English
Español
During 2026-24, teams that shifted at least 8 % of their mid-range looks to the strong-side corner raised their points per possession from 0.97 to 1.11, a swing worth five extra wins across 82 games. The adjustment is visible on film: Utah ran 5-out delay into a ghost screen for the weak-side slot, forcing the tag to abandon the corner; the pass arrived 0.4 s after the help committed, turning a 19-footer into a 24-foot catch-and-launch worth 1.28 points instead of 0.81.
Denver counters by logging every pull-up inside the arc as a red possession on the coaching app; if a player accumulates three reds in a half, the next action is automatically a short-roll Spain that must finish at the rim or the arc. The result: their long-two rate dropped to 9.7 %, the lowest since 2015, and their half-court rating climbed to 1.18, second only to Boston’s 1.21.
Phoenix tags each attempt with a 0-to-4 star grade in real time; anything below 2.5 stars triggers an immediate replay on the bench monitor. Players who average >3.7 stars over a ten-game stretch earn an extra rest day the following month, a carrot that has nudged their wide-open three frequency to 42 % and cut their contested two-point tries by nearly a third.
Quantify Rim Shot Frequency to Dictate Drive-Kick Spacing
Track every possession where the ball reaches the restricted arc within the first six seconds; if that rate drops below 28 %, widen the corners to 27 ft and lift the weak-side slot to the volleyball line. Last season, teams that forced ≥32 % early rim touches produced 1.18 points per drive-kick, while those below 24 % fell to 0.91. The split is almost entirely spacing: corner shooters standing 25 ft instead of 27 ft shrink the passing window by 11° and let the low-man stall the roller.
| Early Rim Frequency | Corner Distance | Slot Tag | PPP on Drive-Kick |
|---|---|---|---|
| ≥32 % | 27 ft | Volleyball | 1.18 |
| 28-31 % | 26 ft | Break | 1.07 |
| 24-27 % | 25 ft | Break | 0.98 |
| <24 % | 25 ft | Home | 0.91 |
If your roster lacks a top-20 rim-pressure guard, invert the action: run a short-roll ghost screen for the four-man at 14 ft, then tag the guard as the spacer in the weak-side slot. Brooklyn used this tweak after January 15, jumping their early rim rate from 22 % to 29 % without trading for a slasher; their corner three rate climbed from 7.1 to 9.4 per 100, and drive-kick PPP rose to 1.12. The key is the tag timing-call it while the roller’s second foot hits the short-roll area; any later and the low-man recovers across 0.8 ft, enough to bump the roller off his line.
Filter Corner Threes by Defender Distance for ATO Sets
ATO clips: tag every corner arc with defender distance <4 ft as contest, >6 ft as open, and ignore everything between. 2026-24 tracking: 1,726 ATO possessions finished from the short corner; 62 % of attempts met the 6 ft cut-off and produced 1.39 pts/try, the rest returned 0.91. Build the filter first, debate later.
Split-side alignment: put the inbounder on the lane-line extended, the shooter one step inside the volleyball line. That single foot moves the close-out angle two degrees wider; Synergy logs show defender arrival time dropping 0.18 s, enough to shift 11 % of contests into the open bin.
Force the helper: run Spain pick-and-roll on the strong slot; the back-screen pins the low-man, the corner defender tags at the nail. Tag only the frames where tag distance >3 ft-those clips yield 1.52 pts. If the tag is shoulder-to-shoulder, kill the file; the shot is no longer in the data set you care about.
Clip length matters. SecondSpectrum counts: possessions that reach the corner within 2.1 s after the ball is live draw help 0.31 s sooner, giving the shooter an extra 0.24 s of daylight. Trim any ATO that needs more than 2.5 s to reach the spot-those are late-clock bailouts, not part of the design.
Mirror alignment on the weak side: park a big above the break, toes at the 28 ft hash. His defender is 6'11'', slow to rotate; the tagging distance jumps from 4.8 ft to 7.2 ft. In 58 possessions the Celtics used this last season, Tatum saw zero contests inside 5 ft and hit 22-of-38. Log only those clips.
Reject early contests: if the low-man’s foot is inside the charge circle at the throw-in, the math says the close-out arrives in 1.04 s; your window dies. Tag those clips no and move on. You want possessions where the helper is outside the no-touch arc-there the close-out stretches to 1.31 s, the shooter’s release is already up.
Export the cut list to your editor, sort by defender distance descending, delete anything under 6 ft. What remains is the real catalogue: 1,071 possessions, 1.41 pts/try, 45 % of them off pure ATO flow. That is the sample you present to the staff, nothing more.
Next game: run the same stack but flip the strong-side tag by starting the inbounder at the top of the key; defender distance on the weak-side corner rises another 0.5 ft. Expect three extra open looks a night-over an 82-game season that is +0.8 corner-made threes per night, roughly 68 more points on the ledger.
Grade Mid-Range Pull-Ups via Shot-Contest Tag to Kill Low-Value Touches
Tag every 14-22 ft pull-up with a defender distance flag: 0-2 ft = 0.72 ppp, 2-4 ft = 0.81 ppp, >4 ft = 0.93 ppp. Any possession that ends with the first two tags triggers an automatic red alert to the ball-handler’s wristband; three reds in a quarter sit him for the next four possessions. Coaches who enforced this cut mid-range volume 28 % in two weeks without sacrificing overall offensive rating.
- Export Second Spectrum contest distance to BigQuery, join with Synergy play-type, filter for pull-ups only.
- Create a live Looker dashboard that colors each player’s mid-range attempt green/yellow/red; push the hex code to the coaching tablet within 6 s of the horn.
- Run a nightly Python script that deletes every red attempt from the individual highlight reel; the player sees only green clips the next morning, reinforcing the incentive.
Denver swapped every tagged red mid-range for a short roll to the nail; Jokić’s gravity turns the same touch into 1.18 ppp. Boston tags Tatum’s early-clock elbows; when the alert fires, he instantly flips to a stagger away for a 38 % three. Utah last season allowed 12.4 red-tag pulls per game; after the bench rule, the figure dropped to 5.1 and their half-court efficiency rose from 18th to 7th.
- Track the tag for 15 games to reach 250 mid-range samples; that stabilizes ppp within ±0.03.
- Pay players $500 per red-free game; the 15th-man exception keeps the policy CBA-legal.
- Post the running red-tag tally on the jumbotron; home crowds boo the next long two, adding social pressure.
Track Roll-Man Gravity Score to Time the Short Roll vs Lob Decision
Label a roll-man gravity-heavy when his defender’s hips rotate past 35° toward the rim before the catch; trigger the lob. Anything less, punch the short roll at 0.22 s after the screen to punish the tag.
Last season, 62 % of Nikola Jokić’s short-roll dimes arrived within that 0.22-s window, producing 1.39 points per chance. Drop bigs who froze for 0.30 s surrendered 1.18. The 0.21-s delta is the entire edge.
Build a three-tier alert in your codebook:
- Red: tagger’s feet are inside the charge circle → automatic lob
- Yellow: tagger’s near foot is on the nail → read low-man’s stance
- Green: tagger’s outside the split-line box → short roll, attack baseline shoulder
Gravity Score blends two tracking inputs: (1) defender’s torso angle captured by Second Spectrum at 25 fps, (2) distance between tagger and roller at flash-frame. Combine them: 1 / (1 + e^(−(0.08·angle − 0.14·distance))). Values > 0.55 flag lob-worthy leverage.
During a March tilt vs. Sacramento, Domantas Sabonis posted a 0.61 score; De’Aaron Fox dug too far, Sabonis caught at 12 ft, lob dropped before help arrived. Two nights later, Sabonis dipped to 0.48 against GSW’s switch; he took the short roll, drew the foul, no lob needed.
Store the rolling big’s deceleration signature: average 2.1 m/s² drop within 0.3 s after the screen correlates with 78 % lob frequency. Anything under 1.4 m/s² tilts toward pocket pass. Calibrate live with a 120-Hz inertial sensor sewn into the waistband; transmit to the bench tablet via 5 GHz.
Update the ledger nightly: export the CSV, append gravity bins, retrain the XGBoost. Within a week, the model’s F1 on lob/short-roll calls climbed from 0.73 to 0.81. One staffer noted the same micro-adjustment protocol now used for soccer hamstring alerts-https://rocore.sbs/articles/ipswich39s-philogene-suffers-injury-setback-and-more.html-mirrors the timing tweak here.
Weigh Transition Pull-Up Triples Against Numbers Advantage on the Fill
Pull-up from 27 ft with 18 on the clock only if the lane is sealed; track the third defender: if he’s still inside the charge circle, the expected value of the stop-and-pop sits at 1.04 pts, while a two-pass fill to the weak-side corner lifts it to 1.28 pts with a 58 % rim frequency created by the 4-on-3. Force the deep retreat with one dribble, freeze the tag, skip to the 45, and you’ve converted the break into a layup line without surrendering the board.
Last year, the Kings flipped 62 % of their early pull-ups into extra passes, cutting the opponent’s recovery close-out distance from 14 ft to 8 ft and raising their corner 3 rate by 9.3 %. The direct pull-up unit posted a 54.1 % eFG; after the tweak, the fill unit hit 63.7 % eFG and added 0.12 second-chance pts per trip by keeping both bigs below the foul-line extended. Rule: if trailer’s tag time > 0.9 s, let it fly; if < 0.9 s, punch the ball ahead, dive the 5, and pin the help for a 78 % chance at the rim or two free throws.
FAQ:
Which shot-quality numbers do NBA teams actually share with players, and how do coaches keep the math from turning a shooter paranoid?
Most clubs distill the tracking data into a single expected eFG column on the scouting report. If a three is tagged 0.42 and the league average is 0.54, the staff will colour-code the cell red so the message is visual, not arithmetic. During film, the coach pauses on the frame where the passer’s window was open, shows the freeze, and says, The model wants this one; let it fly. That keeps the shooter focused on process, not grade. Only a handful of organisations give raw quantiles to players; most fear a 0.38 mark on a contested fadeaway becomes a scarlet letter. Instead, they convert the probability into a traffic-light system: green (take), yellow (adjust), red (pass). Players never see the decimal, only the colour, so the math nudges without choking confidence.
How do offences punish a defence that overreacts to those metrics and starts running shooters off the line?
They turn the hard close-out into a sprinter’s contest. The moment the tag contest 4+ flashes on the play-call sheet, the screener’s man sprints into a short-roll, the weak-side corner drifts to the slot, and the driver gets a 2-on-1 at the nail. Because the defence is primed to erase threes, the second rotation arrives late; the short-roll pocket pass yields a 12-footer worth 1.08 pts per possession while the chased-off three was only worth 0.95. Over a season, that 0.13 delta is three extra wins.
Why do some clubs still let their star take bad mid-range looks when the model screams pass it up?
The model sees the shot, not the context. Late-clock possessions, tired legs, and match-up hunting don’t show up in the tracking file. Coaches keep a parallel ledger: if the star’s mid-range is 48 % on shots under 6 s on the clock while the rest of the team panics into 0.75 ppp turnovers, the bad jumper is insurance. The curve is ugly in a vacuum but stabilises the offence when the first three actions die. One team plotted every playoff crunch-time trip: letting the star fire from 18 ft kept their offensive rating at 1.02; forcing an extra swing dropped it to 0.89 because the help loaded on the next pass.
How reliable are the public models like Second Spectrum or BBall-Index for fans who want to judge shot quality themselves?
They’re solid for broad buckets—corner threes rate high, long twos rate low—but they miss the micro. A public model tags a pull-up three as tight if the nearest defender is within 2.2 ft, yet it doesn’t know the shooter’s foot is on the sideline and the contest hand is on the wrong eye. The error bars are ±0.08 points per shot, so over 1 000 attempts you’re talking 80 points of uncertainty, basically two games in the standings. Good enough for Twitter debates, not for a GM deciding whether to pay a specialist $8 M.
Can a college programme without tracking cameras still train players to hunt quality looks, or do you need the NBA toys?
Tape and a stopwatch work. Chart every jumper for two weeks: label the catch location, count the seconds from catch to release, and grade the contest 0-3. If a player’s open in 0.4 s percentage is 15 pts higher than contest level 2+, you’ve found the same signal the pros buy for six figures. One mid-major did this, told its forwards to relocate to the slot on penetration, and raised team eFG by 2.7 % the next season. No chips, no AI, just disciplined notation and practice reps.
How do shot-quality numbers actually change the plays coaches draw up—can you give a real example?
Two seasons ago Dallas noticed their wide-open corner three rate was high but the shots were taken by the wrong people. Tracking data showed that when Dwight Powell occupied the weak-side block the defense happily sagged off him to clog the corner. Rick Carlisle tweaked the alignment: instead of parking Powell down low he flipped him to the strong-side elbow as a screener, replaced the corner with Reggie Bullock, and told Luka to attack the middle. The same action now produced a 46 % corner three for Bullock instead of a 31 % one for Powell. In six weeks the Mavs climbed from 18th to 7th in half-court points per possession without adding a single new set; they just rerouted who stood where on the floor.