Smash: The game changing tennis wearable | Hacker News Reader

Smash: The game changing tennis wearable

kickstarter.com

126 pointsevjan69 comments

Varsity tennis player here. I have been using my fitbit bracelet for a couple months now, I have to admit that the device is inaccurate. I am pretty sure it undercounts running distance and number of steps. We are talking about big wide steps that happens in a consistent manner. A players movement on court is much more complicated. Such a bracket for tennis players needs to be much more accurate than my fitbit flexband to be useful. A good tennis player never takes a wide big step, almost all the movements on court are quick small steps. Unlike running, a player changes direction all the time. It would be really cool if the bracelet can give the player data about his footwork but I think the emphasis on this particular product is on the actual technique. Man its not easy. Sure there might some easy wins in the beginning, like counting how many forehands or backhand a player hits or things like how hard have you been hitting the ball. But I think its extremely hard to dig more into it. “How much spin are you giving”, “Are you hitting the ball late or early", “How fast can you recover when you hit the ball(combine it with footwork)”, “How high is the ball when you hit it” “Are you bending your wrist too much, or is you elbow really close to your body”. “Are you hitting the ball just with your arm or your shoulder and your waist follows through with the movement" These would be the questions that an intermediate tennis player would be interested in. However I am afraid these all happen in a fraction of a second and I really doubt a bracelet is accurate enough to give the player insightful data. Sure its a really cool device (still not useful) for people starting to learn or maybe little kids competing against each other, but there is a long way until a serious tennis player can benefit from this. I am a software engineer myself and it was always my dream to do build something about tennis. So go for it man..

Thanks for that, appreciate it. I've put all the effort into breaking down events within strokes and writing algorithm's for these. You're right, it is tough but I wanted to ensure it provided useful insight

Have you seen this? Babolat racket has the sensors built in: http://en.babolatplay.com/features. Tennis players might prefer this to wearing a bracelet on their playing arm.

My wife plays a lot of tennis, and I bought the Babolat racket for her for Christmas. Great idea, but it's implementation is awful, primarily because the software is complete crap. We're using it with Mac/Android, so maybe the Windows/iPhone version is better, but the software problems are pervasive and make it pretty much unusable. We called Babolat about it and they were completely unhelpful--their support staff didn't seem to be trained to deal with anything other than physical rackets. This is all fixable--but fixing it would require a company that's great at building rackets to learn how to build great software... unlikely, and there's no sign that it's going to happen. A device like the Smash (or other separately wearable tech) is more likely to do the software right.

Beyond Babolat's bugs, there's a deeper issue: it shows you a lot of data... but it's not clear that that data is actually useful. So 24% of your hits are on the upper quadrant of the racket... so what? Is that good or bad? If bad, what advice would a coach give you to improve? This drove my wife crazy: when we occasionally got her racket to sync properly she was rewarded with a flood of numbers that meant nothing to her. Smash's video says it will "make recommendations like a coach would"--which is what you actually want, so hopefully they are able to deliver on that promise.

I agree-- I feel like there's a long way to go before something like this becomes more valuable than simply watching yourself play on tape.

I don't tennis, but it looks pretty cool.

At the risk of thread hijacking, I have a broader question. My assumption is that most wearable bracelets have the same tech/sensors inside and the difference is in how your software interprets said sensor signals. So why doesn't someone create an open bracelet that people can hack and customize for any sports/fitness application?

I mean, I would have loved something like this for baseball back in the day (both for pitching and hitting). There are wearables for golf too[0].

So, why not just a generic wearable + SDK/app store?

[0] http://mashable.com/2014/05/22/zepp-golf-sensor-review/

Hmm, perhaps an iWatch would fulfill such a role

I'm still willing to bet a steak or sushi dinner that Apple will not release an iWatch. A wearable bracelet, maybe. But not a watch.

What do you call a bracelet with a touch screen? Semantics, anyhow.

Agree, it's as valid as calling your pocket touchscreen computer a "phone".

I believe that they're already doing tooling for it at the manufacturing company where I intern. So don't make that bet with too many people

Foxconn intern?

a watch is a wearable bracelet that just happens to tell time, and often a few other things.

Watches made the transition from time piece to jewelery many years ago. With the time being ubiquitously available watches have been relegated to being a "wearable bracelet" (is there any other kind?) - A Rolex submariner, for example, does the exact same thing as a $5 plastic cheapy, but people still buy Rolex and all the other > $5 watches as jewelery (i.e. social signalling of status)

Why wouldn't they include the time on a wearable bracelet?

Personally, I can't agree enough with this. Positioning the wearable as a "iBracelet" means that it can be work at the same time as a consumer's watch; whereas an "iWatch" potentially limits the market.

That said, I can completely understand the position of making an iWatch and forcing the choice between existing watches and Apple's. However, to me it's better to competing with fitbit than rolex.

The bet has to have a deadline or noone will be eating steak or sushi.

No deadline means he can never win the bet, but whoever he is betting with can, so seems win win ;)

The market for a generic would probably be smaller. My guess is each application probably shares basics (like an accelerometer) but has different requirements (accuracy, noise, processing power, sample rate) if not entirely different sensors (maybe one needs GPS and one needs a barometer). Which means the generic bracelet that would work for every application would have to be the superset of all possible requirements- and that's $$$$.

But you probably could market one bracelet for a couple different sports, so long as they are similar & you can write the software. Baseball & tennis might overlap a lot, for example.

I don't think accuracy of IMU's out there vary dramatically, and the noise shouldn't matter too much (all major ones within some +-3 db?) -- sure, if they were a few orders of magnitude more accurate you would get a whole new type of applications -- but with current ranges you're restricted to some kind of pattern recognition (i.e. not full motion tracking), which is quite noise resilient.

By "noise" I wasn't just referring to noise in the sensor. PSU noise, for example, needs to be controlled to below some limit. That increases the BOM count and thus $$$.

I agree. An open bracelet would be cool but from a bussiness point of view it doesn't make that much sense. There are countless of startups who refused to focus on a core target market and failed. Instead what you should do is focus your development efforts on the core product and listen to community feedback for cool ideas, then provide the necessary API for them to build it. Which is basically what these guys are doing.

The Pebble has the potential to be the generic wearable you describe. It's accelerometer and CPU are very similar to the FuelBand, and it's easy to write apps for. I'm not sure why it hasn't taken off as a sport platform. Perhaps there are software limitations or it's difficult to monetize apps.

Or perhaps there are physical constrains like the durability. On pebble's website it does mention being scratch resistant but I don't know if it can endure the frequent impact it will encounter in many sports.

I think smash aims to be that sort of wearable. From the kickstarter page:

We’re planning to open up an API to developers. There’s going to be a truckload of data, and whilst we have our own visions of where Smash can go, we wouldn’t want to restrict others from creating greatness. We're keen to move Smash into more sports and create more insightful features for players.

Oh man, you have no idea how much this gnaws at me. Especially with the Myo.

So, for a bit of background, the Myo has 6 (or was it 9?) EMG channels, and at least a 6 axis IMU (I think it might have magnetometer as well). The developer API they plan on releasing will only give you the post processed signal + the raw IMU signal - that is to say, they'll tell you the current affinity to each of the pre-programmed gestures (~6 or so) that Thalmic has set up. There is no way to normally get access to the EMG data.

They will however give you units with 'special firmware' if you sign up for their researcher partnership program which costs $10k+ the last time I talked to them (this winter). Needless to say, this is clearly aimed at university research labs.

I can't really say just how frustrating this is, as someone just leaving undergrad, and with a passion for sport and science/engineering. The worst part is just how understandable it is.

Nearly all the players in this field are startups of some sort. Their #1 goal is to get stable positive cash flow out of some niche market. As a hardware startup, they have significant costs to overcome. There just is no incentive for them to split off some effort to try to capture this type of open system/market.

But seriously, I crave an open hardware platform for this. But at the same time, it's obvious that the costs are non-trivial. The 'science' part of the payload isn't even hard. IMU + microprocessor + bluetooth. The hard part is hardware packaging, getting it into a form factor that has enough battery life, will stay in position on the body, will not adversely bother the user, or modify their movement. All of these issues are non-trivial and have to be solved for hardware release.

Seriously, I wish I just had a pile of cash, and a couple years to ticker. sigh

I work for MapMyFitness, and we're aiming at the software side of the equation. We used to be a startup until we were acquired by UnderArmour, so now we're starting to plot out projects that will be taking years to pay off.

Drop me a line if you want to talk more: dave.mclain@mapmyfitnessinc.com

I'm one of the co-founders at Thalmic (Myo).

We get the request for raw muscle data quite often, and are trying to find a way to handle it better in the future. The reason we don't provide it now actually has nothing to do with money (we do have a small number of research groups that pay for this, engineering support, and other services, but it represents a tiny fraction of a percent of revenue).

The reason is actually two-fold:

1. User Experience - Unfortunately you're probably the exception, in that it sounds like you may be able to build a great experience using the raw data. The reality we've found is that it requires a deep machine learning and statistical analysis background to do much useful with the data, considering the need to account for a wide variety of variations in data intra and inter person. What we're trying to prevent is developers attempting to create "custom" gestures improperly (e.g. if (myo.sensor1 > 67% && myo.sensor2 > 30%){output = Rock_On}), which will not work reliably on the population, in different orientations, etc., then the user has a bad experience, and blames the device hardware.

2. Battery Life - Streaming raw data uses much much more power than when we process the data on the device, and recognize gestures directly on-board.

We are considering options to provide this data in some way, as there are certainly good uses in some applications, such as Tennis analysis.

You could quite easily do everything the Smash device does, plus additional information on grip strength, using a Myo armband. It's all about the software!

It's great to hear that. I'm sure you've heard this already, but there are non-trivial amounts of part-time researchers who are salivating over your tech, principally as a tool for scientific exploration, or just tickering. These aren't so much people coming out a developer background, but those coming from a research background, people who can totally understand and stomach difficulties associated with EMG.

Basically, I believe there to be a great deal of people who are in similar situations to your current research group partners in technical aptitude and expectations, but differ in their financial resources - which is why your product in particular is so attractive in the first place.

I really look forward to seeing how this develops.

Tennis player here..ummm... As legohead brought up, there are other things to consider like stance. I can see how this thing can be accurate if we had four of them (ankles and wrists). This can show show most of the moving parts, can replay how the body moves (likely hand location and elbow extention), open vs close stance, how fast were moving from point A to point B.

I don't know how they're measuring ball rotation, purely by the angle of an upward swing? Don't know if this is accurate. Is this an area it would suggest for improvement? RPM has a large range up to ~5,000rpm (Nadal). What would it even cosider optimal to make a shot?

Don't know how the device would know if a ball would land in or not, is this even a major area of concern for the device? Pretty strokes are one thing, but unless the ball lands in, it doesn't matter.

Depending on how accurate this device is (probably would need more than one bracelet per person), it can be used for any sports. More importantly, there's a possiblity to measure bad form and increased pressure on certain joints to prevent injuries.

Looking forward to see technologies brings forth in the coming years!

Shot Stats[0] is another Tennis related kickstarter currently going on.

[0] https://www.kickstarter.com/projects/laviesak/shot-stats-cha...

Attaching to the racquet does seem like a better idea. I'd be more inclined to believe it could do things like measure the speed and angle of the head better, than at the wrist.

Hi, just thought I'd jump on here and answer a few questions. Thanks all for the interest!

You're right in saying the components are pretty ubiquitous, the key part is the algorithms which interpret the data. They would need developing for each sport or particular movement, plus a UI to replay back the insights in a user friendly way. We spent months on this with Smash, it was the real focus of the early work, and will continue to be so. Some kind of platform that allows this to be done easily feels like a great next step...got me thinking :)

Rob

I find your product interesting and just did some brief research. From a tennis perspective, what would differentiate your product from Zepp? (http://www.zepp.com/tennis/)...besides the fact that Zepp is on the racquet while Smash is worn on the wrist. Since Zepp is a mature product and can be purchased from the store or online already...have you had a chance to try it and compare it to your Smash prototypes? Thanks.

I have tried it yes. Our difference is the focus on technique and providing insight. I think Zepp just released a feature on serving which looks pretty cool, I would imagine they're heading down the same path. The key is the experience, it's got to be simple and useful

Former tennis player here, and builder of a similar device for weightlifting form. Unlike a lot of other wearable kickstarter projects, I think most of the technological claims here are legit. That said, I'm not sure if it does enough to be useful for an avid tennis player. To improve my game, I would like to see:

1) Swing form correction

2) The angle of the racket relative to an incoming ball (to predict air or ground balls)

3) Stance information.

Form correction would require position data during the entire swing. Position information is two integrals away from acceleration data collected from the sensors, so it's far too noisy to determine that well enough.

I don't know how would you would measure stance with a bracelet, and I'd imagine that racket angle is unreliable.

I also question the ball speed measurement for a similar reason to form correction. I'd imagine a possibly effective way to measure speed would be to correlate acceleration data with externally measured velocity data collected in real-life trials. That said, unless you're competing at an advanced level, speed is a vanity metric so I don't care if its a ballpark estimate.

Video recording, however inconvenient it may be, addresses all these well. I'd rather see a camera added to tennis courts to record your performance, but that's far less sexy of a product to build.

That said, the bracelet looks sexy and the goal of personal training sure beats the hell out of fitbit/fuel/jawbone's capabilities.

Pair wearable tech with future iterations of Tango/Kinect and you're onto something even bigger.

Tennis for example -

Think about a computer picking up flaws in technique that the human eye would miss; a coach would only be looking at your leg work or arms or upper body but a computer will tell you a multitude of things - wider stance, shoulder's turned to the right a bit more, less follow through, more weight on your front foot etc etc - then imagine it coming straight to your ear every time you place a shot. Pair it up with a ball tossing machine and you've got a coach that knows precisely where you're going wrong, how to fix it and giving you the best possible advice and opportunities to do so (forehand shoddy? that's all your getting until it improves etc).

The next decade has just gotten so much more exciting. I'd love to find someone somewhere working on a "complete" system, not separate bits of it - we already have video analysis, wearables, weight sensors (wii fit?), someone just has to slap them all together.

Side question - aren't there golf swing analysis places already which use a myriad of tech to give feedback?

This is exactly what we are working on at CoachSeek. We are creating a complete management system for coaches, which will in time integrate with a number of wearables to give sports coaches a central place to manage all their client data. http://www.coachseek.com

I would look into this more. I was watching the French Open just yesterday and I heard Darren Cahill say that top juniors now a days have the technology so that when they walk off the court it will tell them what strategy worked what percentage of the time and what they should have been doing more. Pretty interesting comment, although they never did show an example of it I totally believe it. Babolat[0] pure drive meets Sony[1] racket sensor.

The only project I ever built with an Arduino is a wristband with a piezo sensor and it counts the number of balls I hit while was practicing. I thought about expanding it more, but that was the only data point I was really interested in.

0. http://www.tennis-warehouse.com/Babolat_PLAY_Pure_Drive/desc...

1. http://www.theverge.com/2014/1/20/5326558/sony-smart-tennis-...

This is a very interesting concept. A side-note: while receiving immediate feedback is useful, a coach is still incredibly useful. They tell you how to improve the one stat and more importantly broader strategy (where to stand, where to hit, etc).

I wonder what other sports this can be applied to. My guess is that swimming is a prime candidate and maybe shotput/javelin or baseball pitching.

A few come to mind: boxing (incl thai boxing and MMA), golf, ice hockey, bandy. So many swingin' sports!

A smart boxing glove that can iteratively learn the required knock-out force for different opponents?

I think we're just about to be in the world of putting accelerometers on everything.

Probably applicable to a ton of sports. Golf?

I have a high res accelerometer + 10hz gps in my race car, too.

I made a similar product (albeit a lot less refined) for a hackathon, using the Pebble. It counted number and type of tennis shots, displayed useful HUD data, and then let you see pretty graphs in a desktop webapp. Some problems I ran into:

If you have acceleration, to get position, you have to integrate twice, which means two unknown constants, really throwing off your data accuracy. To fix this, the user would have to go back to a defined position (perhaps the tennis 'ready' position) frequently, but that isn't guaranteed or user friendly. They could absolutely work around this with some refined heuristics, but I wasn't able to get one working in the few hours I had.

The algorithms are the hard part. I was using naive machine learning techniques to identify swings and such, and they were not nearly adequate; I had many false positives. If they can get all of the stats they claim, I'll be very impressed.

It seems that no one else has a problem with this, but I'm still not completely sold on Kickstarter projects that essentially function as pre-order mechanisms for products in development. Shouldn't this money be coming from investors or lenders?

Yeah, Kickstarter is pretty clear that they're not aiming to be a pre-order site. And there has been a lot of discussion about the fact that backers have no real stake in the company, especially after Oculus got acquired by Facebook.

I am skeptical about this device becoming a game changer in tennis. I played tennis for a while and what really helped me then was video analysis. While data is helpful, it's not as helpful as video. If the wearable could depict my form and technique from data, that'll be a game changer. For game improvement, we don't lack knowledge. After playing any game for a while, we'll know our limitations - but to tide over those limitations you need hours and hours of practice along with a coach. During those rigorous practice hours, your mind gives up and you just want to quit, but your coach won't let you do that :).

Playsight does this:

http://www.usatoday.com/story/sports/tennis/2013/11/28/play-...

As a tennis player I don't see how this is going to help me. At least the video & description of the product don't explain it very well.

I've been teaching my daughter tennis, and the issues she has (and that I have as an amateur) need very specific attention -- how you are holding the racquet, follow through, feet position, anticipating ball placement..

The description says it gives "tips and advice" - like what? Can it tell me I need to rotate my grip 10 degrees? That my backhand starting swing position is too high? That I am trying to hit the ball too soon or late? etc.

Thanks for the feedback, really appreciate it. I envisaged the player profile being the place to set your grip, e.g semi western. It's a reference point I've been using to develop the algorithm's. The device can help you with follow through, point of impact, backswing. Footwork and movement is another story and is kind of what I have in mind for phase 2, doing a bit of investigation with iBeacons but line of sight is tricky

Interesting from a gadget standpoint - I wonder if any "regular" tennis players will be interested. I am reminded of this very clever lean startup experiment about a similar concept (though it was in the racquet, not a wearable).

MVP experiment using a walkie-talkie to fake a "smart racquet": https://www.youtube.com/watch?v=2eJbu4EtHMk

Results of the experiment: https://www.youtube.com/watch?v=gZzYUW3-JHQ

What I don't understand yet, is what hardware on this band makes it different from other wristbands (besides form factor)? If all the calculations are done on the mobile app, and the wristband is just a data collector, why not implement this app to integrate with existing bands?

We've seen so many wristbands (and now watches) releasing, it doesn't make sense to have separate bands for separate tasks. We'll eventually have a similar problem where every website has it's own app.

I was sort of expecting a responsive bracelet that would actively help you. For example by vibrating more or less to help you swing properly (stay on the correct path, buzz until complete follow through etc, buzz when you are at correct spot in serve to help you time the ball etc etc).

This looks more like a good, smart application specific flight recorder. Very cool, but not as game changing cool as I first thought it would be.

Keep it up though - it is a great game!

Isn't this the same thing - just as a bracelet.

http://preorder.moov.cc/

Apart from the dedicated Tennis application.

This looks very well thought out and phenomenally useful for general fitness alone, with the potential to handle all sorts of applications, including tennis analysis. It's unfortunate that they're incredibly vague about what "summer launch" means, considering summer is only a few weeks away.

It is also very unfortunate that Android compatibility will not exist for another 4-6 months at the earliest. Focusing on a single mobile platform isn't something that should still be happening these days, especially when Android's market share dwarfs that of iOS. I look at the decision to focus on iOS in 2014 as a highly questionable business decision, at best.

Yeah, this is way better. The tennis one is too focused, and the market is too small.

I'm a nerd and love data... so I am happy with the response to this question being "because it's cool"... but is anyone actually reaping any reward beyond the basic placebo effect by wearing any of these devices? It might make me want to train more just to see the data... but is the data really helpful?

I wonder if we'll be looking at these devices like Nike Pumps in a few years... all gimmick.

Consistency is a major factor in pretty much any sport. It's difficult to improve your performance if you don't have a consistent baseline from which to diverge, just like how you change only a single variable when using the scientific method. I know little about tennis mechanics, but I imagine a hypothetical experiment could be something like "Do I get a lower bounce if I put the angle the racket slightly later during the swing?" Being consistent is the first step to being able to test that theory.

Gathering data during practice and play allows you to know if you're being consistent, and if you're not, when those inconsistencies occur. Once you've hammered out the inconsistencies, you can start to experiment and improve your technique.

Weight lifters make meticulous notes about their workouts. Even amateur race car drivers spend thousands of dollars on data logging equipment. Professional sports players spend hours watching videos of themselves. At a high level, it's all about data and analysis to provably improve technique.

I am not questioning the value of data collection... I guess I am more questioning the data quality and applications.

I suppose it's less relevant if we agree some data is better than none. I also wonder about the relevance to the weekend warrior. Is this something that really will improve their game? Or is it just a fun way to engage in a hobby? Maybe it doesn't matter from a market viability standpoint.

> I also wonder about the relevance to the weekend warrior. Is this something that really will improve their game?

Yes, if they're interested in improving it. For basically all the reasons coldpie has already pointed out. It's true that you'd need to practice specifically as well as play games--it's much harder to develop higher consistency from the dynamic environment of a real game--but that's not unusual.

I'm hoping that something like the Vidonn X5 ( http://www.vidonn.com/en/ )gets open sourced - it's $40 device, inc postage, and seems to have everything it needs to do most of what I want... except the app is pretty poor.

If you could use this with your OWN apps then it would be a killer device.

I quit working on this same idea about a year ago after concluding that the resolution on (existing) 9-axis IMUs wasn't good enough for serious tennis players.

Then again, maybe just being able to see your racquet head speed increase as you improve your technique will be enough for a "quantum of utility" for beginners.

Good luck to them!

This is a great idea! I've just pledged 140 AUD. I don't play tennis, but I do play a lot of badminton.

My hope is that they'll eventually give software devs access to an API so that I may refine their product to work for badminton.

It looks like they are planning on doing so. From the Kickstarter page:

"We’re planning to open up an API to developers. There’s going to be a truckload of data, and whilst we have our own visions of where Smash can go, we wouldn’t want to restrict others from creating greatness. We're keen to move Smash into more sports and create more insightful features for players."

Really, really, really impressive design. But I think this is a really, really, really small market, and I don't think there will be enough interest to meet the fundraising goal. Hope to be proven wrong, though.

I wonder if it's possible to build Smash for the Google Glass, to get instant feedback instead of looking at the app post-game. Actually, it is possible, and I hope it happens soon.

I'm curious if there are any computer based training tools for tennis that make use of a high speed camera and do image processing to analyse your swing etc.

I love that logo.

~$200 for wearable accel/gyro chip? You have got to be kidding me.

It is the price of a quality tennis racquet and comparable to running GPS watches which perform a similar function (though with different hardware and a different sport). Not to mention tennis is a pretty pricy sport. Frankly $200 sounds cheap to me.

Reply on news.ycombinator.com