Visible hip travel is projected along the erg rail and used as a proxy for seated-position and slide movement.
The physical seat is not instrumented.How ErgScreen Reads a Video
Video becomes evidence when the movement behind it can be traced.
A coach should not have to trust a mystery score. ErgScreen shows what the clip supports, which strokes carried the read, and where the evidence stops.
ErgScreen turns fixed side-on video into a stroke-by-stroke movement comparison and keeps the supporting stroke evidence connected to each report-facing finding.
Within-athlete comparisonAthlete's own reference strokesVisible support limitsNext-best evidence
What the camera can see
Three visible movement channels. One stroke-by-stroke story.
The monitor measures the result. ErgScreen shows how the visible stroke was organized during the submitted clip — without pretending the camera is a force plate.
Visible wrist movement relative to the hip is used as a proxy for handle travel and connection timing.
The physical handle is not separately tracked.The visible shoulder-to-hip axis creates a two-dimensional trunk signal for timing and stroke comparison.
This is not a 3D anatomical measurement.ErgScreen measures visible two-dimensional movement. It does not directly measure handle force, monitor watts, mechanical efficiency, boat speed, or on-water performance.
From clip to comparison
The software does more than find poses. It determines which strokes meet the comparison requirements.
Before any movement finding reaches the report, the clip passes through a sequence designed to protect the comparison.
Framing, visibility, frame rate, camera stability, and usable duration set the evidence ceiling.
The better-supported body side is used when one side is obscured or less reliable.
Movement is aligned to the visible rail so supported camera tilt is less likely to be mistaken for athlete movement.
Catch, finish, recovery events, and the next catch define the stroke boundaries.
Incomplete, clip-boundary, low-support, or temporally inconsistent strokes are retained for audit but kept out of the scored comparison.
CatchDriveFinishRecovery → next catch
Each usable stroke is mapped onto the same drive and recovery timeline. That lets ErgScreen compare movement shape and timing even when individual strokes differ slightly in duration.
Clip-boundary strokes are excluded when needed to reduce edge effects. Incomplete, warm-up, or otherwise unsupported strokes may also be omitted from the scored comparison, with the reason retained in Stroke Evidence. An omitted stroke is not automatically a poor stroke.
A complete evidence path
One 90-second clip. Every step accounted for.
In this sample recruiting clip, the software did not jump from video to conclusion. It preserved the full stroke population, separated the supported comparison sample, and kept the strokes behind each finding visible.
The complete clip population remained available for audit.
Only strokes meeting the comparison requirements entered the report read.
Warm-up, end-of-clip, and incomplete strokes were excluded with stated reasons.
Two strokes moved beyond the report threshold relative to the session reference.
One stroke supported a qualifying Drive Organization finding.
Nothing becomes a report-facing finding until the required stroke and signal criteria are met. The supporting strokes remain beside the conclusion, and omitted strokes remain visible as omitted—not silently discarded.
Added Movement
Not “good” or “bad.” A measurable change from the athlete’s own session reference.
ErgScreen combines normalized movement changes from seated position, trunk orientation, and hand path into a video-derived comparison. The reference comes from this athlete in this clip — not from a universal model of perfect rowing.
+5%
The report threshold
A stroke enters the Added Movement count when its composite movement value is at least 5% above the session-specific clean baseline.
Why the threshold exists
Every stroke is different. Not every difference deserves a finding.
The strokes with lower composite movement form the session’s comparison band. The median of that band becomes the session comparison baseline used in the report. The threshold helps separate small within-clip variation from a clearer change.
Session-specific
The athlete is compared with their own supported strokes.
Software-defined
The 5% rule is a reporting threshold, not a universal biomechanical standard. Trigger calculations use the underlying value; displayed percentages may be rounded.
Evidence-linked
The strokes that cross it remain visible in Stroke Evidence.
What “session reference” means: the session comparison baseline is the statistical reference used for calculation. Representative reference and changed examples are real supported strokes selected to illustrate the difference, not the most flattering or extreme frame.
Drive connection & carry-through
The drive can change without the monitor explaining why.
The report currently groups qualifying visible drive findings under Drive Organization. It is a compiled video-based category describing visible drive connection and carry-through — not a direct force, power, or energetic-efficiency measurement.
Drive Organization is an ErgScreen reporting category, not a universal rowing biomechanics standard.
Connection timing, catch continuity, and early hand-path speed build can show where the drive begins to change from the more consistent session pattern.
Peak visible hand-path velocity, carry-through, and finish-side consistency can help show whether the athlete’s better-supported drive pattern holds.
Endpoint travel and finish carry-through can support a qualifying drive-pattern change.
One isolated signal does not create a drive finding. The category requires either a qualifying change in visible drive travel or finish carry-through, or agreement across multiple supported drive signals. Recovery behavior is evaluated separately, and Added Movement remains a separate measure.
How uncertainty is handled
When the video cannot support the read, the answer should get smaller.
Support is not one magic confidence number. ErgScreen tracks whether the required landmarks, signals, rail alignment, and stroke boundaries are available for the finding being considered.
The required channels and boundaries are sufficiently supported for report-facing use.
The result may remain visible for coach review, but it should not carry the same weight as a fully supported report finding.
If a required source is missing, that category is marked unavailable rather than silently converted into a zero-event result.
Measurement confidence is not decision confidence
A clear read can still leave the important question unanswered.
ErgScreen separates confidence in the measurement from confidence in a larger coaching or recruiting decision. A clip may support a reliable stroke-to-stroke comparison while leaving fatigue response, rate sensitivity, cue response, or repeat-session consistency untested.
Video quality, landmark support, rail alignment, stroke boundaries, and signal continuity determine whether the comparison can carry the report.
One controlled piece may not show what happens under fatigue, at another rate, after a cue, or across another session.
Clear measurement does not become false certainty. “Not tested” is different from “no concern,” and missing evidence is not converted into a negative judgment or a clean bill of health.
The clip has to earn the conclusion
Not every clip answers every question.
A standard side-on screen can support a strong stroke-to-stroke movement comparison without supporting a fatigue, late-piece, or race-rate conclusion. ErgScreen keeps those questions separate instead of stretching one short clip beyond what it contains.
Movement comparison, longer-clip trends, and fatigue reads each have their own sample and support requirements.
ErgScreen checks whether the same pattern also appears in reference strokes before making it the main read.
The same screened-stroke population carries through every report-facing calculation. Sequencing events are checked for a coherent timeline, and unsupported longer-clip or fatigue conclusions remain unavailable.
What would clarify the read next
The next useful evidence is part of the answer.
When the current clip leaves an important question unresolved, ErgScreen can suggest the smallest follow-up recording likely to clarify it. The goal is not more video for its own sake. The coach decides whether that follow-up is appropriate.
A steady-rate clip captured during early-to-moderate fatigue can test whether the change persists when the athlete is less fresh.
A short clip with enough post-cue strokes can show whether the athlete can organize the movement differently.
A matched repeat or small controlled rate step can separate a one-clip exception from a more stable tendency.
ErgScreen does not just show what changed. It shows what the video supports—and what follow-up evidence may clarify the decision next.
What can weaken the read
The camera sets the ceiling.
Recording conditions can change which channels are supported, which side is used, and which strokes are eligible.
Camera movement or rollNon-side-on viewHands or hips leaving frameFar-side occlusionLow frame rate or heavy compressionCropped or edited footageClothing or lighting that obscures landmarksIncomplete strokes
What ErgScreen does not determine
The limits are part of the product.
Handle force or loadMechanical efficiency3D joint motionLeft-right symmetry from one viewBlade workBoat speedInjury riskAthlete potential, lineup value, or recruiting outcome
ErgScreen can show an association between visible movement and a changed stroke. It does not claim that one visible pattern caused an outcome.
The credibility standard
Structured video evidence — not laboratory measurement dressed up as certainty.
Under the same source video, software version, and configuration, ErgScreen applies the same measurement and classification logic. The same input produces the same report logic and evidence population. ErgScreen is not presented as formally validated against instrumented biomechanics, motion capture, or independent expert labeling.
The software should get quieter when the evidence gets weaker. ErgScreen shows what the clip supports, keeps the supporting strokes visible, and marks the rest as limited or unavailable.