The Fantasy Basketball Encyclopedia

Strategy

Points leagues versus category leagues

Points and category leagues price the same box score with different mathematics β€” points is a linear functional on eleven stats, 9-cat is nine independent binary contests β€” and the practical consequence is that punting is unavailable in points, availability dominates it, and no two platform "points" defaults are the same formula.

Up: 2026-27 Fantasy Basketball Encyclopedia

Related: 2026-27 points-league rankings Β· 2026-27 category rankings Β· 2026-27 fantasy basketball rankings Β· 2026-27 fantasy basketball tiers Β· 2026-27 fantasy basketball player projections Β· 2026-27 NBA minutes projections Β· 2026-27 fantasy basketball projection methodology Β· 2026-27 fantasy basketball injury and availability ledger Β· 2026-27 fantasy basketball evidence and confidence policy Β· 2026-27 sleepers, breakouts, and fades Β· Games played versus per-game value in fantasy basketball Β· Robust punt builds in fantasy basketball Β· Percentage categories and volume

The five defaults, verbatim

Everything below rests on this table. "Points league" is not a format; it is a family of formats that disagree with each other.

Stat ESPN default Yahoo default CBS Free Sleeper standard Fantrax / RotoWire "standard"
PTS +1 +1 +1 +1 +1
REB +1 +1.2 +1 +1.2 +1.2
AST +2 +1.5 +2 +1.5 +1.5
STL +4 +3 +1 +3 +3
BLK +4 +3 +1 +3 +3
TO βˆ’2 βˆ’1 βˆ’1 βˆ’1 βˆ’1
3PM +1 β€” β€” +1 (default on, value league-set) β€”
FGM +2 β€” β€” β€” β€”
FGA βˆ’1 β€” β€” β€” β€”
FTM +1 β€” β€” β€” β€”
FTA βˆ’1 β€” β€” β€” β€”

Sources: ESPN's default points-league scoring explainer Β· Yahoo default league settings Β· CBS Free official rules Β· Sleeper's scoring primer and Sleeper's scoring-settings list Β· RotoWire's standard formula.

Two caveats worth carrying. Sleeper publishes which settings are on by default (points, rebounds, assists, steals, blocks, turnovers, 3PM, double-double, triple-double, technical and flagrant fouls) but does not publish the numeric defaults in its help centre; the values above are the ones Sleeper's own primer calls standard, and you should read them off the league-creation screen rather than trust this row. Fantrax exposes scoring per league in the League Rules Summary and does not publish a single canonical default page that survives a fetch; the row above is the value Fantrax's own editorial and RotoWire both call standard, not a verified product default. Both rows are lower-confidence than the ESPN, Yahoo, and CBS rows, which come from official product documentation. Note also that a sixth family exists β€” DFS-style scoring such as FantasyData's 3/2/1 shot-value system with STL and BLK at +2 β€” which is not a season-long default anywhere but shows up in third-party projection exports.

CBS Free is the outlier that nobody flags: steals and blocks at +1 collapse the entire defensive-specialist archetype, and CBS averages a player's score across games in the scoring period rather than summing.

ESPN's formula is not "points plus extras" β€” rewrite it

ESPN's eleven terms look intimidating and get misquoted constantly. Collapse them. Since \(2\cdot\mathrm{FGM}-\mathrm{FGA}=\mathrm{FGM}-\mathrm{FG_{miss}}\) and \(\mathrm{FTM}-\mathrm{FTA}=-\mathrm{FT_{miss}}\):

\[ \mathrm{ESPN}=\underbrace{\mathrm{PTS}+\mathrm{3PM}+\mathrm{FGM}-\mathrm{FG_{miss}}-\mathrm{FT_{miss}}}_{\text{shooting}}+\mathrm{REB}+2\,\mathrm{AST}+4\,\mathrm{STL}+4\,\mathrm{BLK}-2\,\mathrm{TO} \]

Per shot event that is: made three = +5, made two = +3, made free throw = +1, any miss = βˆ’1. ESPN says exactly this in its explainer.

The useful question is not "does ESPN punish volume" β€” in isolation it barely does. A two-point attempt is worth $4p-1$, positive above 25%; a three is $6p-1$, positive above 16.7%. The useful question is ESPN minus Yahoo per attempt, because that is the rank-swing driver:

Shot type ESPN value Yahoo value ESPN βˆ’ Yahoo Break-even
Two-point attempt $4p-1$ $2p$ $2p-1$ FG% = 50.0%
Three-point attempt $6p-1$ $3p$ $3p-1$ 3P% = 33.3%
Free-throw attempt $2p-1$ \(p\) $p-1$ FT% = 100%

Read the third row again. ESPN taxes every free-throw attempt by \((1-\mathrm{FT\%})\). There is no free-throw volume that is neutral. Luka DončiΔ‡ shot 10.1 FTA at 78.0% in 2025-26 β€” that alone is βˆ’2.2 FP/G in ESPN relative to Yahoo, the largest FT penalty in the elite pool.

Worked example: 2025-26 lines, both formulas

All inputs are real 2025-26 regular-season per-game stats from ESPN. Team labels are 2025-26; dossier links go to 2026-27 rosters, so several players changed teams in the offseason.

Player 25-26 Tm GP FG% (FGA) FT% (FTA) ESPN FP/G rank Yahoo FP/G rank swing
Nikola Jokić DEN 65 .569 (17.4) .831 (7.4) 66.2 1 62.1 1 —
Luka DončiΔ‡ LAL 64 .476 (22.8) .780 (10.1) 58.8 2 57.5 2 β€”
Shai Gilgeous-Alexander OKC 68 .553 (19.4) .879 (9.0) 55.8 3 50.6 4 +1
Victor Wembanyama SAS 64 .512 (16.9) .827 (7.0) 55.3 4 53.4 3 βˆ’1
Trey Murphy III NO 66 .470 (15.9) .886 (3.7) 40.7 16 37.9 23 +7
Jaylen Brown BOS 71 .477 (21.7) .795 (7.5) 43.8 13 45.2 8 βˆ’5
Bam Adebayo MIA 73 .442 (15.7) .778 (5.8) 39.4 22 41.0 15 βˆ’7
Mikal Bridges NY 82 .490 (11.7) .827 (1.2) 33.6 30 29.8 33 +3
Jalen Duren DET 70 .650 (11.5) .747 (6.1) 38.6 23 38.0 22 βˆ’1
LaMelo Ball CHA 72 .407 (17.3) .899 (2.5) 39.7 19 38.2 21 +2

Ranks are within a 37-player pool of 2025-26 top-100 scorers, so they are relative, not draft ranks. Across that pool the two formulas correlate at Spearman ρ = 0.974, Pearson r = 0.986 and the largest rank swing is 7 slots. The honest headline is therefore: the formulas mostly agree, and the disagreement is worth about half a round. Anyone who tells you a Yahoo board is useless in an ESPN league is overselling; anyone who tells you they are interchangeable is costing you a pick.

Decomposing ESPN βˆ’ Yahoo into its six sources makes the archetypes explicit (units: FP/G):

Player shooting FT tax REB (βˆ’0.2Γ—) AST (+0.5Γ—) STL (+1Γ—) BLK (+1Γ—) TO (βˆ’1Γ—) net
Shai Gilgeous-Alexander +3.9 βˆ’1.1 βˆ’0.9 +3.3 +1.4 +0.8 βˆ’2.2 +5.2
Mikal Bridges +1.8 βˆ’0.2 βˆ’0.8 +1.8 +1.3 +0.8 βˆ’1.0 +3.8
Trey Murphy III +2.3 βˆ’0.4 βˆ’1.1 +1.9 +1.5 +0.4 βˆ’1.8 +2.8
Luka DončiΔ‡ +2.8 βˆ’2.2 βˆ’1.5 +4.2 +1.6 +0.5 βˆ’4.0 +1.3
Alperen ŞengΓΌn +1.2 βˆ’1.6 βˆ’1.8 +3.1 +1.2 +1.1 βˆ’3.2 0.0
Karl-Anthony Towns +1.5 βˆ’0.7 βˆ’2.4 +1.5 +0.9 +0.5 βˆ’2.5 βˆ’1.2
Jaylen Brown +1.1 βˆ’1.5 βˆ’1.4 +2.5 +1.0 +0.4 βˆ’3.6 βˆ’1.4
Bam Adebayo βˆ’0.2 βˆ’1.3 βˆ’2.0 +1.6 +1.2 +0.7 βˆ’1.6 βˆ’1.6

The winners in ESPN are efficient, low-turnover, three-point-volume wings and guards (Mikal Bridges at 11.7 FGA/1.2 FTA/1.0 TO is almost purpose-built for it). The losers are volume bigs who shoot under 50% from the floor and rebound for a living β€” Bam Adebayo and Karl-Anthony Towns both go negative, because ESPN's rebound weight is 1.0 instead of 1.2 and their shooting term stops subsidising them. The intuition "ESPN rewards efficient bigs" is only half right: it rewards efficient bigs like Jalen Duren (+3.5 shooting) and punishes inefficient high-usage bigs, and the rebound downweight partially cancels the gain for both.

Steals and blocks: the format's real fault line

In points, a steal is 3 or 4 fantasy points and nothing more. Dyson Daniels led the NBA with 2.0 STL/G in 2025-26 and finished 28th of 37 in Yahoo FP/G in the elite pool above. His league-leading skill bought him 6.0 FP β€” about 17% of his own line, and roughly 3.6 FP of margin over a 0.8-STL replacement. That is a tenth of a starter. Points leagues do not care that the steals are scarce; they only care about the coefficient.

Nine-cat cares about scarcity, because each category is a separate binary contest and aggregation compresses extremes. Within the same 37-player pool:

Quantity mean SD max max z
STL/G 1.11 0.37 2.00 (Daniels) +2.37
BLK/G 0.79 0.61 3.10 (Wembanyama) +3.78
PTS/G 20.98 5.80 33.5 (DončiΔ‡) +2.16
Yahoo FP/G 39.62 8.58 62.1 (Jokić) +2.62

(These SDs come from a 37-player convenience sample of high-minute scorers and are narrower than a true 150-player rosterable pool would produce; the ordering is the point, not the decimals.)

Victor Wembanyama is +3.78σ in blocks. No player is +3.78σ in fantasy points — the best in the pool is Jokić at +2.62σ, because a sum of eleven imperfectly-correlated terms always has lower relative variance than its most extreme component. That is the whole format difference in one line: 9-cat lets a single roster slot be more extreme than any player can ever be in aggregate, and lets that extremity decide a contest outright. A points league cannot express that, by construction. It is also why CBS Free's +1 stocks is a materially different game from Yahoo's +3, and why ESPN's own format explainer frames category leagues as rewarding "diverse collection of statistics" while points leagues make accumulation king.

Turnovers

Nine-cat treats TO as a ninth category you can simply concede β€” punt-TO builds are a standard archetype and cost nothing beyond one of nine slots. Points leagues make TO a permanent linear drag that follows the player everywhere. The size of the drag differs by a factor of two across platforms: DončiΔ‡'s 4.0 TO/G costs βˆ’4.0 on Yahoo and βˆ’8.0 on ESPN. Between an ESPN and a Yahoo league, his turnovers alone move 4.0 FP/G β€” more than the entire ESPN-vs-Yahoo net differential for any player in the table above. High-usage creators (Cade Cunningham 3.7, Jaylen Brown 3.6, Alperen ŞengΓΌn 3.2) are the ones this hits, and on ESPN it is partly repaid by the doubled assist weight β€” Cunningham's +4.9 assist gain outruns his βˆ’3.7 turnover loss.

Punting does not exist in points leagues

This is structural, not stylistic. Punting works because in H2H categories a 5–4 win counts the same as an 8–1 win, so you can trade a guaranteed loss in FT% for a decisive edge in five others. Formally: 9-cat scoring is a sum of nine step functions of category margins, so margin above the win threshold is worthless and can be reallocated. Points scoring is a single monotone linear functional \(\sum_j w_j x_j\) β€” there is no threshold, no wasted surplus, and no category you can concede for a rebate. Conceding a stat in points is simply scoring fewer points.

What replaces it is volume, minutes, and games played, in that order, plus a lineup-slot discipline that has no 9-cat equivalent:

Games played: why it dominates points and merely matters in 9-cat

Nine-cat H2H normalises much of the schedule automatically, because per-game rates decide FG%, FT%, and ratio contests regardless of volume, and roto leagues typically cap games at 82 per position (Yahoo, ESPN) β€” Yahoo's H2H default carries no games cap at all. Points leagues sum an unbounded total. The result is that availability is not a tiebreaker, it is a primary term.

Player GP Yahoo FP/G per-game rank Season FP season rank
Anthony Edwards 61 44.0 11 2687 20
LaMelo Ball 72 38.2 21 2751 18
Mikal Bridges 82 29.8 33 2755 17
Desmond Bane 82 33.7 29 2761 16
Scottie Barnes 80 42.1 13 3364 6

Desmond Bane was the 29th-best per-game player in this pool and out-produced Anthony Edwards, the 11th, over the season. Mikal Bridges β€” 33rd per-game β€” matched LaMelo Ball exactly. In 9-cat those same three players are not remotely equivalent, because Bridges wins you nothing and Ball wins you 3PM and AST. This is why 2026-27 points-league rankings uses a draft index of \(\mathrm{FP/G}\times\sqrt{\mathrm{GP}/82}\) rather than pure per-game value, and it is the single largest reason a points board and a category board should never be the same list. Cross-check every points pick against the 2026-27 fantasy basketball injury and availability ledger.

Efficiency: ratio in 9-cat, linear in points

The mathematical difference is real and produces reversals, not just re-weightings.

Nine-cat FG% is a team-level ratio \(\sum_i m_i / \sum_i a_i\). A player's marginal impact is approximately

\[ \Delta \approx \frac{a_i\,(p_i - P)}{A + a_i} \]

β€” attempts multiplied by the gap from your team's rate. Three consequences: it is volume-weighted (a 67% centre on 4 FGA is nearly inert), it is context-dependent (the reference \(P\) is your own roster, so the same player helps a bad-shooting team more), and it is non-additive (you cannot pre-compute a player's FG% value in isolation). See Percentage categories and volume.

ESPN points is \(\mathrm{FGM}-\mathrm{FG_{miss}}\): an absolute, context-free, additive constant per player. No reference rate, no teammates, no punt.

The reversal case is free throws. Compare 2025-26:

Player FTA FT% FT misses/G ESPN FT term 9-cat impact vs .780 ref
Zion Williamson 7.5 .716 2.13 βˆ’2.13 βˆ’0.48
Evan Mobley 4.6 .606 1.81 βˆ’1.81 βˆ’0.80

ESPN says Zion Williamson is the worse FT liability (he misses more shots). Nine-cat says Evan Mobley is roughly 1.7Γ— worse (his rate is far further from reference). Both are correct within their own arithmetic, and they disagree about which player to avoid. The general rule: ESPN prices raw misses, 9-cat prices the rate gap times volume. Low-volume terrible shooters (Dyson Daniels, 1.6 FTA at 61.5%) are close to harmless in ESPN points and only mildly damaging in 9-cat; high-volume mediocre shooters are the ones ESPN quietly hates.

Critical finding: this vault's two "points" boards are not the same formula

2026-27 points-league rankings is computed on the Yahoo formula β€” \(\mathrm{PTS}+1.2\,\mathrm{REB}+1.5\,\mathrm{AST}+3\,\mathrm{STL}+3\,\mathrm{BLK}-\mathrm{TO}\) β€” and states so explicitly. Meanwhile a large number of individual player dossiers carry a bolded "ESPN-default points:" verdict line, and many cite ESPN's April 2026 way-too-early points board as a market anchor. These are different scoring systems and the verdicts are not interchangeable.

Worse, the discrepancy is currently uncloseable from the vault's own data. In Data/player-projections-2026-07-18.csv, of the 262 rows that clear the 20 GP / 15 MPG screen, only 54 carry fga, fta, fg_pct, and ft_pct β€” the four fields ESPN's formula requires. Coverage is not random; it is by team desk:

Desks with complete shot data Desks with none
DAL (11/11), MEM (11/11), DEN (8/8), MIL (8/8), PHX (8/8), LAL (7/7), FA (1/1) the other 24 teams, 0 rows each

So no ESPN-default board can be produced for 79% of the projection pool, and the 54-player subpool that can be scored is a documentation artifact, not a talent tier. Within that subpool the Yahoo and ESPN draft indices correlate at ρ = 0.993 with a maximum swing of 5 slots β€” reassuring, but it is a sample of six teams' rosters and cannot be extrapolated.

Where the discrepancy bites hardest, based on the archetype analysis above:

Remediation: the 24 team desks should backfill fga/fta/fg_pct/ft_pct into the projection export before draft season, at which point a genuine ESPN-default column can sit beside the Yahoo one. Until then, every dossier line that says "ESPN-default points" should be read as a directional claim, and the board rank next to it as a Yahoo number.

The decision, for a 10-or-12-team autumn draft

If you want the draft to be decided by preparation and in-season management, take 9-cat: punt architecture, ratio management, and scarcity pricing give a prepared manager far more edge surface. If you want the season decided by roster throughput and injury luck, take points β€” it is a simpler, higher-variance, more football-like game where the winning skill is squeezing games played out of thirteen roster slots. And whichever you pick, read the scoring settings screen before you rank anything: the CBS-versus-Yahoo stocks gap (+1 vs +3) is a larger distortion than the entire ESPN-versus-Yahoo differential analysed above.

Open questions