In the thread you allude to, Dan has several times suggested real world tests like this, done in a velodrome.

I agree with you that results of such would add to, rather than subtract from, the fun of these discussions.

Dave,

Have you ever seen the power output for a PowerTap for a one mile ride, let alone a 10 mile ride? It is almost impossible to maintain a constant power on the bike when the engine is a human.

Whatever one does to try to "prove" a theory is subject to criticism and interpretation, which is why these discussions occur. If it were simple there would be no controversy. Those who try to make it simple open themselves up for criticism.

Dan's original post tried, if I remember rightly, to invoke a recognition of the complexity of these problems to put some of the issues and claims into perspective. I agreed with is thought, except in his discussion of weight. Then, in reply, he pointed out a complexity I had not thought about before, which then got me and many others thinking and started in a very interesting discussion. It was fun to me.

I can't see however, how the answer to this discussion could be found easily using a Velodrome and the real world. tough study to design and carry out, hence the need for mathematical models and theoretical discussion. It is part of the academic process.

---

--------------
Frank,
An original Ironman and the Inventor of PowerCranks

 
I think the point of velodrome-type tests would be to get a better real-world picture of the performance impact of varying components and such (over, say, waving single components around in a wind-tunnel), rather than to obviate the need for math and thought.

I agree that it is unlikely "the answer to this discussion could be found easily using a velodrome". However, I think velodrome tests could help inform and extend the discussion, and at least be a nice counterpoint to wind-tunnel tests.

Its a tricky loop that you get stuck in and I'm not sure there's really a way out of it. People won't completely believe mathematical models or windtunnel tests because given the technological limitations there is no way to accurately simulate 'real-world conditions.' Afterall, it really wouldn't be called a simulation then, huh? The benefit of simulations, depsite the fact that they can not describe every variable in the equation (a limitation in itself), they are or at least can be repeatable.

On the other side you have the so-called holy grail of tests, the 'real world conditions.' Trouble here is that things will never be exactly the same, and thus never repeatable. So, it would be theoretically impossible to do a velodrome test simply because while there may be no wind, no elevation change, no environmental difference, there can and almost always will be some other difference (human being the biggest, even in the same person).

So, you've got one way which is repeatable but not complete. You've got another that is complete, but not repeatable. Which way to go?

Oh, and yeah, one more thing. Suppose you can somehow create this utopian aero test, not sure how, but we'll just say its possible. No two people are 100% identical in build or style. Thus, whatever works for one person doesn't necessarily have to work for another, and now you're back to where you started, room for debate.

So, whats my point? You can argue this til you're blue in face. Meanwhile, the other guy has been on his bike and getting stronger despite his complete disregard for aero-ness. True, information is your friend, but if its a debate you're after, there are some pretty good ones at roadbikereview.....just search for campy vs. shimano.

BTW, Dan is right.

 
Rather than a tricky loop that should be avoided lest we get stuck, I prefer to think of it as a process of iterative refinement. Revising ideas, methods, and thoughts is a useful process and should not have a hard endpoint. Models are *about* something. When they're about real-world phenomenon they need to be informed by real-world observations. This does not mean you chuck the model the first time the tiniest fleck of recalcitrant data lands in your lap.

Math with no connection to anything in the world of experience is a purely intellectual exercise; a valuable and diverting exercise to be sure. Real-world observations devoid of theory and math is a purely practical exercise; also valuable and diverting. What's wrong with liking both? And letting each inform and revise the other?

I have to disagree that real-world tests are not repeatable. They are, in fact, repeatable, but with varying degrees of error. What we want is testability. We need repeatability to do this, with results showing correlation in excess of margin of error to be valid, right? Here's an example:

I've got 2 "aero" forks. One is the Acme Forkotronic, the other is the Yoyodyne Motherforker. My hypothesis is that there is NO performance difference between the two. To test this, I take several highly trained riders, build several pairs of bikes identical except for the forks, take them to a velodrome over the course of several days, and have them each ride several trials over a given distance alternating bikes. You're absolutely right in that conditions are far from identical for these tests. However, suppose I find that in every single trial the Forkotronic times are lower? Can this be explained entirely by margin of error (or random factors, of whatever you want to call it)? Yes, it can. However, I crunch the numbers anyway and end up with, say a 99%+ confidence interval that my hypothesis (that there is no performance difference) is false. So... have I proven the Forkotronic is better performing? Nope. But I've shown that it *probably* is better. Furthermore, and more germaine to the original discussion, if you go waving each fork around by itself in a wind tunnel and announce the Motherforker is better, I've got some credible alternate "real-world" observations (using a repeatable methodology) with which I can reasonably disagree.

Anyway, I think velodrome tests would be useful as (a) a counterpoint to wind-tunnel test and (b) to possibly help extend and continue, not end, the discussion on the theory side. Both good things in my opinion.


And finally, I think your points were well-taken until this:
>Meanwhile, the other guy has been on his
>bike and getting stronger despite his
>complete disregard for aero-ness. True,
>information is your friend, but if its a
>debate you're after, there are some pretty
>good ones at roadbikereview

This notion that anyone involved in a discussion is somehow doing so at the expense of training (instead of riding or whatever) seems to pop up in almost every theory debate around here. It's completely fatuous. Frustratingly so. If this mythical "other guy" has no intellectual life at all, and scampers back onto the saddle anytime life presents him with a stimulating conversation , I don't want to be that guy. I'm a well-rounded renaissance man after all (a little too round, maybe, but still). Sheesh!

And no, I don't like debate for the sake of debate, so I'm not going to read archived shimano vs. campy debates just to see people arguing. I like debate on topics I find interesting, ideally debate that challenges my assumptions and teaches me something, and, sometimes, let's me get the gratification of contributing in a positive manner as well.

Great googly moogly, I think I need a coffee break or something.

 
Perfect. So this suggests that when yoyodyne and acme send me eval copies of their new forks I can do similar "real world" tests without having to go to the greater expense and difficulty of booking wind tunnel time and I have a precedent to point to showing such tests can be valid. Right?

"However, suppose I find that in every single trial the Forkotronic times are lower? Can this be explained entirely by margin of error (or random factors, of whatever you want to call it)? Yes, it can. However, I crunch the numbers anyway and end up with, say a 99%+ confidence interval that my hypothesis (that there is no performance difference) is false. So... have I proven the Forkotronic is better performing? Nope. But I've shown that it *probably* is better. "

Bingo...that was the essence of what I wanted to say. Sure, you can do tests until you're sick and all your money is gone, and YES, it will be beneficial to you, BUT, there will always be some guy out there who has data to prove that you're wrong. Thats the devil about tests. Nothing is certain and is therefore always open to some level of interpretation. Yes, things are 'probably' accurate and 'probably' correct, but debate will always happen. That just leaves Company X and Company Y to bicker like two little kids.

I'm not saying that testing isn't useful. Its just that some of the debates over it are.

"This notion that anyone involved in a discussion is somehow doing so at the expense of training (instead of riding or whatever) seems to pop up in almost every theory debate around here. It's completely fatuous."

Actually, I wrote all that at work, so if the 'other guy' was indeed out training while I was writing my own nonsense, I'm screwed anyways because it means he probably isn't working and thus on his bike the whole damn day.

"B) you have the time and inclination to perform enough trials to avoid making a type II statistical error. "

That's the key, but you left out the larger factor - money. This will always be a problem with products developed for a sport like triathlon. There just isn't enough money involved in selling a limited number of carbon forks to justify spending a million bucks in testing each model.

Watching the Giro d'Italia, the importance of the engine was again emphasized to me. Compare that to an event in which the technology rules like the America's Cup. These teams spend a couple million each building two boats out of a total budget of up to $60 million. The rest is spent on refining and testing the boats, both theoretically in tunnels, tanks and computers and practically against each other day after day after day.

I guess all we need are a few billionaires who are also triathletes.

There is room in this world for both real world testing and mathematical models. To be valid a mathematical model should be able to predict real world results. One develops mathematical models to see if one understands what is really going on in the real world. If one has a valid mathematical model one can make improvements without the "need" for expensive guesswork, modeling, and trial and error. People are still developing experiements to see if Einstein's Theories stand up to real world experimental observation - so far they have.

It turns out that Michael Dell's wife Susan is a triathlete. Too bad she or he isn't interested in squandering their wealth to answer these small, in the big picture, questions.

---

--------------
Frank,
An original Ironman and the Inventor of PowerCranks

Mr. Not a rotor wrote: "You mean like this? "

Yes, that would be one way modeling can be used to help with real world questions. Now, the only question is whether the model is sufficiently accurate in predicting real world results and/or whether it can be improved by including additional variables or improving algorithms (or, is it accurate enough for this purpose).

---

--------------
Frank,
An original Ironman and the Inventor of PowerCranks