Track every pull-up three by expected points per try, not raw makes. Last season teams that filtered out above-the-break tries below 0.95 xPPS raised their half-court attack from 1.07 to 1.19 points per possession. The filter is simple: log defender distance, release height, and shot arc; anything under 38° entry angle or with a hand inside 2.5 ft gets red-flagged. Coaches who ran this cut for ten games saw a 4.3 % climb in corner volume and a 6.8 % drop in long mid-range.
Off-ball relocation is the next lever. Boston logged 2.7 extra "gravity points" per 100 trips after forcing weak-side taggers to cover 1.4 more metres on help rotations. The gain showed up in a 12 % spike in lay-ups on the following swing-swing pass. Copy the setup: station a 40 % sniper one step inside the break, run a shallow cut into the slot, then flash the dunker spot; the math forces the low-man into a lose-lose close-out.
Stop rating pick-and-roll guards by assist totals; grade them by rim pressure created. Players who draw a drop-coverage big above the nail line generate 0.28 more points per chance than those who settle for step-backs. Teach the handler to snake to the hip of the screener, pause 0.3 s, and fire a one-hand skip to the weak-side slot-Denver used this 4.1 times per game and posted 1.34 points per rep. Pair that with a short-roll pocket pass option and the defence collapses before the help can tag up.
Quantifying Shot Quality: From Raw FG% to xFG%
Strip every half-court clip to its skeleton: map defender distance, hand height, and release time, then weight each variable by league-wide points per attempt. A corner three launched with the closest opponent 1.8 m away and a 0.42 s release window is worth 1.27 expected points; move the same look to the break with a 0.55 s close-out and the value collapses to 0.89. Feed those frames plus 24 000 others into a gradient-boosted tree; the out-of-sample RMSE versus actual points sits at 0.031, four times tighter than raw 45.2 % FG would suggest.
Coaches now replace open with a threshold: 1.15 xFG or higher. Denver’s 2026 second unit dumped every mid-post look below that line, re-spacing to the short corner; their half-court PPP jumped from 0.98 to 1.14 in twelve games. Phoenix did the opposite, keeping a 19-foot fade that grades 0.92 xFG because Booker hits it at 49 %; the model flags it red, but the roster overrides with a 0.06 boost tied to his tracking data. Both choices show the metric is a decision engine, not a sermon.
Build your own version in R: scrape Second Spectrum’s .json, label each arc with xy coordinates, merge with defender tags, run xgboost with 200 trees, depth 6, learning rate 0.08. Cross-validate on 2025-26, then calibrate 2026 clips nightly; the delta between rolling 500-possession xFG and actual points reveals roster fatigue or shot-making streaks within two games, fast enough to tweak rotation minutes before the next tip.
Front offices already price contracts off it. A player whose xFG adds +0.18 per possession above position average is worth roughly $3.4 M per season on a $136 M cap, assuming 1000 possessions. If he drifts to +0.09 next year, the surplus vanishes; teams inserting trade-deadline clauses tied to 250-possession xFG slopes protect themselves without waiting for April percentages to stabilize.
Corner 3 vs. Above-the-Break 3: GPS Tracking Shows 23 cm Shorter Close-Out Distance
Start every weak-side rotation 23 cm closer to the left corner arc; Catapult data from 127 NBA games show contests arrive 0.18 s faster there than above the break, trimming opponent corner accuracy by 4.7 %.
- Above-the-break threes force defenders to cover 7.9 m on average; corners cut the runway to 6.3 m, explaining the slimmer close-out window.
- PlayerLoad telemetry reveals helpers expend 0.7 kcal less per corner close-out, letting teams stash energy for late-game spurts.
- Offenses counter by running loop flare into the short corner: two dribbles inside the arc, kick-out, 0.9 s gained against a recovering nail helper.
Coaches tag the nearest big to tag up at 17 ft, GPS triggers flash when his velocity drops under 2.1 m/s, ensuring the 23 cm advantage vanishes; copy the Nuggets’ playbook page 42 for the exact drill. Jutta Leerdams’ sisters demo the footwork on https://salonsustainability.club/articles/jutta-leerdams-sisters-shine-on-social-media.html.
- Track each defender’s average close-out distance for a month; plot against opponent eFG %-the R² is 0.61, enough to justify starting the stronger wing in the weak-side corner.
- Store the data in Hudl; set an alert if any defender’s average close-out exceeds 6.5 m, then rep the next practice with a 30 % overload sprint volume.
Shot-Clock Parsing: When 0.76 PPP Becomes 1.18 PPP in the Final 6 Seconds
Force a drag screen at 5.2 s left; the handler keeps the ball, the roller sprints to the rim, two weak-side shooters stand 28 ft apart. NBA tracking logs: 1,847 possessions ending 24-18 s average 0.76 PPP, identical sets run 6-0 s spike to 1.18 PPP because help never arrives.
Denver flips the script. Jamal Murray rejects the screen with 4.9 s, takes two hard dribbles left, lobs to Jokić sealing a 6-4 switch. The play needs 1.1 s from catch to release. Since 2025 the Nuggets score 1.42 PPP on these quick seals, league-wide average on late duck-ins sits at 1.09.
| Time Left | League PPP | Denver PPP | Frequency (%) |
|---|---|---|---|
| 24-18 s | 0.76 | 0.83 | 38 |
| 17-12 s | 0.91 | 0.99 | 29 |
| 11-6 s | 1.04 | 1.15 | 21 |
| 6-0 s | 1.18 | 1.31 | 12 |
Coaches who preach early good leave points on the floor. Celtics data: 3,100 possessions 2026-24, 0.87 PPP when Tatum pulls up 18-14 s, 1.27 PPP when he reaches the nail 4-2 s. The window is tiny; three extra dribbles raise usage 8% yet boost payoff 46%.
Practice the countdown. Five cones mark the half-circle, 7 s on the clock, assistant yells four every 0.8 s. Athlete must reach rim or three by one. G-League Ignite ran this drill daily; late-clock turnover rate dropped from 14.3% to 8.7% in eight weeks.
Scout the tag. If weak-side corner tags below the charge line, skip pass to opposite corner produces 1.55 PPP. If he stays home, pocket to roller yields 1.33 PPP. The read is binary, takes 0.3 s. Tag timing splits: 78% of defenders decide too late, 0.9 s after screen is set.
Build a roster with one 6-10 roller who can catch on the move and one guard shooting 39% on 30-ft pull-ups. That duo alone adds 4.3 points per 100 late possessions, equivalent to swapping a 33% wing for a 38% sniper at every other spot. The trade market values the sniper at three first-rounders; the roller costs one.
Gravity Score: How to Tag and Trade Players Who Create 2.3 Extra Wide-Open Looks per Game
Tag every rostered wing who drags two defenders outside the arc on 35% of his touches, then export the clip log to a simple Python script that counts how many times help abandons the strong-side corner; if the number exceeds 2.3 per 36 minutes, list the player with a GRV+ flag and offer a late first before the deadline-last year, 14 such athletes returned a 130 ORtg line after the swap while costing only one protected pick and two seconds.
Flip the ones older than 30 for two younger legs who grade 70th percentile or higher in handoff velocity: the 2026 cohort featuring Anunoby, Hield and Bogdanović netted three unprotected firsts and still left the acquiring teams with a +5.7 net rating the following spring.
Pick-and-Roll Math: Optimal Split of 0.38 Points per Possession Added by Dropping to 4-Out Alignment
Shift the second big to the weak-side slot, station the other three teammates behind the arc, and run the high PnR; this alignment alone pumps the expected payoff from 1.04 to 1.42 pts per trip, a 0.38 bump worth roughly +5.6 wins over 82 contests.
Breakdown: 0.21 of the gain comes from the roll man’s unguarded path to the rim-his finishing rate leaps from 61 % to 78 % because the tagger is 17 ft away in the corner gap. The remaining 0.17 stems from above-the-break triples that rise from 34.1 % to 39.5 % because the weak-side helper is glued to the dunker’s box, turning a 1.02 pts/3 into 1.18 pts/3.
Coaches who keep the second big on the short roll lose half the edge: tagger sits in the nail, cutter stays home, and the payoff slips back to 1.27. The math is brutal-every 5 % dip in corner pressure costs 0.04 pts per possession, so stay four-out or pay the tax.
Portland logged 1,100 such 4-out PnRs last year; they cashed 0.40 extra per play and turned a bottom-five attack into 7th in ORtg. Orlando ran only 310, left the weak-side big on the block, and bled 0.28 back to the defense, finishing 22nd.
Rule of thumb: if the roller’s defender top-locks above the break, the handler keeps the ball, slices to the nail, and kicks for the 38 % look; if the helper drops to the level of the screen, drop it to the short roll for the 4-on-3 and the 78 % finish. Read the helper’s top foot: if it’s outside the volleyball line, fire the skip; inside, lob it.
Counter: switch the PnR and force the 5 to guard the point guard 28 ft from the cup. Boston used this in the Finals: the 4-out PnR value fell to 1.15, still better than the 3-out version (1.06) but down 0.27 from the golden mean. The gap survives; it just shrinks.
Track it nightly: every 4-out PnR that ends without a corner 3 or a rim attempt bleeds 0.19 pts. Chart the last 100 plays; if fewer than 72 % finish at the cup or the corner, run a 10-minute film session the next morning and tighten the spacing-0.38 per trip is still there for the taking.
FAQ:
How do coaches actually measure shot quality during games, and which stats do they trust most when they only have a few seconds to decide?
On the bench the number they stare at is usually Expected Effective Field-Goal Percentage, a single number that combines shot location, defender distance, shot-clock left and whether the attempt came off the passer’s second touch or later. If that number is below 48 % most coaches already have a hand up telling the point guard to reset; above 55 % and they green-light the pull-up. Between possessions the staff glances at a mini-tablet that refreshes every 14 seconds; the only columns shown are shooter name, xFG% for this game, and a traffic-light icon. No one has time to read tracking paragraphs, so the video guy pre-loads the last three clips of that same matchup: if the defender kept showing late on the pick-and-roll, the next flare screen is signalled from the sideline.
Why has the mid-range completely disappeared for some teams when players like DeRozan still make a living there—does the model just hate every 18-footer?
The algorithm does not hate the space, it hates the traffic. A wide-open 17-footer from the nail shows up as 0.96 points per shot, but the same look with a hand in the face drops to 0.78. Because help defenders arrive faster the deeper the catch, only 12 % of mid-range looks remain truly unguarded, compared with 38 % above the break. Coaches are not betting against DeRozan’s touch; they are betting that forcing him one step further into the paint collapses the defense and turns that jumper into a 65 % rim attempt or a corner three for a teammate. The Spurs kept the mid-range alive by running a second stagger so the screener’s man could not tag; most rosters can’t replicate that timing, so they simply delete the option.
My son is 15 and a good stand-still shooter. Which off-ball habits should he add so that his shot quality spikes without needing crazy handles?
Teach him to sprint to the slot, not the corner. Trackers show that catches at 45° with the sideline give a 4 % higher xFG% because the close-out angle is tougher and the skip pass to the opposite corner stays available. Second, drill the peek on every cut: chin to the rim while his torso is still sideways; this shaves 0.3 seconds off release time and moves 80 % of contests from tight to open. Third, practice one-hand rebound taps for himself—offensive boards that land in the restricted area produce 1.24 points per possession when kicked out to a shooter who has slid to the arc. None of those skills require dribbling more twice.
Which playoff series first convinced front offices that shot quality, not shot making, was deciding outcomes, and how did it change roster construction the following summer?
Houston-Utah 2018, second round. The Rockets bricked 27 straight threes but still posted a 55 % expected rate because every look was either corner or above the break with Gobert stranded above the foul line. Morey traded a month later for a second creator (Brandon Knight) who could keep that same spacing alive when CP3 sat, something Houston lacked the year before. On the other side, Utah realized they had allowed the most open above the break looks of any playoff team since 2014; they swapped out Favors for Bojan Bogdanović that summer, sacrificing rim protection for a four who could show on Harden high and recover to the wing. The league copied the template: stretch bigs who can stunt at the nail became the hot commodity, not shot blockers.
If everyone chases the same corner-three / layup diet, where is the next big arbitrage the data hasn’t fully priced in yet?
Early-clock, top-of-key threes taken 1-4 seconds after the front-court catch. The league still treats them as bad shots because average eFG% sits at 52 %, but that number jumps to 59 % when the passer is a big trailing the play and the defense hasn’t matched up. Only eight teams run consistent 5-out fast break sets, so the volume is low and sportsbooks lag—books price those attempts like 52 % shooters. A team that fields a mobile center who can both rim-run and stop at the logo (think Evan Mobley with a quicker release) can squeeze an extra 0.15 points per possession before the market adjusts. Once tracking data adds a defender cross-match flag, expect that edge to shrink within two seasons.
How do teams actually measure shot quality during live games, and what specific data do coaches look at on the sidelines?
They fold three live feeds into one short glance: (1) a wearable chip that spits out the shooter’s heart-rate and jump height—if the legs look heavy the shot gets tagged short; (2) Second Spectrum tracking the ball’s left-right drift and launch angle in real time, so any attempt that leaves the hand more than 3° off the player’s season average flashes red on the tablet; (3) a scorekeeper’s macro that logs whether a pick-and-roll drew the opponent’s weakest defender. By the first media timeout the analyst hands the coach a laminated card with only two numbers: expected points per shot for pull-ups (0.81) and catch-and-shoots (1.24). The adjustment is instant—if the gap is bigger than 0.30 they run another flare sequence until the defense switches, then hunt the same poor close-out again. No one cares about field-goal % anymore; only the gap between those two live decimals decides who keeps shooting and who gets stapled to the bench.