The Snowcookie...no it's not edible

[Noodler]

Hi Scruffy. I'm Chris, a ski instructor in Snowcookie team. Answering your question:

Snowcookie measures correlation between chest sensor and two ski sensors giving an exact angle of hip angulation ( common problem for skiers is that they don’t angulate symmetrically left vs right). Since the hip dump is a movement which often starts in subtalar joints followed by the rotation of pelvis you’d have to put another sensor on your pelvis to verify it. (and putting to much sensors on someone is pretty much killing the user experience). What Snowcookie can really measure is the skier's torso alignment with the direction of the movement of the skis. But when it comes to a score we assume it won’t be benefit to "hip dump" during turns. That's because it can generally decrease inclination of the edges, so carving would be less effective and that’s where you’ll get even less points.

The problem then is trying to correlate the actual position of the pelvis as it sits between the chest and the skis. It is quite possible for the chest device to be "fooled" into thinking the hips are being kept level, when in actuality spine bend is being used to keep the shoulders level, while the hips are actually banked/inclined incorrectly.

IMHO, the sensors on the body should be at least 2 and be placed on the area of the iliac crest of each hip joint.

 

[Mike King]

How does this product differ from Carv? What’s it key advantages? Weaknesses? Who are the ski instructors/coaches devising the algorithms?

 

[Snowcookie]

Hi guys,


I am yet another Snowcookie team member here My name is Mateusz (but pls call me mati if you want) and I am the Chief Scientist. Nice to meet you all! If there are any questions or blames that something does not work in our system measures, or it does work, but in a different way you would expect it to (sometimes misleadingly called a bug you know who should you call.


Firstly, sorry for our such a long absence here, but we are preparing hard for the upcoming season, hoping to amaze some of you with our product.

Now, answering your questions:

How does this product differ from Carv? What’s it key advantages? Weaknesses? Who are the ski instructors/coaches devising the algorithms?

Please see this short comparison:

It should already give you an overview about differences between us, Carv and present also other players in that field.

The key advantage of Snowcookie system? For sure chest sensor giving you much wider technique and style analysis. Of course - Carv have their strong sides too, like direct pressure measurement with insoles inside the ski boots, but we can also observe it indirectly with accelerometers mounted to skies and their vibrations (or their differences to be specific). We are also not limited only to skiing, but that is another, separate topic.

Weaknesses? We are aware advanced skiers can fool around the system in some aspects to get worse score than they should get, even when skiing properly, but as there is no point in doing that (or I am missing something in getting valuable and best possible score) we accept it for now. This partially connects to pair point made by @Noodler in his last post under this topic. It does not mean that system doesn`t work, it only should point you out what you all already know for sure - how complex skiing is in therms of motion analysis. It was already raised here on the forum (i.e. https://www.pugski.com/threads/tracking-your-com-and-your-bos.16511 about CoM and BoS). In that domain handling all possibilities is sometimes hard. I will just spoil here that in parallel to current version of algorithms we work on a new generation based on more than 500 skiers (both recreational and professional). This is 6000+ runs we have collected so far in our database (by the way that is more than 500k+ turns to analyze! How cool is that? )We also do not have coaching yet, but it is in advanced progress and is expected to be introduced at the beginning of 2020 (February/March).

As for the question about supporting coaches - they are from both sides: US and EU (Switzerland to be precise here). That is why our skiing level is inspired with PSIA levels for recreational skiers combined together with professionals.

If you have any other questions about Snowcookie feel free to ask - we will do our best to answer them as precise as we can as soon as possible.

 

[Snowcookie]

Thank you for that info. How long should a transtiond take?

Hi Josh,
Sorry for this late answer – I thought other team members have left me only the last question in this thread to elaborate. Let me quickly fix our mistake keeping your question unanswered for such a long time.

Let me start with a question - how do you define the term “transition time”? It is probably not a surprise that there are different definitions given by different coaches for different terms and this one is not an exception. I have heard at least two definitions for the “transition time”. One defines the “transition time” as a time required to start next turn being in the current one, so a time required to make skies start turning in opposite direction. On the other hand some treat it as a max angle to max angle time in consecutive turns. For me, the first definition sounds better, and this is how we define it in our system. Again - this is my opinion – and I do not know how you understand it. If our definition is OK for you, I can come back later with some statistics about this measure. I hope it is clear that those times would be different in different conditions, for different skiers. Nevertheless, I can provide some meaningful range in a well defined cases.

Answering your question more generally - how long should transition take - I would say that for recreational skiers this time should be as long as required, just to feel comfortable and being sure to have a full control of the skis. On the other hand, for semi and professional skiers, this should probably tend to zero, as they are pushing themselves to the limits while skiing on the edge of keeping control over the skis and fighting for each millisecond. Don`t know if that satisfy you? If you are still curious about it I am open for the discussion.

 

[Skitechniek]

Interesting point of view about the short transition.

I have a racing background and in racing, contrary to popular believe, you want the transition to take as long as possible.

Edging means lots of resistance, transition means little resistance. Hence shortening your transition means skiing with resistance for a longer period of time. Fast transitions are merely used due to dinstance/time constraint. If you don't transition fast you'll be skiing in the wrong direction.

I think every racer has heard the phrases 'be patient' and 'ski with flow' before. Which probably roughly translates to don't rush your edge angle and wait for the pressure to come to you. That is what you get taught at racing academies in Austria at least.

I'd love to hear other peoples take on this. Maybe people can share experiences from race academies in other countries as well?

 

[Snowcookie]

Interesting point of view about the short transition.

I have a racing background and in racing, contrary to popular believe, you want the transition to take as long as possible.

Edging means lots of resistance, transition means little resistance. Hence shortening your transition means skiing with resistance for a longer period of time. Fast transitions are merely used due to dinstance/time constraint. If you don't transition fast you'll be skiing in the wrong direction.

I think every racer has heard the phrases 'be patient' and 'ski with flow' before. Which probably roughly translates to don't rush your edge angle and wait for the pressure to come to you. That is what you get taught at racing academies in Austria at least.

I'd love to hear other peoples take on this. Maybe people can share experiences from race academies in other countries as well?

Hi Skitechniek,

Could you tell us what is your definition of transition? According to my previous answer it is the first or the second definition? This might help us to discuss this topic further with others

 

[Skitechniek]

I mean time required to start the next turn. I have been taught to delay this and be patient. Like I said, the only reason to change edges fast is to keep on skiing in the right direction, not because it is faster. If the course set gets easier I try to stay on my base for as long as I can. As the course set gets tighter, I transition as fast as I can. Not because it is fast, but because otherwise I won't make the gate.

This is what I have been taught and which makes sense to me from a physics point of view. Edging/turning slows you down, because the resistance of edges is much higher than the resistance of the base of the ski.

Fun fact: When I was still racing lots of ski racers used to tune away as much edge as they possibly could, because lots of metal makes the ski slower. They probably still do this, wouldn't surprise me.

 

[Scruffy]

I don't think you can define good skiing by transition time; too many variables that are all correct for the all levels of skiing.

 

[Snowcookie]

I don't think you can define good skiing by transition time; too many variables that are all correct for the all levels of skiing.

I agree - transition time can not be used alone to directly define good skiing, but it is for sure one of the crucial parameters to do this. We started with a question from @Josh Matta about how long should transition last in general. I did not find him asking if that defines good skiing. We can provide obviously some transition time ranges for various skiers at some well defined slopes and conditions, but as you have mentioned - there are simpy too many variables to describe it with some simple rules.

 

[Snowcookie]

I mean time required to start the next turn. I have been taught to delay this and be patient. Like I said, the only reason to change edges fast is to keep on skiing in the right direction, not because it is faster. If the course set gets easier I try to stay on my base for as long as I can. As the course set gets tighter, I transition as fast as I can. Not because it is fast, but because otherwise I won't make the gate.

This is what I have been taught and which makes sense to me from a physics point of view. Edging/turning slows you down, because the resistance of edges is much higher than the resistance of the base of the ski.

Fun fact: When I was still racing lots of ski racers used to tune away as much edge as they possibly could, because lots of metal makes the ski slower. They probably still do this, wouldn't surprise me.

Could you describe your definition of transition time in more details. I think it differs from mine. As I understand you correctly you want to be "patient" and to postpone next turn as long as possible, because taking turns slows you down. I agree with that. So basically you want to go straight as long as possible. From my perspective transition time is between two consecutive turns, if they are really one after another, up to some maximum delay. Making turns with a delay in between them larger that this max value from my perspective would be a completely different evolutions sequence: turn - straight - turn, not turn - turn with transition time in between. I would say you started a completely new, independent turn outside of some sort of streak of turns. How about that @Skitechniek ?

 

[karlo]

I would rather know what my body motions are


For example, am I hip dumping? Is my pelvis stable? Those, like me, who are rehabilitating from an injury, or those who never figured it out in the first place, would benefit much more from that than where the edges are. Ideally, the bodysuit has strain sensors that, when I do something wrong, activates my shock collar

 

[Skitechniek]

I'd say it's probably from where you release the pressure to where the pressure starts again. But maybe it is easier to show with video.

This is what I would call rushed skiing:
(from 0:18 onwards)

And this is what I would call patient and flowing skiing:
(from 4:05 onwards)
(this is me btw)
(from 0:30 onwards)

Forcing the edge change and edge angle VS. flowing and waiting for pressure.
Max edge angle before the fall line VS. progressively increasing edge angle till after the fall line

In the attached files I have posted a graph from an interesting study, which shows a transition time from 0,7 seconds on an easy course with low speeds. 0,7 is a long time. That time will definitely go down once the course set gets more challenging. But when you have a choice between long transition time vs. short transition time, the transition will always take as long as possible in ski racing. That is at least what I've learned and what is faster from personal experience.

 

[Snowcookie]

I would rather know what my body motions are


For example, am I hip dumping? Is my pelvis stable? Those, like me, who are rehabilitating from an injury, or those who never figured it out in the first place, would benefit much more from that than where the edges are. Ideally, the bodysuit has strain sensors that, when I do something wrong, activates my shock collar

Hi @karlo,

Nice to have you in this thread. I am glad you have mentioned this technology, as I have worked with it for more than six years, wrote my PhD based on it and thanks to it I joined the Snowcookie team You can see in this paper that we have tested our system extensively in one of the best human motion laboratories in the Europe in collaboration with a great researchers team. Unfortunately, Vicon is hard to apply outside lab, on the slopes – it uses passive markers and light reflections, which in snow conditions behave really bad. Moreover, setting up such a system on the slope is extremely difficult, as this hardware (cameras, processing unit, network devices) is very fragile and must be precisely setup.

You also mentioned a bodysuit that you could wear. You won`t believe it, but I have also work on that technology designing and developing firmware and algorithmic part for it (quaternions algebra for analyzing, compressing and de-noising rotations in joints with lifting scheme - my PhD topic). You can check the results here. This technology is still in development stage as it is a research project, but some commercial solutions are already available if you are interested in, i.e. Rokoko. Costumes are much easier to use on the slopes, but this technology has its weaknesses too. This is a cloth with IMU sensors within it (accelerometer, gyroscope and magnetometer). As all clothes it is moving around our body and we have no control on it. Considering skiers dynamics you can imagine what can happen - recorded data can be useless or hard to interpret. With proper data analysis we can fix it to some point but still – costume might be uncomfortable for skiers and is prone to various damages (usually broken cables, if any, holes in pockets for sensors, other communication issues and so one). There are also problems with cleaning those clothes, with batteries, calibration and many more. If you are interested in details I will be happy to answer all your questions about it.

As for our system – we are currently working with data collected close to sensor attached places – skis and chest. What you were mentioning (hip dumping, stable pelvis) are parts of kinematic chain that we cannot observe directly. We are working on dedicated motion models that learn to estimate what is happening in between chest and skies based on recorded forces, angular velocities and orientations as a function of skiers speed. As this is very complex task, it might appear as soon as winter season 2021. On the other hand, placing sensors on hips/pelvis will allow easily to analyze mentioned behavior based on vibrations (accelerations). If you go through this paper you will notice, that at those places we have mainly frequencies in range 4-10 Hz, where our sensors can work easily with 100Hz sampling, allowing to analyze signals from 50Hz up, according to Nyquist–Shannon sampling theorem. But skip this signal processing theory – what I wanted to say is that either we build dedicated motion models analyzing mentioned behavior indirectly, based on what we have right now, or we could scale up our system in the future with additional sensors (within a belt?) to observe those places directly. In the end – this is not something we could not do, just a matter of time, founds and user requirements that have to convert in the end into a revenue.

And the feedback - if you have all the data, it is just a matter of your imagination how would you like to get it

Cheers
Mati

 

[Noodler]

@Skitechniek - it's way harder to ski (takes more skill) to make the turns in the first set of videos (what you call rushed) than the mostly park-n-ride stuff shown in the later videos.

Let's not muddy up what could be a fairly nice product for recreational skiing with discussions of what movements are fastest in the gates. That's completely irrelevant when it comes to defining what most recreational skiers want; speed control with smoothness, good balance, lack of grinding the bottom of the turn, etc. If you want to be the fastest down the mountain, then don't turn.

 

[karlo]

Hi @karlo,

Nice to have you in this thread. I am glad you have mentioned this technology, as I have worked with it for more than six years, wrote my PhD based on it and thanks to it I joined the Snowcookie team You can see in this paper that we have tested our system extensively in one of the best human motion laboratories in the Europe in collaboration with a great researchers team. Unfortunately, Vicon is hard to apply outside lab, on the slopes – it uses passive markers and light reflections, which in snow conditions behave really bad. Moreover, setting up such a system on the slope is extremely difficult, as this hardware (cameras, processing unit, network devices) is very fragile and must be precisely setup.

You also mentioned a bodysuit that you could wear. You won`t believe it, but I have also work on that technology designing and developing firmware and algorithmic part for it (quaternions algebra for analyzing, compressing and de-noising rotations in joints with lifting scheme - my PhD topic). You can check the results here. This technology is still in development stage as it is a research project, but some commercial solutions are already available if you are interested in, i.e. Rokoko. Costumes are much easier to use on the slopes, but this technology has its weaknesses too. This is a cloth with IMU sensors within it (accelerometer, gyroscope and magnetometer). As all clothes it is moving around our body and we have no control on it. Considering skiers dynamics you can imagine what can happen - recorded data can be useless or hard to interpret. With proper data analysis we can fix it to some point but still – costume might be uncomfortable for skiers and is prone to various damages (usually broken cables, if any, holes in pockets for sensors, other communication issues and so one). There are also problems with cleaning those clothes, with batteries, calibration and many more. If you are interested in details I will be happy to answer all your questions about it.

As for our system – we are currently working with data collected close to sensor attached places – skis and chest. What you were mentioning (hip dumping, stable pelvis) are parts of kinematic chain that we cannot observe directly. We are working on dedicated motion models that learn to estimate what is happening in between chest and skies based on recorded forces, angular velocities and orientations as a function of skiers speed. As this is very complex task, it might appear as soon as winter season 2021. On the other hand, placing sensors on hips/pelvis will allow easily to analyze mentioned behavior based on vibrations (accelerations). If you go through this paper you will notice, that at those places we have mainly frequencies in range 4-10 Hz, where our sensors can work easily with 100Hz sampling, allowing to analyze signals from 50Hz up, according to Nyquist–Shannon sampling theorem. But skip this signal processing theory – what I wanted to say is that either we build dedicated motion models analyzing mentioned behavior indirectly, based on what we have right now, or we could scale up our system in the future with additional sensors (within a belt?) to observe those places directly. In the end – this is not something we could not do, just a matter of time, founds and user requirements that have to convert in the end into a revenue.

And the feedback - if you have all the data, it is just a matter of your imagination how would you like to get it

Cheers
Mati

I’ll PM you with an idea. If it’s novel, please let me invest.

 

[Snowcookie]

I’ll PM you with an idea. If it’s novel, please let me invest.

Sound good to me Waiting for your message then!

 

[Snowcookie]

@Skitechniek - it's way harder to ski (takes more skill) to make the turns in the first set of videos (what you call rushed) than the mostly park-n-ride stuff shown in the later videos.

Let's not muddy up what could be a fairly nice product for recreational skiing with discussions of what movements are fastest in the gates. That's completely irrelevant when it comes to defining what most recreational skiers want; speed control with smoothness, good balance, lack of grinding the bottom of the turn, etc. If you want to be the fastest down the mountain, then don't turn.

You are right there that currently we are mainly for recreational skiers but also for the aspiring ones In the near future however we would like to propose solution for pros too We keep learning from each piece of information you post here

 

[Philpug]

I was hoping Mt. Rose would be open from the top this past weekend. I will get our test model out this coming weekend.

 

[Snowcookie]

I'd say it's probably from where you release the pressure to where the pressure starts again. But maybe it is easier to show with video.

This is what I would call rushed skiing:
(from 0:18 onwards)

And this is what I would call patient and flowing skiing:
(from 4:05 onwards)
(this is me btw)
(from 0:30 onwards)

Forcing the edge change and edge angle VS. flowing and waiting for pressure.
Max edge angle before the fall line VS. progressively increasing edge angle till after the fall line

In the attached files I have posted a graph from an interesting study, which shows a transition time from 0,7 seconds on an easy course with low speeds. 0,7 is a long time. That time will definitely go down once the course set gets more challenging. But when you have a choice between long transition time vs. short transition time, the transition will always take as long as possible in ski racing. That is at least what I've learned and what is faster from personal experience.

Ok @Skitechniek,
I think we are back on track We are seeing transition time in a very similar way, but we are looking at it from different perspectives - you see it as a pressure changes (results) and I see it as a body dynamics (source of the changes you observe). Additionally, you see transition time between each consecutive turns, when I consider transition time between turns, but up to some defined maximum duration. After that I consider that this combination of turns was separated by a straight section and for the second turn instead of a transition time I would call it initiation time.

From body dynamics perspective I define the transition time as a time between a moment when you start transferring your legs (and whole body - CoM) in opposite turn direction, which leads in the first place to lowering your current turn edge angle, than allow you to ski for a reasonably short time with flat skies straight and later on to gain edge angle in opposite direction, up to a moment, when your skis start truly rotating around vertical axis in the next turn direction. Hope that is clear. I think that this plays nicely with your interpretation based on pressure release (start ending the current turn) and gain (begin turning in opposite direction) apart being patient where I start to consider straight section and cut turns strike to two independent sets.

How do you feel about it?

 

[Snowcookie]

I'd say it's probably from where you release the pressure to where the pressure starts again. But maybe it is easier to show with video.

This is what I would call rushed skiing:
(from 0:18 onwards)

And this is what I would call patient and flowing skiing:
(from 4:05 onwards)
(this is me btw)
(from 0:30 onwards)

Forcing the edge change and edge angle VS. flowing and waiting for pressure.
Max edge angle before the fall line VS. progressively increasing edge angle till after the fall line

In the attached files I have posted a graph from an interesting study, which shows a transition time from 0,7 seconds on an easy course with low speeds. 0,7 is a long time. That time will definitely go down once the course set gets more challenging. But when you have a choice between long transition time vs. short transition time, the transition will always take as long as possible in ski racing. That is at least what I've learned and what is faster from personal experience.

Could you also provide the title (or full text, link to it) of the work you have presented the chart with edge angles? Have not seen that or I do not remember it, while it looks really interesting