Pitch to the Pros 3

Was the Best Pass
Really Playable?

A modeling framework for adding decision-window context to pass value.
Ryan Inghilterra headshot
Ryan Inghilterra Senior Data Scientist @ CrowdStrike
The pass moment

The best pass depends on when you look.

Simplified timing read Same option, one second apart
1.0s before pass P R
Pass moment P R
The missing layer is time

Pass options change inside
the decision window.

Player A in possession before pass actual pass to D
Pass to B
playable early, harder later
Pass to C
valuable, hard to play throughout
Pass to D
becomes playable late
Review question: when was each option playable
before the actual pass?
The approach

Turn pass options into decision-window context.

1Start with pass candidate optionsreceiver, value, option window
2Score execution contextinside the decision window
3Return review contextbest frame, last playable, primary constraint
Algorithm pass-window review
for each pass event:
define pass frame + carrier window
options = candidates before pass
for each option:
score active frames
trace playable value over time
keep peak + pass-frame value
return evidence:
ranked pass options
playable value + timing status
Output: top alternatives, playable timing, and what changed before the pass.
Playability calculation

At each frame, five playability components produce the score.

LaneIs the path open?S_lane
PressureCan the carrier execute?S_pressure
ReceiverIs the target free?S_receiver
MovementDoes the run fit?S_movement
TimingIs the option still alive?S_timing
Playability score
Splay(o,t) = geometric mean ( Slane, Spressure, Sreceiver, Smovement, Stiming )
Each component is a 0-1 tracking-derived score. One severe constraint pulls the full score down.
TheoryVtheory = SkillCorner xThreat
Playable valueVplay(o,t) = Vtheory(o) * Splay(o,t)
Decision frictionFdecision(o,t) = Vtheory(o) - Vplay(o,t)
Real match data

Built on open SkillCorner tracking and event data.

10Australian A-League 2024/25 matches
9.6Mplayer-frame tracking samples
47.9Kdynamic events
24.4Klinked passing options
How the data is used

SkillCorner passing options define the candidate actions. SkillCorner xThreat estimates each option's potential value. Tracking supplies the geometry for lane access, pressure, receiver space, movement fit, and timing.

Example 1 / Snapshot

Start with the pass snapshot.

Howieson completes the safe pass.
Moreno is the higher-value option to review.

Howieson selected pass with reviewed higher-xT Moreno option at the pass frame
Review question Higher value is only the start. Was Moreno playable in time?
Example 1 / Timing read

The better window came earlier.

Moreno's cleanest look came 1.7s before the pass. By release, the same lane had tightened.

Moreno option 1.7 seconds earlier compared with the same option when Howieson passed
How this frame was chosen identified from option-window data
Option window 48 frames appeared 4.8s before pass
Opportunity frame S_play 0.94 xT 0.034 at the earlier frame
Pass-frame audit S_play 0.81 same option, 1.7s later
Visible change lane 1.00 -> 0.36 playability 0.94 -> 0.81 clean route became contested
Timing read 1.7s opportunity frame to pass
Example 2 / Snapshot

Start with the pass snapshot.

Ishige is the higher-value option on paper.
At release, playability has already collapsed.

Nagasawa selected pass to Wootton with reviewed higher-xT Ishige option at the pass frame
Pass-frame verdict Ishige xT 0.019 playability 0.00
Pressure0.00
Lane0.16
Receiver0.52
at release; rewind to test whether it was ever fair
Example 2 / Constraint read

No clean window appeared.

Ishige had value on paper, but pressure was already limiting the option before release.

Ishige option 1.2 seconds earlier compared with the same option when Nagasawa released the pass
How this frame was chosen identified from option-window data
Option window 31 frames 3.1s before release
Review frame S_play 0.53 xT 0.019, already constrained
Pass-frame audit S_play 0.00 same option, 1.2s later
Pressure never cleared 0.11 -> 0.00 lane 0.73 -> 0.16 the constraint was present, then fully collapsed the final-frame demand
Fairness read not a clean demand the completed Wootton pass becomes more justified
Analysis automation system

From match data to automated review workflows.

Inputs
tracking frames
events + options
analyst question
Metric pipeline
option windows
playability by frame
visuals + scores
review store
AI automation
case discovery
query generation
report generator
Analyst outputs
scenario reports
aggregate metrics
grounded answers
AI automation runs on stored calculations, frames, and source artifacts.
Analyst export workflow

From pass scenario to shareable report.

AI-enhanced automation packages the review evidence into a readable report and PDF practitioners can save, share, and revisit.
Future improvements
01Calibrate playabilitylearn what "open" looks like from outcomes and expert labels
02Add video contextconnect clips for body shape, disguise, contact, and technique
03Scale the samplemove beyond ten matches to test patterns across teams and roles
Closing takeaway

Turn pass value into time-aware decision context.

01 Find the moment Surface the pass scenarios worth reviewing.
02 Explain the context Was the option open, constrained, or already gone?
03 Package the insight Package frames, scores, and reasons into a shareable report.

Thank you

Ryan Inghilterra Pitch to the Pros 3