100 / 100
Inside the central range—the middle half of reference players.
Baseball coaching, made clear
See the movement.
Know what to
work on.
Turn one filmed swing or pitch into understandable measurements, clear comparisons, and one focused practice step.
From movement to meaningIllustrated example
College · right-handed hitter · stride53 cm
Off Target37 / 100 · estimate
Separate report example: beyond the College reference range. The report explains the result and selects a practice focus.
9hitting measurements
10pitching measurements
4movement groups per report
1focus and a practice drill
01 / The experienceA clear path from
A clear path from
video to practice.
One report covers one swing or one pitch. Select each step to see what happens next.
- Choose the player, level and side. Add one swing or pitch and confirm its key moments.
- The video engine estimates body positions and calculates supported angles, distances and timings.
- Fixed rules compare measurements with the available player references and calculate ratings.
- The report presents one focus, a practice cue and a drill. AI receives measurement data to explain the results.
Measured first. Explained second. The video engine measures the movement. The AI coach receives those measurements and movement data—not the video—and cannot change scores, ratings or drill selection.
02 / Reading the resultsThree ratings.
Three ratings.
A reason behind each.
Ratings show how a measurement compares with recorded players. “Typical” does not mean “ideal,” and a score is not a prediction of performance.
Measurements → groups → overallA group averages its rated measurements. The overall score averages the rated groups, so every group has equal weight.
Not rated is not zeroMissing data, an unsuitable comparison or no reference range can leave a result unrated. These results are excluded from averages.
Estimates stay visibleCurrent video measurements and their scores are approximate. A high overall score can still contain an Off Target measurement.
How the 0–100 score is calculated
A value inside the On Target range scores 100. Moving toward either edge of the wider Satisfactory range reduces the score steadily to 70. Beyond that edge, the score drops to zero over twice the gap between the central and wider limits on that side. Scores are rounded; values outside the central range cannot score 100, and values beyond the wider range cannot score 70. Narrow reference ranges can make small measurement changes affect a score substantially.
03 / Explore the movementWhat we measure.
What we measure.
What it means.
Choose a movement, player level and side. Each measurement includes a rendered motion-capture example, an example rating and both levels’ available reference ranges.
All explorer values are illustrative examples, not a player’s analysis. Ratings assume a usable measurement and a comparable activity. Animations render public motion-capture examples to show what is measured; the player appearance is illustrated and does not prescribe an ideal motion. The example ratings are separate from these recorded movements. Units: ° = degrees, cm = centimetres, ms = milliseconds (1,000 ms = one second).
All 19 measurements · complete level and side reference
Green shows On Target. Amber shows the outer limits of Satisfactory; the central green interval takes priority. Red means below or above the amber limits, provided the value passes the quality checks. These are the guide’s published, rounded ranges; live reports use the stored precision.
04 / The outcome
One clear next step.
Documented test examples show the complete report: group scores, a focus, a practice cue and one drill. These are examples, not results from the illustrated players.
Work on itThe lowest-rated eligible measurement below On Target becomes the focus. Hitting gives reliable results priority over tentative estimates.
Keep itWhen the usable results are On Target and no excluded result contradicts that conclusion, the report may give a maintenance message and drill.
Keep practicingWhen the evidence cannot support a specific correction or maintenance message, the report gives a general rhythm drill. It does not invent a fault.
Pitching can also include a short AI summary and up to five evidence-linked cards: what the data supports, what to review, and what could not be determined. Rejected cards are withheld; these cards do not replace the rule-based scores or drill.
05 / Keeping feedback groundedClear results.
Clear results.
Honest limits.
The report separates a measured estimate from a supported comparison and from advice that is appropriate to show.
The recording matters.
A hidden joint, uncertain frame or unstable estimate can leave a measurement unavailable. Timing requires the original capture speed; playback speed alone is not enough.
Measurement checks
- Moving a marked moment one frame earlier or later checks how sensitive the result is.
- If this changes the rating or spans at least 20 score points, that measurement does not provide a tip or become the focus.
- Implausible values are treated as tracking problems, not technique faults. Pitching rejects values beyond the wider reference band expanded by two band widths; hitting uses the observed full range expanded by half its spread.
- Foot landing means the first visible touch, not full weight transfer. Torso tilt combines forward and sideways lean. Knee and elbow angles do not measure force or joint stress.
- Shoulder follow-through measurements cover roughly the first tenth of a second after contact or release, not the whole finish.
The activity matters.
A normal swing, a standard pitch and a movement drill are different activities. Their ratings and practice advice follow different rules.
Context and player rules
- For pitching, standard baseball fastballs can use the available reference. Unknown context also uses that comparison with a note; the leg-lift and stride-related measurements then receive no practice advice.
- Rocker and step-back pitching drills exclude leg-lift torso lean, lift-to-landing time and ankle spacing from comparison. Other supported measurements may still be rated.
- Dry pitching deliveries, explicitly different pitch or ball types, and “other” pitching drills are described without scores. Dry hitting swings can keep comparisons but do not receive corrective advice.
- A walk-through swing does not receive a stride-length rating. An unknown hitting drill prevents stride-based advice. Missing or unsuitable timing also limits advice on affected results.
- If the analyzed side conflicts with the player profile, the analyzed side is used for ratings but specific corrective advice is withheld. If no side can be established, there is no rating.
- Hitting uses High School references for youth or an unspecified level, and College for adults. Left-handed High School hitters use the left-handed College reference. All pitching levels use College references, with a notice.
The evidence matters.
AI explains the data in simple language. Fixed rules control scores and drills. Unsupported claims are rejected or replaced with the existing rule-based wording.
Feedback boundaries
- Generated wording preserves the result’s direction and estimate label. It does not add numbers, links, medical claims or promises about outcomes.
- Pitching AI cards must cite their evidence. A separate check can withhold individual cards. They describe observations and review steps rather than diagnose technique or prescribe corrections from unvalidated estimates.
- Six hitting measures currently offer general comparison and repeatability guidance rather than a specific technical correction.
- For tagged standard, rocker and step-back pitching drills, extra unscored observations can describe knee-angle change and the order of peak hip and shoulder turns when supported.
- Rocker and step-back loading observations stay “can’t tell yet”: weight transfer is not measured. Knee straightening versus bending is withheld if shifting the marked frames could reverse that conclusion.
06 / Practice & reference
A drill behind the focus.
The current library contains 21 drills. Open a drill for its instructions and cue. Drill choices and coaching wording remain subject to coach review.
Where the comparisons come from. The guide uses the project’s calibration of the OpenBiomechanics Project, by Driveline Baseball. The calibrated source subset contains 649 swings from 97 hitters and 399 pitches from 100 pitchers. Usable level-and-side reference groups are smaller: 12 High School right-handed hitters; 50 College right-handed and 23 College left-handed hitters; 60 College right-handed and 15 College left-handed pitchers.
Current coverage. There is no separate High School pitching reference or left-handed High School hitting reference. Pitching ankle spacing has no benchmark. Phone-video body estimates have not yet been validated against lab measurements; the reference ranges describe observed players, not proven ideal technique. Some High School ranges are narrow because the reference group is small.
Source & attribution. Based on coach-logic main · 8c71e98; scoring and eligibility checked against product main · 9a9b646, 4 November 2026. Benchmark numbers and the two documented test examples are preserved. Movement animations use public OpenBiomechanics College motion-capture samples; left-handed views are mirrored. Body and hitting bat paths are recorded; head, clothing and shoe appearance are illustrated. Displayed rating examples are separate from those samples. Adapted benchmark and motion data: CC BY-NC-SA 4.0.