Assessing a Coach - Process-Based versus Outcome-Based Judgment of Decisions

Track per-play success rate inside the red zone before you declare a hire a bust. Sean McDermatt’s crew just added John Fox to the room; the move only makes sense if you inspect how Buffalo’s short-yardage calls ranked 24th in EPA per rush last fall. https://salonsustainability.club/articles/bills-hire-former-super-bowl-hc-john-fox-to-staff.html

Look at the pre-snap motion rate and late-box identification speed instead of glancing at the standings. Matt Rhule posted a 11-27 record in Carolina, yet his 2021 offense manufactured 6.2 net yards per play when motion exceeded 55 %, top-five that season. The record stunk, the structure didn’t.

Owners fire staff after 8-9 campaigns, but the same clubs ignore adjusted sack rate spikes from 7 % to 3 % in twelve weeks. Keep the play-caller, swap the left tackle, watch the curve flatten. Wins lag behind process fixes by an average of 1.4 seasons since 2011.

Use win probability added on 3rd-and-medium and expected points on 4th-and-2 as the real report card. The score is just the receipt; the menu is where you spot the chef’s skill.

Map the 5-Phase Chain from Practice Design to Final Score

Start every micro-cycle with a 3-column spreadsheet: drill name, target speed (m·s⁻¹), and expected RPE. Anything lacking a number gets deleted.

Phase 1 - Blueprint: convert match analytics into 8-12 game-speed scripts, each 30-45 s long, rehearsed at 95 % peak heart rate. Track success rate with a stopwatch and a head-count; if less than 8 out of 10 reps finish inside the time gate, redesign the script before the next session.

Phase 2 - Rehearsal: run the scripts in 4-min blocks, 1:3 work-to-rest ratio. Clip a 10 Hz GPS unit on every player; flag any unit that drops below 85 % of the previous block’s average distance. Those athletes repeat the block while others move to corrective mobility. Outcome: 7 % rise in high-speed metres within two weeks, verified across 38 sessions in an U-19 Bundesliga squad.

Phase 3 - Feedback Loop: export the GPS file to a cloud sheet that auto-calculates a 5-variable load score (distance > 19.8 km·h⁻¹, accelerations > 3 m·s⁻², decelerations < -3 m·s⁻², PlayerLoad, and HRpeak). If the z-score of any variable exceeds ±1.5, the algorithm texts the staff within 30 s. Staff then shrink the next drill volume by 15 % or insert an extra recovery day. Over 11 matches this cut second-half goals conceded from 0.9 to 0.4 per game.

Phase 4 - Transfer: stage a 15-min chaos period at the end of training: scoreline flipped, one sub down, 20 % narrower pitch. Record every touch with a 25 fps camera; run Python code to count passes completed under pressure. The code pushes the metric to a Slack channel; if the weekly average dips below 78 %, the next tactical session swaps 10 min of pattern play for 3 v 3 + 2 rondos.

Phase 5 - Scoreline Link: align the load score with match-day expected goals (xG) difference. A Pearson r of 0.62 (n = 24 fixtures) shows that keeping the 5-variable load score inside the -0.8 to +0.8 band raises xG difference by 0.34. Action rule: if pre-match rolling 28-day load score drifts beyond ±1.0, start the player on the bench and limit him to 60 min. The chain ends when the final whistle confirms the scoreboard reflects the same upward slope plotted on Day 1 of the cycle.

Build a 7-Metric Dashboard Before the Season Starts

Lock the seven cells in a Google Sheet before the first practice: raw point differential per 100 possessions, turnover rate on first three touches after a rebound, corner-three frequency allowed, free-throw rate gap, bench minutes share, medical red-flag minutes lost, and training-load minutes above 85 % max heart rate. Each metric carries a 2026-24 league average baseline pulled from Cleaning the Glass, NBA.com, and Catapult CSV exports so the color scale turns red the instant your squad slips below the 25th percentile.

Point differential updates every quarter-hour via the league’s stats API using a 30-line Python script; set a 500-possession minimum so a single blowout does not repaint the trend line. Last year Orlando climbed from -7.1 to +2.4 in twelve weeks; they hit the playoffs because the dashboard pinged the staff when the second-unit bleed exceeded -9 for five straight games and forced a nine-man rotation tweak.

Turnover rate after rebound is scraped from Second Spectrum’s ZIP files; if the value tops 18 % for two weeks, the staff adds a three-man trap drill at mid-court the very next practice. Indiana did this in January, sliced the rate to 11 %, and turned the extra touches into 4.3 more transition points per night.

Corner-three defense is tracked by tagging every possession that ends within 6 ft of either corner; the goal is to keep attempts under 7 % of total rival shots. Milwaukee’s early-season spike to 9.4 % triggered extra weak-side help reps, dropping the figure to 5.9 % and saving 6.1 points per 100 on their slide to 110.3 defensive rating.

Free-throw rate gap (your trips divided by opponent trips) must stay above 1.15; fall below and the staff schedules a 20-minute draw contact small-side scrimmage. Utah’s youth roster started at 0.93, added the drill, and finished at 1.22, translating to +2.7 more points from the stripe each night.

Bench minutes share is auto-calculated from the play-by-play; aim for 38-42 % to keep starters under 32 min yet maintain rhythm. When OKC dipped to 29 % in November, fatigue spikes showed up in the heart-rate column; raising bench share back to 36 % cut the high-load minutes by 11 % and stabilized fourth-quarter efficiency.

Medical red-flag minutes and high-load minutes feed from the wearables CSV; set a combined alert threshold at 15 % of total player-minutes. Cross the line and the next practice drops to 75 % speed, no exceptions. Brooklyn kept the combo at 12 % all year and played 93 % of possible starter minutes, the healthiest mark in the East.

Run a Mid-Game Process Check with a 60-Second Clipboard Audit

Stop play, flip the sheet, mark four columns: shot zone, passer, screen angle, rebound result. Tick each possession for 60 seconds. After 20-25 possessions you own a snapshot that exposes hidden leaks long before the final horn.

Track only these four micro-stats; extras dilute clarity. A Women’s D-II squad boosted PPP by 0.18 after spotting that 62 % of their early-clock threes originated from the left slot with zero inside touch. Two practice days later the same slot produced lay-ups on 9-of-11 sets.

Checkpoint	Target	Red Flag	Fix in 1 Practice
Ball enters paint	≥60 % trips	<45 %	5-on-4 disadvantage shell, 7 straight scores before rotation
Close-out distance	≤0.9 m	>1.5 m	Mirror drill, sprint to 18 cm from cone, contest, sprint back
Off. rebound tag	3 bodies hitting	1 or 0	2-on-2 hit-and-find, coach pads ball off glass
FT differential	+6 per 10 mins	-4	Drop shoulder line, ref whistle every bump

Fold the sheet twice; the crease creates a vertical line. Left side logs your four metrics, right side keeps a running score. When deficit hits 8 or lead shrinks to 3, glance: if paint touches sit below 45 %, call horns-fist, ram the ball to short roll, force two defenders to commit, kick for corner three. You just converted data into points without a huddle.

Color code with a red pen for errors that chain-react: live-ball TO, surrendering an and-one, or leaving the strong-side corner untagged. One red mark equals one 30-second lane slide for the culprit in the next dead clock. Players police themselves; the clipboard never lies.

College men’s programs average 7.4 passes per possession when leading; only 4.1 when trailing. If your tally drops under 5 mid-game, insert a double-drag to force help, restart the pass count, restore flow. No speeches, just geometry.

End of quarter, rip off the top sheet, stash in back pocket. In the locker room you have 45 numbers, not 450. Circle the worst two, assign one 3-minute micro-clip, one 2-minute 3-on-3 constraint. Ten minutes later the squad re-takes the floor with the fix ingrained, not scribbled.

Convert Box-Score Noise into Coachable Micro-Actions

Tag every rebound with a shoulder-check timestamp; if the big glances over the wrong shoulder 0.4 s late, log it as "late-scan" instead of "DREB". Over a 14-game stretch, teams that cut late-scan frequency below 2.3 per game allow 4.7 fewer second-chance points.

Strip assists into two micro-stats: rim-assist (pass travels ≤3 m and shooter catches inside 2.4 m) versus swing-assist (pass travels ≥6 m and travels two zones). Rim-assists convert at 68 %, swing-assists at 39 %. Tell the playmaker to raise rim-assist share by 5 % and expect a +2.4 offensive rating bump.

• Record "early-crossover" when the ballhandler crosses mid-court with ≥12 s and ≤1 dribble; these yield 1.18 pts per possession.
• Log "late-crossover" at ≤6 s; efficiency drops to 0.87. Push early-crossover share above 60 % to add 5.6 pts per 100.

Track close-out distance at release: 0.9 m contests drop opponent eFG to 43 %, 1.8 m raises it to 54 %. Arm the wings with a wrist-strap sensor that vibrates at 1.5 m; within three weeks average contest distance shrinks 18 cm and opponent corner-3 accuracy dips 3.2 %.

Turn live-ball turnovers into "snap-to-pass" clips: freeze the frame 0.2 s after steal, measure how many outlet options are tagged. Elite transition teams average 2.8 tags; bottom-third sit at 1.4. Run 5-minute VR reps forcing the thief to find the third runner; snap-to-pass count rises 0.7 within a month.

Box-outs recorded only by result miss the point. Log "first-contact" time: when a player's hips hit the opponent's thigh. Every 0.1 s faster than 0.9 s after rim release adds 0.06 expected rebounds. Post players shaving 0.15 s reclaim 1.3 boards per night.

1. Export every contested rebound as a 3-dot clip: release frame, first-contact frame, possession frame.
2. Colour-code red if hips face the rim at first contact; green if hips face the opponent. Target 80 % green.
3. Send the clip to the player's phone within 30 s of subbing out; hit rate climbs 11 % in two weeks.

FAQ:

How can a club owner tell whether a coach is building something solid or just riding lucky breaks when the scoreboard keeps flipping?

Stop staring at the league table for a week and start counting repeatable actions. Is the team generating the same—or better—expected-goal difference every match? Are the same passing lanes open against both high and low blocks? Are minutes shared by U-23 players rising without results tanking? If those numbers hold steady while the goals bounce in or out off shins and VAR reviews, the process is intact. When the luck evens out, the points follow; if the metrics collapse, the coach was only surfing variance.

My kid’s youth side wins 8-0 every weekend, but they just boot the ball long and chase. Should I care if the score stays shiny?

Yes, because puberty is a brutal equaliser. A 12-year-old who never learns to receive under pressure or swap positions will be overtaken by smaller kids who did. Ask the coach for one training session where goals don’t count and see if he can still run a rondo. If he can’t, the scoreboard is lying about development.

Can a board really keep a manager after relegation if the underlying data say the squad improved?

They can if they sell the story to fans and investors before the drop, not after. Show the stadium graphs of shots conceded falling, show the wage bill shrinking, show the academy kids who now start. Relegation hurts, but parachute payments plus a coach who already raised the floor can return the club richer than a roulette hire who might bounce them straight back down.

Which single metric fools chairmen the most, and which one quietly warns them?

Points per game over the last six fixtures flatters or damns faster than any stat, yet it swings on two late penalties. Expected goals minus expected goals against hides in plain sight; if it stays positive while results rot, smart owners ignore the noise and keep the project alive.

Players say trust the process on Twitter while fans boo—how does a dressing room not split in half?

The coach posts the clips, not the scoreline, in the group chat. He shows the winger tracking 45 m to block a cross, the midfielder shielding the ball under a wall of whistles. When the squad sees their graft turning into arrows on the analyst’s laptop, they back each other instead of the jeers. The moment they stop believing those small wins matter, the process dies and the scoreboard never recovers.

Our high-school basketball team has lost four straight games, but the coach keeps running the same press-break and motion sets. Parents want him fired; the AD says trust the process. How long can a coach stick to a system before the scoreboard proves it’s wrong?

Four games is a blink; most schemes need at least a full season to show what they can do once players recognise cues without thinking. Ask the AD for two numbers: points the team gives up per possession after dead-balls and how many of those come from the press-break miscues. If both keep rising after 12-15 games, the scheme, not the kids, is the anchor. Until then, film every night, clip every turnover, and let the coach see the same reel the parents see. When he can’t explain three clips in a row, the scoreboard has spoken.