Hey everybody, I’ve been away from the community for a while now due to life distracting me but I’m back and I wanted to share something that I worked on about a year ago, right before I took a break from throwing.
I’m not sure if anyone has posted anything like this but this was a hypothesis test that I ran in order to test the significance of a concave bearing on the average spin time of a throw. I was writing this for a non-thrower so the intro is very low level so he would understand the background.
I won’t spoil the results, go ahead a take a peek!
Let me know if you guys have any questions, I’d be happy to clarify things.
Interesting. Seems to me that the variation of your tests is large enough to indicate the quality of your throw is not consistent. I don’t see how you can have a 40 second sample and a 100 second sample and consider the testing method to be good - that is a huge variation in the quality of the throw. Couple that with the small sample size and I’m not sure this result can be counted on. The longest throw is on the flat bearing. Given the conditions for the throw aren’t changing (string, wind, pad condition, etc), the only difference in each sample is the quality of how you threw it.
It’s all kind of a non-starter because of the variables.
A high-quality manufacturer of a flat bearing might produce something that spins longer than a shoddily-manufactured concave-profiled bearing. At which time the determining factor is physical quality rather than profile.
If you had two bearings from the same manufacturer, with the same materials and produced to the same tolerances, with the only difference being profile… then the “throw” would have to be identical. Mechanically reproduced, with the same string and same response, etc…
But part of the reason many people prefer profiled is that it automatically centers the string and keeps it from precessing and also from rubbing against yoyo walls/pads. If the mechanical throw was successful in centering the string on a flat bearing, the profile wouldn’t really be as much of a factor anymore.
I just think there are too many variables that you’re not in strong control of.
When I was throwing I wasn’t doing any tricks, the throw consisted of a standard forward throw and no purposeful string contact. So I did not correct any off kilter throws. The main thing that I took from the variances is that the variance of the centertrac throws was much smaller, showing that any change in the pitch of the yoyo wasn’t as effective on the centertrac, I.e the string didn’t touch the response system as much.
As far as the sample size, the law of large numbers takes care of any concerns about the validity of the test. That being said, randomly I could have gotten a good sample or a bad sample, that’s just how the cookie crumbles. I would’ve loved to do more trials and I would totally be open to repeating the experiment, but at the time 30 of each was all I managed.
All that being said, I don’t have a throw like the pros, at the time I was fading out of the scene as I said so my throw wasn’t 100% what it was. But I feel confident in my test’s validity, at least for an avid enthusiast like myself.
When designing my experiment I planned on to account for controllable things such as string wear, so I used 2 Type X strings from the same package, 1 for each set of throws. As for the bearing selection, I chose those 2 because of their reputation in the community as being decent choices for bearings and their comparable price points. From what I can tell, the CTX is better quality than a standard YYF bearing.
If my personal throw is in question with this analysis, I would love to have someone with a perfected throw do the throws and if anyone wants to help me find identical bearings with different profiles to test. It would’ve also been better to have 20-25 of each bearing to eliminate the variability in the manufacturing process.
My throws in theory should be normally distributed. So my throwing both samples theoretically should have eliminated any bias in my throws.
I think as a starting point you would need to devise a machine to execute the “throw”. It doesn’t have to be particularly elaborate. Might just be letting go of the yoyo that has its slipknot hooked onto something. If you want it to be a high-speed throw, forget about the winding and use a simple motor rig to ramp the yoyo up to X RPM.
But again, part of the spintime equation when it comes to yoyo is the practical application of technique. Proponents of profiled bearings claim that reduced string friction is what allows the yoyo to perform longer without losing spin. If you just spin and then hang from a string, the entire premise of the bearing’s benefit is thrown out the window.
So, in that sense, your original “thrown by a human” at least introduces imperfection into the throw.
Or, your “throwing machine” can introduce angle variations up to a full couple of degrees.
I gotcha ;D ideally that would’ve been the plan. But my project was to test how the shape effects the spin time with respect to string walk on the bearing so that would remove shape from the picture and would be more of a quality of materials test. A robotic arm would be great for testing it. Somebody out there got one I can borrow?
Exactly my point. If you are doing this and getting 40 second differences, then the most important variable is the quality of your throw. Your math is on point, but the data is not. It’s a garbage in, garbage out thing. All you really have proven is that you need to work on the consistency of how you throw. The bearing had no bearing on those differences.
They are normally distributed for how you throw which is highly erratic.
The point of the curved shape is to lengthen the duration of the throw correct? The curve doesn’t care how bad of a thrower I was, right? So the bearing did its job, which was lengthen the duration of my throw, which my math found significant. Human error is what the shape is there for, so the shape did its job, so to speak.
The main thing here that I think they are trying to point out is repeatability. There is no way for a human to exert the exact same amount of force every time and to throw at the exact same angle. If you were to have it roll down an incline or something from the exact same point or something would have, (theoretically anyway) more consistent results.
No, the point is string centering when there are wraps of string. A no tricks throw isn’t taking advantage of this. Even if it did to some degree, the margin of error of your throw is far greater than than any benefit gained by the bearing. You pretty much need the aforementioned throwing machine or at least someone who can throw more consistently.
You could test spin lengths of different bearings by throwing one on a bearing seat and blowing it with compressed air for x seconds until it reaches say, 9k rpms (about the rate a yoyo thrown hard will spin for if I remember correctly) and then find how long different types spin.
A throwing rig or a rig to mimic the throw would then just be testing the differences between the two bearings without consideration for the shape. It wouldn’t provide any significance as to the shape of the bearing better centering the string.
My hypothesis was that given a forward throw, the bearing shape would account for some percentage of my error, which is what happened. I wasn’t testing the effect of the bearing on multiple string wraps as I didn’t feel repeating a trick hundreds of times was realistic to provide results worth looking at. Nor was I testing which bearing was going to spin longer in and of itself.
All I’m trying to test in that experiment was: does the curved bearing provide for a longer spin time, specifically by prolonging a throw, good or bad. From what forge said, no human can repeat the same exact motion every time, there’s bound to be a bad throw somewhere in there. A pro may do it once in a few thousand, me at that time was a different story I hand a few bad ones in 60 throws so I wasn’t testing the best thrower in the world. But what I did do was get a sample of lost of imperfect throws and if the bearing had a significant effect on my throws then for a good thrower it may just give them the extra half a second of their 1 bad throw.
I think your are overestimating that percentage. I would guess if you repeated the experiment you would get very different results simple because they effect of your throw is SO MUCH GREATER than the effect of the bearing. For example if you throw a little sideways, you will get a much shorter spin obviously (34 seconds vs 104 seconds - this is a massive difference and it’s random whether or not the happened on either side). In your small sample of 30 throws all it would take is a few extra to happen to occur on one side or the other to get the result you got. And if you look at the numbers, the really short ones occur more frequently on the flat bearing side - and it’s factual the bearing won’t matter if you throw it that far off.
KYO has done this with a method that gives consistent starting spins. I believe he had a machine that just dropped the yoyo, then measured the spin times and RPM. No doubt he will weigh in here on his method and results.
I believe his method was to measure spintimes of different bearings, period-- not whether the profile had an impact or not. But yes. He had exactly what you describe…!
Using a machine to create repeatable, consistent “throws” would be biased against non-perfect throws.
The assertion that:
“There is no way for a human to exert the exact same amount of force every time and to throw at the exact same angle. If you were to have it roll down an incline or something from the exact same point or something would have, (theoretically anyway) more consistent results.”
is exactly the point of bias. The idea that multiple human throws produces a “normal” distribution of throws seems more statistically accurate than a large-sample of “perfect” throws. If no human can make “perfect” throws, what conclusions can be drawn from a large sample of equally non-reproducible throws?
