Metra Performance validation summary
Preliminary results from our own testing. This is not a peer-reviewed paper. A peer-reviewed study is in preparation, and the numbers may change as the app is updated.
We validated Metra against force plates, the gold standard for vertical jump testing.
* Excludes DRI: 6.6% on the drop jump and 7.9% on the hop test. DRI divides by contact time squared, which roughly doubles contact time's error. Its ICC is in the range shown.
Methods
- ParticipantsN = 23.
- PhoneiPhone 14 Plus on a tripod, front-on to the athlete at 6 ft (1.83 m), with the lens 18 in (45.7 cm) above the force plates and the phone level. No markers and no calibration.
- Force platesVALD ForceDecks dual force plates sampling at 1000 Hz, recording the same jumps at the same time.
- ProtocolAll tests with hands on hips, in this order: countermovement jump (stand still for 2–3 s, then one continuous dip and jump); hop test (10 continuous hops after one warm-up hop); drop jump (step off a box 16 in (40.6 cm) above the force plates, land, and jump straight back up).
- AnalysisEach jump recorded by Metra was paired with the same jump on the force plates.
- ExclusionsDecided before the results were calculated:
- Two or more consecutive dropped video frames during the jump: 5 of 450 jumps (1 countermovement jump, 4 drop jumps), left out of every measure.
- Landing partly off the force plates: 6 of 116 countermovement jumps, 49 of 229 hops and 4 of 105 drop jumps, left out of the measures that use that landing or contact.
- No still stance before and after the jump: 22 of 115 countermovement jumps, left out of the measures compared with the Owen et al. (2014) method.
How the force plate values were obtained
Timing measures
Flight time, contact time, contraction time and the RSI values built from them were measured from the force trace using the force plate software's own event definitions.
Measures from integrating the force trace
Jump height from take-off velocity, countermovement depth, mean propulsive force and mean landing force come from integrating the force trace to get the velocity and position of the centre of mass. These values depend on two things: the athlete's body weight, and the instant the integration starts. If it starts after the athlete has already begun to move, for example after a small rise before the dip, the error carries through the whole calculation. Owen et al. noted that around half of jumpers start a countermovement jump this way.
The Owen et al. (2014) method
We compared these measures with this published criterion method, applied to the raw force trace of every jump:
- Body weight is the average force over 1 s of still standing before the jump.
- The jump starts at the first point where force moves more than 5 standard deviations away from body weight, in either direction, so a small rise is caught as well as the dip.
- Integration begins 30 ms before that point.
Results
| Measure | Jumps | Relative error (RMSE) | Absolute error (RMSE) | Bias ± 95% LoA | ICC |
|---|---|---|---|---|---|
| Countermovement jump · 23 athletes | |||||
| Flight time | 109 | 1.2% | 6.6 ms | −0.2 ± 13.1 | 0.99 |
| Jump height, flight time | 109 | 2.4% | 0.9 cm | 0.0 ± 1.8 | 0.99 |
| Jump height, take-off velocity * | 93 | 2.4% | 1.0 cm | −0.1 ± 1.9 | 0.98 |
| Contraction time | 115 | 2.4% | 19.4 ms | +1.0 ± 38.0 | 0.98 |
| Countermovement depth * | 93 | 3.8% | 1.4 cm | −0.1 ± 2.8 | 0.97 |
| RSI-modified | 109 | 3.5% | 0.02 m/s | 0.00 ± 0.03 | 0.99 |
| Mean propulsive force * | 93 | 1.6% | 0.03 ×BW | −0.01 ± 0.07 | 0.98 |
| Mean landing force * | 93 | 2.9% | 0.08 ×BW | −0.01 ± 0.16 | 0.99 |
| Hop test · 22 athletes | |||||
| Flight time | 180 | 1.7% | 6.3 ms | +0.1 ± 12.3 | 0.99 |
| Contact time | 180 | 2.7% | 5.7 ms | −0.1 ± 11.2 | 0.99 |
| Jump height | 180 | 3.4% | 0.6 cm | 0.0 ± 1.2 | 0.99 |
| RSI | 159 | 4.0% | 0.07 | 0.00 ± 0.14 | 0.98 |
| DRI † | 125 | 7.9% | 0.08 | +0.01 ± 0.15 | 0.98 |
| Drop jump · 21 athletes | |||||
| Flight time | 97 | 1.4% | 7.3 ms | −0.2 ± 14.5 | 0.99 |
| Contact time | 101 | 2.9% | 7.4 ms | −0.7 ± 14.4 | 0.99 |
| Jump height | 97 | 2.8% | 0.9 cm | 0.0 ± 1.9 | 0.99 |
| RSI | 97 | 3.9% | 0.08 | 0.00 ± 0.17 | 0.98 |
| DRI † | 97 | 6.6% | 0.09 | 0.00 ± 0.18 | 0.98 |
* Compared with the Owen et al. (2014) method.
† DRI = (drop height + jump height) / (g × contact time²). Drop jump: the box height. Hop test: the previous hop's jump height. Because contact time is squared, DRI's error is about double contact time's.
- Relative error (RMSE)
- The root-mean-square of each jump's difference from the force plate, as a percentage of the force plate's value.
- Absolute error (RMSE)
- The root-mean-square difference from the force plate, in the measure's own units.
- Bias ± 95% LoA
- The average difference (app minus force plate) ± the 95% limits of agreement (Bland–Altman). 95% of jumps fall within this range. Same units as the absolute error.
- ICC
- Intraclass correlation for absolute agreement, ICC(2,1). 1 means the app and the force plates agree exactly.
Supported iPhones
Metra runs on:
- iPhone 13 Pro, 13 Pro Max
- iPhone 14, 14 Plus, 14 Pro, 14 Pro Max
- iPhone 15, 15 Plus, 15 Pro, 15 Pro Max
- iPhone 16, 16 Plus, 16 Pro, 16 Pro Max, 16e
- iPhone 17, 17 Pro, 17 Pro Max, iPhone Air
- iPhone 18 Pro, 18 Pro Max