: Victor Wembanyama (3.4 AST but elite everywhere), Walker Kessler (2.1), Chet Holmgren (1.8), Jaren Jackson Jr (2.0), Myles Turner (1.2), Kel'el Ware (0.8), Michael Porter Jr. (2.8), Trey Murphy III (3.6), OG Anunoby (2.2), Lauri Markkanen (2.2), Keegan Murray (1.8). Declare: rounds 4–6, after you see whether the assist shelf has collapsed. Only 12 players project ≥7.0 AST and 26 reach ≥6.0; if six of the twelve are gone by round 3, the punt is being declared for you. Failure mode: assists and steals correlate at +0.53, so a punt-AST roster drifts into a steals deficit it did not intend. Budget one dedicated steals source — Cason Wallace (1.9) or Herb Jones (1.7) — as insurance.

Punt points

Anchors: Dyson Daniels (12.0), Ausar Thompson (11.2), Cason Wallace (9.2), Herb Jones (9.0), Rudy Gobert (10.2), Isaiah Hartenstein (8.8), Moussa Diabaté (8.2), Draymond Green (7.9), Keon Ellis (9.2), Kris Dunn (6.3). Declare: rarely, and never before round 6. Failure mode: structural. Points has no negative bridge to anything (FG% +0.01 is the closest), so conceding it buys you only low turnovers. Worse, PTS correlates +0.56 with 3PM, so punt-PTS silently becomes punt-PTS-and-3PM — a double punt requiring 69% in the surviving seven. Treat this as a consequence of a defensive-specialist build, not a plan.

Punt three-pointers

Anchors: Rudy Gobert, Ivica Zubac, Jalen Duren, Daniel Gafford, Jarrett Allen (all 0.0 projected 3PM), Giannis Antetokounmpo (0.7), Zion Williamson (0.1), Domantas Sabonis (0.4), Amen Thompson (0.3), Alperen Şengün (0.5). Declare: rounds 4–6. Failure mode: it is the near-twin of punt FT% (both buy FG%/REB/BLK), so running both is not a double punt in resource terms but is a double punt in the win-probability table — you now need 69% across seven. It is also the easiest punt to accidentally reverse: 77 qualified players project ≥2.0 3PM, so a single waiver add can un-punt you and waste the concession.

Punt blocks

Anchors: Nikola Jokić (0.8), Luka Dončić (0.5), Shai Gilgeous-Alexander (0.7), Trae Young (0.2), Tyrese Maxey (0.5), Stephen Curry (0.3), Tyrese Haliburton (0.5), Karl-Anthony Towns (0.6), Payton Pritchard (0.1). Declare: rounds 1–3 — this is the punt most often implied by simply drafting the best perimeter players available. Blocks are the scarcest counting category: only 13 of 241 project ≥1.5 BLK, and Elite Fantasy Basketball's count for 2024-25 was nine players at 1.5+ against twenty a decade earlier. Failure mode: rebounds. BLK–REB correlate at +0.67, so the guard-heavy roster that abandons blocks usually abandons rebounds too. Force one high-rebound, low-block big — Domantas Sabonis (11.8 REB / 0.3 BLK) and Karl-Anthony Towns (11.5 / 0.6) are the exact profile — or accept the second punt.

Punt steals

Anchors: Karl-Anthony Towns (0.9), Domantas Sabonis (0.8), Ivica Zubac (0.5), Lauri Markkanen (0.9), Myles Turner (0.6), Damian Lillard (0.7), Zach Edey (0.5). Declare: don't. Failure mode: it has no upside to fail from. Steals' largest negative bridge is FG% at -0.12 — statistically nothing. You concede a category and receive no structural freedom in return, which is the worst possible trade under the 0.6118 breakeven. Only 11 players project ≥1.5 STL, so steals is genuinely scarce; but scarcity is a reason to buy a category cheaply, not to concede it. Punt steals only when it is the residue of a punt-AST build you already committed to.

When to declare

Rarely in round 1. On this board the four Tier-1 engines — Jokić, Wembanyama, SGA, Dončić — are explicitly build-agnostic in 2026-27 fantasy basketball tiers, and none of them forces a punt. The two round-1-to-2 exceptions are Walker Kessler at #7, whose .700 FT / 0.3 3PM / 2.1 AST line declares three punts simultaneously, and taking Dončić or Trae Young early, which declares punt-TO for free.

The realistic sequence:

SportsEthos' punt framework makes the same point qualitatively — punting means ignoring a category, not actively degrading it — and warns beginners off three-and-four-category builds, which the double-punt breakeven of 0.6887 quantifies.

Format matters more than the punt does

H2H each-category. Records are kept per category, so a 6-3 week is recorded as 6-3 (Yahoo, ESPN). This is the worst format for punting: a punted category is a guaranteed loss every single week, roughly 22 losses over a 22-week season, and you receive no threshold bonus for the fifth win. Each-category scoring is closer to roto in incentive structure than to most-categories. Soft-punt only.

H2H most-categories. One W or L per week regardless of margin. This is the format punting was invented for. All surplus above five categories is discarded by the scoring rule itself, so the wasted-surplus recovery that punting performs is pure profit. Everything above assumes this format.

Roto. Punting is materially weaker, and the usual explanation — "you lose 1/9 of the standings points" (RotoWire, ESPN) — is true but shallow. The real mechanism: roto has no wasted surplus. In a 12-team roto league, moving from 6th to 1st in rebounds is worth 5 standings points, exactly as much as moving from 6th to 1st in any other category. Marginal production converts linearly everywhere. Punting's entire edge in H2H comes from the majority threshold making surplus worthless — delete the threshold and you delete the edge, leaving only the cost. Arithmetically: punting to last place in a 12-team roto costs \(6.5 - 1 = 5.5\) points and requires recovering \(5.5/8 \approx 0.69\) standings places per surviving category, with no compensating nonlinearity. See also the rotisserie optimization literature, which treats roto as a linear-objective allocation problem — precisely the structure that punts do not exploit.

10-team versus 12-team

The difference is replacement level, and it points in a clear direction: punt less in a 10-team league.

In a 10-team league every roster holds genuine contributors and the waiver wire still contains ~111 qualified players, including — on this board — real block and rebound sources like Ryan Kalkbrenner (1.2 BLK), Adem Bona (1.1) and Mitchell Robinson (7.6 REB). A punted category is therefore repairable in-season, which means the concession was never necessary: you can be balanced and strong, and balance at \(p = 0.58\) across nine (\(P = 0.6903\)) beats a punt with a 62% reallocation (\(P = 0.6401\)). Shallow leagues reward best-available drafting and aggressive streaming.

In a 12-team league the marginal roster spot is a sub-replacement player and the waiver pool thins to ~85 qualified names, most of them role-fragile. Category repair costs a real asset rather than a free add, so the resource concentration a punt provides is genuinely purchasable — and the categories with only 11–13 elite sources (steals, blocks) become impossible to win by accident. Punt in 12; consider it in 12 with 14+ roster spots almost mandatory; be sceptical of it in 10.

One 10-team caveat that cuts the other way: with fewer teams, each opponent is stronger on average, which raises the bar in every category and can make a concentrated build the only path to five wins against the top seeds. This is a real effect but a second-order one, and it does not survive the repairability argument in most leagues.

Open questions

• The binomial model assumes independent categories. Within a real matchup, a manager's categories are positively correlated through games played and minutes; the true breakeven above 0.6118 is probably somewhat higher, but the desk has not simulated it.
• The correlation matrix uses raw FG%/FT% rather than attempt-weighted impact, because most dossiers still lack projected FGA/FTA. Recomputing FG%/FT% correlations on impact scores could change the -0.59 bridges materially.
• No verified win-rate data exists for punt builds versus balanced builds at the league level. Every claim here is structural inference, not measured outcome — none of the sources surveyed publish success rates.
• The 10-versus-12 recommendation ignores roster-spot count, IL slots, and weekly transaction caps, all of which change repairability more than league size does.