Bringing IoT to sports analytics | Hacker News Reader

Bringing IoT to sports analytics

blog.acolyer.org

63 pointsgvb13 comments

The resolution of UWB radio ToF measurement is not 15.65ns, but 15.65ps. That corresponds to a 64GHz counter.

I wonder which UWB radios they use for positioning. It's clear that this is IEEE 802.15.4 UWB PHY with the ranging extension, but I'd love to know what is the chip vendor.

EDIT: read the original paper, it's Decawave, same chips that I use :-)

Out of curiosity, what do you use it for and do you have any project you can share?

For positioning and communication :-) — unfortunately I can't provide any more details than that right now.

I figured. That's why I said "any project you can share" :) Thanks for taking the time.

I've actually thought about using those Decawave chips for tracking players in hockey. You could mine a wealth of analytics from this including who wins puck battles and face-offs, zone entries and exits, player acceleration and speed, how player dynamics change through a shift and through the game, etc.

Couple technical issues with this kind of system though, including position resolution (maybe Kalman filter with an IMU to improve?), determine players that are active on the ice as opposed to being on the bench (maybe position based), tracking all players and puck simultaneously, plus making the hardware small enough for placing it in the players clothing and in the puck.

Business problem would be in trying to get teams signed up to collect data. Getting data from one team would be good, but from both teams and the puck would be best. You almost need to get a league to agree to use the system. Then you have to deal with data ownership issues, CBA agreement, etc. Plus the whole barrier of people in hockey thinking this stuff is useless and unless you've #playedthegame you shouldn't be making the decisions.

There are a few companies doing something similar. Zebra technologies is using RF tracking in the NFL (they do warehouse tracking normally). A bunch of others try to do it through IR systems with cameras or video processing of broadcast video, but these methods have their own issues (occlusion, HW needs to be visible, etc.).

In any case, the wealth of data and statistics you could mine would be fascinating to say the least.

I believe NFL already heavily uses UWB sensors for analytics as well. Could someone describe how this is different? Is it just extending to the rules and semantics of cricket?

It's used in rugby too, but it's only player tracking involved. That's a much simpler problem and, as mentioned in the paper, is fairly easily solved.

I think the USP here is being able to measure the path, velocity and spin of an object in flight, and how this is much more easily done using instrumented equipment, rather than ultra-expensive cameras.

I can imagine this solution could also be applied to many other ball sports, eg baseball (as mentioned), golf, tennis, etc. Maybe also for T&F throwing events.

In my own sport, rowing, equipment to measure stroke angles and force curves is become ever more readily available and cheap (important in a sport with not much money flowing around). Low-cost ubiquitous IoT in sport is definitely a growth market.

Sorry should've elaborated. I know there are UWB RFID tags in NFL player shoulder pads. I was also under the impression that the ball itself is tracked. Is this not the case?

I imagine a football is easier to sneak sensors into than a cricket ball though. Restrictions of the ball itself. Which I realize now that you call it out.

Out of curiosity, could you link to some of the rowing sensors? A quick google brings up http://www.smartoar.com/ i'd be interested in seeing what there is in the way of cheap options you've used.

http://www.nkhome.com/rowing-sports/empower-oarlock

This comes from a manufacturer of standard rowing equipment and is priced at a level suitable for small programs or individuals.

2 others that are currently under development:

http://www.sweetzpot.com/

https://www.kickstarter.com/projects/1605381989/the-rowing-a...

Both of these use smartphone apps for the UI.

I wrote my thesis on biomechanics analytics under sports equipment department, which primarily focused on improving the motion for people suffering with lower body injuries. I think the field need more exploration I believe with advent of AI in this field can even boost its reach.

For those who're not familiar with Cricket, it's very similar to baseball, to first order. The most substantial differences that I can think of:

Instead of throwing the ball directly towards the batter's upper body, it must usually bounce once on the ground. There must always be two batters, one at each end of the strip. The batters whack the ball and run up and down the strip instead of around a diamond. Each batter keeps going till they get out (and they don't get out for playing and missing a few times).

And, of course, refer: https://en.wikipedia.org/wiki/Cricket

The key difference between throwing and bowling is that a bowler's elbow must not extend or hyperextend at an angle greater than 15 degrees during the bowling action.[1]

If interested in cricket analytics there are two very nice R packages available to support this.[2][3]

Cricsheet is a good source of data.[4]

[1] https://en.m.wikipedia.org/wiki/Bowling_(cricket)

[2] https://cran.r-project.org/web/packages/cricketr/index.html

[3] https://cran.r-project.org/web/packages/yorkr/index.html

[4] http://cricsheet.org

Finally, something interesting coming out of the Gentlemen's game. Nice read.

Reply on news.ycombinator.com