Latest revision as of 10:09, 5 November 2018

A selection of the running sensors I've tested. From top left: Moxy, BSX, Wahoo TICKR Run, Lumo Run, TgForce, Garmin Running Dynamics, SHFT, MilestonePod v2, MilestonePod v3, Stryd, and a Gu for size comparison.

There are a growing number of wearable devices that will analyze your biomechanics, mostly from small startup companies, though the bigger players are also contributing to the space. Many of these devices a making use of the cheap and accurate accelerometers that are now readily available, though there are some other approaches being used. These are cheap accelerometers have created an explosion of products, and it seems likely that we'll see a number of companies fail. Here are my sound bite summaries:

• Stryd is a running sensors that I think every runner should have. It's the most accurate way of measuring distance and pace, and has integration with a wide array of running watches. Its estimate of "Running Power" is moderately useful.
• Garmin's Running Dynamics is worth having if you own a Garmin watch that supports it, though I wouldn't buy a Garmin just for its support of Running Dynamics.
• RunScribe provides lots of detailed and useful information, but I consider it a 'running lab' rather than an everyday training tool.
• MilestonePod is amazingly cheap and provides a wealth of data. It's worth the cost just for its ability to track the miles you put on your shoe, but it adds in more data than most other systems that cost many times as much.
• Wahoo TICKR Run gives some interesting metrics, but you have to use their Smartphone App to get them. Their "3D Smoothness" would be awesome if it worked. It's also a Heart Rate Monitor that supports both Ant+ and Bluetooth.
• Moxy can provide a new way of evaluating exercise intensity by looking at the oxygen saturation of the blood within the working muscles. However, not only is it expensive, but you'll also need to dedicate significant time and effort into getting the best out of it.
• TgForce is a "one trick pony", but it's a great trick. It measures the Impact on the lower leg rather than the foot, providing great real time metrics. Sadly, there are production issues at the moment that are causing the sensors to fail.
• Moov Now This is a cheap and interesting sensor. Like the TgForce it can measure lower leg Impact, but it's got software problems that have not been fixed.
• Lumo Run Needs more development before it's ready for prime time.

1. Where does the sensor go?
2. What are your options for analyzing the results?
3. Can you see the data in real time? This makes a huge difference if you're trying to improve your biomechanics. "Watch" means you have real time information so on the watch which is one of the ideal approaches. "Phone" means you get real time data shown on the phone display, which if it's done right can show an awful lot of data. "Audio Coaching" means you get periodic audio messages which I find more annoying than useful. "RTAF" is Real Time Audio Feedback, which means that your hearing an indication that you're past a threshold. For example, the TgForce will be if your impact is higher than a given threshold, and it will be part of every footstep.
4. Do you have to carry your phone with you? I hate carrying a cell phone when running, so this is a factor for me.
5. Phone needed to get anything beyond HRM.
6. You can use two pods and compare the data, but the system won't compare the results for you.
7. Impact can be measured in various places, such as the foot impact, the lower leg (tibia), or the chest. The impact can be measured as peak acceleration (g's) or as rate of change of acceleration (AKA "Jerk").
8. This is Rate of Impact, which is slightly different to simple impact g's (see review for a few more details).
9. The $200 kit only comes with one sensor, so you'd need to by a second.
10. The Sensoria measures a number of metrics, but it does so remarkably badly.
11. This is a measure of the impact on the lower leg, which excludes the impact that the foot as absorbed.
12. This is the rate of change of acceleration (AKA "Jerk") in three separate axis: up/down, side-to-side, and left/right.
13. A sensor that doesn't allow for data export locks you into the vendor's tools.
14. A 9-axis sensor can detect twisting motions through its gyroscope, in addition to the acceleration measurement of a 3-axis sensor.
15. Measuring Vertical Oscillation from a footpod is a dubious proposition and the results seem unsurprisingly poor.

1 What To Look For In A Running Sensor

Having tested a fair number of running sensors, I think there are 3 aspects of the metrics that a sensor provides that you should consider.

• Accuracy. Not unreasonably, a running sensor should provide a reasonably accurate metric. Depending on the metric, the value could have some degree of error without affecting its value if the error is always proportional. For instance, if a sensor gives a value for Ground Contact Time that is always 10% too high, that may be perfectly acceptable as you can look for relative changes. Or a sensor could have a synthetic value of "smoothness", which might not be based on a real-world measurement, but could still provide actionable information. Of course, accuracy is always preferable, and an inaccurate sensor could cause more problems than it solves. A Cadence sensor that reads 10% too high might lead you to believe that your Cadence is fine when it's actually far too low.
• Responsive. A useful metric needs to be responsive enough that you can see the results of any changes to your running form. A metric that has too much smoothing can be quite frustrating, while one with too little smoothing can be so twitchy that it's equally useless. Of course, the responsiveness also depends on the feedback method, and a metric that you only get in post run analysis is far harder to use.
• Meaningful. A valuable metric is one that has some meaning to you as a runner, either because it has a bearing on your Running Economy or because it may have a bearing on your injury risk. The current state of the available research doesn't provide a huge amount of confidence linking many of the available metrics to either injury rates or Running Economy, but there are some that seem to have potential.
• Actionable. The final characteristic of a valuable metric is one that you have some control over. For instance, Cadence is relatively easy to change and therefore the metrics are actionable. Other metrics may be rather tricky to modify with your running form, such as a braking force (at least, I found it very hard to change that metric.) Another reason why a metric may not be actionable is because correlation is not causation. For instance, lower Ground Contact Time is associated with a better Running Economy, but it's not clear if reducing Ground Contact time will directly improve Running Economy, or if there is some other variable that is responsible. (My personal suspicion is that improvements in Running Economy are more related to Cadence, which in turn changes Ground Contact Time. I'd like to see research into Ground Contact Time normalize the values against Cadence.)

2 Stryd

Main Article: Stryd

The Stryd Footpod is my favorite running sensor. It provides vastly better distance and pace information than any GPS watch I've tested. This alone makes it well worth the purchase price and if for any reason I lost mine, I'd replace it immediately and without hesitation. Stryd is primarily marketed as a "running power meter", though I think this sells it a little short. Running Power Meters are somewhat flawed concept, as they are really an estimate of power, and the estimate is of far less value than a cycling of power meter is to a cyclist. That said, I found that Stryd provides a useful way of providing even effort on uphill sections (but not down hills.) It also has the best watch integration of any running sensor.

3 Garmin's Running Dynamics

Main Article: Running Dynamics

Many newer Garmin watches ( 620, 920XT, Epix, Fenix 3, etc.) combined with a special chest strap will provide extra metrics that can give insight into your Running Form. The Garmin watches will provide these metrics in real time, allowing you to see the effect of changes in your form. These metrics include:

• Vertical Oscillation. This is how much the torso moves up and down with each stride. It is generally believed that less Vertical Oscillation is a better, but I suspect that this is an oversimplification. It's sometimes a thought that a greater Vertical Oscillation will result in greater Impact, but this is not the case. Impact is how quickly you decelerate, so landing hard can result in less vertical movement but a shorter, more intense deceleration. It's even suggested that greater Vertical Oscillation will result in more braking force, but that does not seem reasonable. I believe that some are part of a runners vertical movement is likely to be elastic in nature (consider a bouncing rubber ball), some of the vertical motion will be while the runner is airborne (ballistic), and some of the vertical motion is the vertical deceleration as the runner lands. So it seems likely to me that excessive Vertical Oscillation is bad, but it's a tricky to know what excessive is likely to be, or how to correct it.
• Ground Contact Time (GCT). The Science of Running Economy generally shows that longer Ground Contact Time is correlated with poorer Running Economy.
• Ground Contact Time Balance. This is the relative Ground Contact Time ratio of the left and right feet, which will reveal potential imbalances in the body.
• Computed metrics. The Garmin watches will use the basic metrics to calculate things like stride length (based on cadence and pace), and Vertical Ratio (Vertical Oscillation to stride length ratio).

Garmin's Connect web site shows the metrics gathered using the HRM4 and Garmin Fenix 3.

4 RunScribe

Main Article: RunScribe

RunScribe is a small pair of Footpod's that provide a wide variety of Foot Strike metrics. These include things like Impact G's, GCT, Braking G's, Pronation, and more. I love the detailed metrics that are provided, and the insight into my running form, and possible imbalances. The main disadvantage with the RunScribe approach is that the data is not available during the run, and has to be analyzed afterward. RunScribe is rather more sophisticated than most Footpods as it not only has a 3-axis accelerometer, it also has a 3-axis gyroscope and a 3-axis magnetometer allowing it to sense far more movement. You can read about my testing at RunScribe.

An overview of the data from RunScribe

5 MilestonePod

Main Article: MilestonePod

MilestonePod is a vastly cheaper alternative to RunScribe, and while it doesn't provide the detailed analytics, it's a great value for money. It's also a cheap and effective way of keeping track of the mileage of your shoes, and is worth its price for that feature alone. Like RunScribe the MilestonePod doesn't provide real-time metrics, and you have to analyze the results post run. You can read about my testing at MilestonePod.

Foot Strike, Cadence and Stance Time (Ground Contact Time)

Rate of Impact, and Stride Length

Leg Swing and "Runficiency"

A sample of the graphs in the app.

6 Moxy

Moxy uses the same approach as BSX, using infrared light to detect the oxygen saturation of the muscles of blood. Moxy is not as easy to use as the BSX, but it's the data seems to be rather more reliable, though it requires rather more effort to interpret. I'll post a full review soon.

Here data from a Moxy test, showing a clear decline in SmO₂.

7 TgForce

The TgForce measures just one thing; the peak Impact on your lower leg. While this is a far cry from the broad array of data gathered by devices like RunScribe, TgForce may add some particularly valuable insight. Measuring the Impact that the foot can be a little misleading as the movement of your ankle can absorb a significant portion of that shock, so knowing how much of the force is transmitted to your lower leg may give a much better insight into injuries. The Impact that your lower leg (tibia) receives will be transmitted into your knee, so reducing that stress might be quite helpful. The companion app only runs on iOS devices, not android, which will limit the appeal for some, and unless you buy two devices, it will only measure one leg at a time. Even if you get two devices, it doesn't appear that the app readily supports this approach, unlike RunScribe that does a lot to provide side to side comparisons. That said, I really like to have real-time feedback, something that TgForce provides, either visually for use on a treadmill, or audibly outside. I've ordered a TgForce and my initial testing shows that the data from TgForce might be quite valuable, but there is a quality control problem and I've had several sensors fail. They have a fix and I should have a fully working sensor soon.

This is the real time display, showing the Impact of each step (from one leg) in the bar graph, plus the current and average Impact. There are also metrics such as total steps and Cadence.

8 Moov Now

Main Article: Moov Now

Moov Now is the second generation of Moov activity monitors, and is a small ankle mounted pod. It supports various sports, including cycling, swimming, boxing, and general workout routines like crunches, plank, etc. The Moov Now will also act as a general purpose activity tracker, including sleep tracking. However, it's the running functionality that I'll focus on. The Moov Now uses ankle mounted accelerometers to give a number of metrics, including Cadence, Impact (at the tibia), and "range of motion" which is how far your leg swings from the furthest forward to the furthest back. I don't see the range of motion as a particularly useful metric, as it gives no indication if you are Over striding or not. However, I like the idea of getting Impact metrics on the tibia (lower leg), as I suspect this is more relevant than the Impact measured at your foot. It seems to me that your foot could undergo massive deceleration as your forefoot touches down without necessarily generating a huge amount of stress on your ankle, and even less stress on your knee. Unlike most activity monitors, it uses a replaceable battery that should last months, which is really nice, especially if you're traveling. It's a shame the Moov Now needs to be linked to a smart phone to work, as it means taking your phone on the run. The Moov Now seems to be an interesting device given its price, especially if you are interested in the other sports that it supports. Error: Could not parse data from Amazon!.

The Moov Now sensor in Red is quite small (the squares in the background are 1"/2.5cm).

The display of impact in real time is rather poor at the moment; there's no chart or average, and the impact has no decimal to give precision.

9 Wahoo TICKR Run

Main Article: Wahoo TICKR Run

The Wahoo TICKR Run is a chest strap based system that's similar to Garmin's Running Dynamics. It supports both Bluetooth and Ant+, which is nice, but most of the functionality beyond heart rate requires you to have your phone with you on your run. If you do, then you get Cadence, Ground Contact Time, Vertical osculation, and 3D smoothness. This 3D smoothness shows the jerk (rate of change of acceleration) in three planes; forward-backward, up-down, and side to side. This has great potential, but the smoothness depends far too much on the tightness of the strap, even varying with your breathing.

The smoothness that Wahoo displays seems like it could be really valuable if only it was usable.

10 Lumo Body Tech Run (Bankrupt)

Main Article: Lumo Run

Update: Lumo has gone bankrupt and as far as I can see their sensors are effectively bricked. Unlike BSX, who made some of their software open source, Lumo has just disappeared. Like many other devices, Lumo Run uses accelerometers to measure body movement, but uniquely (so far) Lumo places the accelerometers at the small of your back. This allows Lumo to not only detect Cadence and Vertical Oscillation, but also breaking and pelvic movement. The measurement of braking force is rather different from RunScribe, as Lumo measures how much your overall body slows up with each stride, rather than measuring the deceleration of your foot in touch down. I believe that both approaches are important, and give valuable insight into possible biomechanical problems. In addition, Lumo will measure how much your hips (pelvis) will drop from side to side, and how much it rotates (twists). Lumo is available Lumo Run is available for Error: Could not parse data from Amazon!. I really wish that Lumo would measure Impact, as it would be great to know how much of the foot strike Impact reaches the hips.

This is all you see when you're running; so much screen space with so little data.

After the run is complete, Lumo displays some summary statistics.

11 SHFT

Main Article: SHFT

The SHFT system uses two pods, one on the chest and one on the foot. The SHFT system includes some unusual metrics such as toe off angle and body angle, as well as Cadence, GCT, Vertical Osculation, Impact, Braking, and Foot Strike. They also claim to measure power for running, but my testing suggests this is far from useful. The system requires you to carry your phone, and the main feedback is via audio through the headphones which I generally find rather ineffective. There are some good post run analytics available via the app and on their web site, as well as data export of the basic running information. The price of the two SHFT sensors is $300 which is a little high compared with other systems, but it does provide quite a bit of information. I've found the SHFT metrics are not as accurate as I'd like; read more at SHFT.

The SHFT pod is delightfully rounded and rather organic.

12 Sensoria

Main Article: Sensoria

The Sensoria smart socks are made of materials that can detect pressure changes under your foot and transmit them back to anklet that communicates with a smart phone via Bluetooth. The Sensoria anklet also includes accelerometers to enhance the data gathered from the pressure sensors. While these socks are interesting, unlike a true pressure plate that has a matrix of pressure sensors, the socks have just three sensors, one in the heel and two in the forefoot. (Error: Could not parse data from Amazon! for a pair of left & right socks, plus one anklet.) Note that the socks last for 60 washes and cost $45, which works out to $0.75 per run. That might be cheap to a triathlete, but to most runners that works out pricy. Read my full review of Sensoria and the list of problems.
(Sensoria also make a tee that shirt includes an embedded Heart Rate monitor strap, you can plug a standard heart rate transmitter module into. They sell a dual Bluetooth/Ant+ transmitter, as well as working with the Polar H7 and some Garmin modules. )

13 BSX

Main Article: BSX

The BSX Lactate Threshold Monitor attempts to estimates your Lactate Threshold by measuring the oxygen saturation of the blood within your muscles. The results of the first version were extremely disappointing, but I have not had the opportunity to fully test the updated hardware. However, even if the BSX works perfectly, its value is limited by the relative ineffectiveness of Lactate Threshold Training. You can read about my testing at BSX.

The dashboard view of the BSX app, showing previous results.

14 RUNTEQ Zoi

Zoi places sensors on both the foot and the torso. This allows it to gain a little more insight than other sensors as it knows about the movement of the foot and the torso independently. Eventually I expect to see a company produce a group of sensors placed on each foot, each knee, and the pelvis, which would give insight into the movement and Impact forces across most of the body. The Zoi only has one Footpod, so it doesn't give you detailed foot strike information in the way that RunScribe can, though you could see them adding that functionality in the future. The Zoi gives quite a few metrics, including Cadence, Ground Contact Time, Vertical Oscillation, breaking (at the torso, not foot breaking), of Foot Strike type (fore, mid, heel), Foot Impact, and some Pronation information. Zoi has a smartphone app that provides real time feedback and post-run analytics, but I've not seen any support for displaying metrics on a watch. Currently Zoi is on pre-order in Europe for 150 EUR. I'd like to test this system if I can get hold of one. The approach is similar to the SHFT system, though the SHFT uses a 9-axis sensor and it's not clear what the Zoi uses.

15 RPM²

RPM² (Remote Performance Measurement/Monitoring) is a pair of insoles that fit into your normal running shoes. These insoles measure pressure and use accelerometers to measure movement (a little like Sensoria). The details are not entirely clear from the web site, but they claim to measure Cadence, Ground Contact Time, Foot Strike, and "acceleration power". The system also claims to measure running power, though I'm not sure of the methodology. The RPM² system measure pressure in four areas (Sensoria has three), giving a Pronation measurement. There are notes that RPM² insoles are not waterproof, which is rather disconcerting, and if you run in different shaped shoes the fitment is likely to be problematic. The sizing of the insoles needs to take into account the position of the ball of the foot to ensure the sensor is in the right place. The system supports Bluetooth to connect to your phone as is common with running sensors. The RPM² can also connect to an Ant+ watch, though it's a bit of an ugly kludge. You need to have your phone with you, and then plug in the Wahoo Key adapter via a lightning-to-30 pin adapter to transmit Ant+ to a watch. The Wahoo Key and adapter are all extra bits you have to buy. The system can also be used as a power meter for cycling. Their web site is Remote Performance Measurement/Monitoring.

16 Kinematix TUNE

Kinematix went out of business in April 2017.

17 Testing Approach

Where possible, I like to verify the metrics that these running sensors are providing. In some cases, this is a fairly trivial, such as Cadence which you can measure by simply counting your steps. Some other metrics I've verified by using High Speed Video (HSV.) This requires a little bit of time and effort, and a lot of attention to detail but it can be quite an effective approach.

• Ground Contact Time. I've used HSV to verify Ground Contact Time, and found more variation than I expected. Using HSV to measure Ground Contact Time you can see the very earliest and latest stages of contact as well as estimating the pressure changes from the compression of the foam midsole. Examining the video makes it clear that some devices are including time where the shoe is completely airborne, which is a little surprising.
• Vertical Oscillation. I've evaluated the accuracy of Vertical Oscillation by looking at the movement of my torso using HSV and lasers. This proved to be rather more tedious and time-consuming than I'd anticipated, but the results were worthwhile (to me at least.)
• Foot Strike. I've done some simplistic testing of sensors that measure Foot Strike. I've run with these sensors in my usual heel strike, plus running forefoot both with and without my heel touching down. I've used HSV to verify that my foot strike pattern is what I think it is, as it's easy to be mistaken. I didn't include midfoot strike as this is tricky to get exactly right, and tricky to verify using HSV. This approach to measuring Foot Strike is based around the point of first contact, not the point of highest force. Personally, I believe that the latter approach is far more relevant, but harder to measure.
• Power. It's tricky to measure running power without access to VO₂ equipment, so I've resorted to some simple tests. An incremental treadmill test should show an increase in power with an increase in pace; if it doesn't pass that test, then it's probably not much use.

18 Ideas For The Future

This section documents a few ideas I've had for running sensors, partly to inspire manufacturers, and partly to disclose these ideas so that they cannot be patented in the future.

18.1 Proportional Audio Feedback

Many running sensors will provide information on your biomechanics using an audio message. Typically, this is a spoken message such as "you're landing on your heel" or "your braking is 1.31 feet per second." I generally find these audio messages far more annoying than they are useful. The messages tend to occur too infrequently for me to get a sense of how any modifications in my form are changing the metric being measured. The spoken word means it's a little tricky to combine this with music or radio. A better approach is to have a simple signal like a beep when your metric is outside the desired level. For example, TgForce will beep when your Impact is too high, and you can combine this simple audio signal with music for the spoken word from a in audiobook or the radio. I believe that a superior approach is to have a variable audio signal that indicates how well you're doing. So instead of having a simple threshold such as 7g you'd have a range such as 5-9g. Then, when your Impact is above the lower threshold you would get an audio signal, but the audio signal would vary in one or more of volume/pitch/duration depending on where you are in the range. So a 5g Impact would produce a quiet, short, deep beep, and 9G Impact would produce a loud, longer, high-pitched beep. This way you'd get quantitive feedback on how well you're doing against your chosen metric.

18.2 Deriving Practical Impact From Acceleration

Often running sensors will provide a value for impact based on the acceleration measured at the foot or leg. However, the bulk of the stresses on the lower limbs comes from the acceleration of the overall body mass during landing. Therefore, I believe it would be more effective to measure the acceleration of the torso on landing to provide an estimate of the stresses on the lower limbs. I suspect that both peak acceleration and the area under the acceleration curve would provide insight into the stresses of running. Peak acceleration is fairly obvious, and I suspect is more useful than a jerk. However, it may be useful to evaluate the area under the acceleration curve, as the time spent under stress could also be a significant factor on injury rates.

18.3 Measuring Foot Strike

Several running sensors will measure Foot Strike, though so far, I've only found RunScribe and SHFT to provide useful data. However, even these devices tend to measure the foot angle on initial contact, and I suspect that it would be far more useful to measure foot and call at the time of maximum deceleration of the torso. This requires at least two sensors, one on the torso and another on the foot, and ideal three so there's one for each foot. Synchronizing the timing between the devices is likely to be problematic, but not insolvable.