Public methodology
What the tags can—and cannot—say
Goalie Vision is designed to support a conversation between a goalie, parent, and qualified coach. It is not a scouting grade, psychological assessment, medical tool, or replacement for coaching judgment.
Evidence rules
- Observed: state only what was tagged or visible in the event data.
- Possible focus: suggest an area worth reviewing without claiming an unseen cause.
- Confidence: low for fewer than two useful examples, medium for two to four with reasonable completeness, and high only for at least five examples with strong completeness.
- Evidence: link the clips that produced the pattern.
- Next step: ask the goalie’s own qualified coach to confirm the technical cue.
Data completeness
Every report measures how many on-goal events have shot locations, how many saves have a save type, and how many goals have a net zone. A warning appears below 70% completeness or below five on-goal events. Percentages should always be read with their numerator and denominator.
Model card
Goals Saved Above Expected
- Model type
- Rule-based shot-difficulty estimate; not a population-trained statistical model.
- Population
- No fitted population. It is not calibrated for age, gender, league, or playing level.
- Base estimate
- 0.17 for a tagged shot inside the defined high-danger box; 0.04 outside; 0.08 when location is missing.
- Situation multipliers
- Screen 1.4×; breakaway 1.8×; lateral pass 1.2–1.65× depending on the tagged pass origin; cross-slot pass 1.6×; backdoor 1.65×; one-timer 1.15×; odd-man rush 1.35×; second chance 1.3×; deke 1.35×; turnover 1.3×; tip or tagged deflection 1.6×; power play 1.3×. Two detailed defensive-context observations can add small 1.05–1.1× refinements when they do not duplicate an existing situation tag.
- Output cap
- Each event is capped at 0.74 so overlapping situation tags cannot make a single chance look nearly certain to become a goal.
- Calibration
- The original scale was reduced after manual review found that it overstated expected goals in real tagged games. This is a product calibration, not an external population study.
- Validation
- No external validation sample, calibration error, or confidence interval is available yet.
- Minimum sample
- No number makes this heuristic a validated grade. The interface warns below five shots and shows sample counts throughout.
- Allowed use
- Add context to the shot mix in a parent or coach review.
- Do not use for
- Youth ranking, league comparison, selection decisions, mental-performance claims, or Player Mode.
Goals Saved Above Expected is the sum of these shot-difficulty estimates minus actual goals against. A positive value means more goals were stopped than the estimate predicted. The baseline is not calibrated by age, league, or player population, so this is context for reviewing a goalie's games, not a validated comparison between goalies.
Next-shot results
This counts the outcome of the next on-goal event after a goal within the same session. A goal with no later shot is excluded. The report shows “saved opportunities / total opportunities.” It does not measure resilience, confidence, composure, or any other psychological trait.
Drill library
Recommendations come from a fixed, structured starter library with age range, equipment, duration, instructions, and a measurable outcome. The product selects a drill that matches the observed tag pattern. A credentialed goalie coach must sign off on the library before launch, and the goalie’s own coach should adapt technique, pace, and volume.
For an established external baseline, coaches can also use USA Hockey’s beginner, intermediate, and advanced goaltending practice plans and its Mobile Coach age-appropriate practice tools. These links are references, not an endorsement of Goalie Vision.