Login required to started new threads

Login required to post replies

Prev Next
Re: The Aero Bike shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
kileyay wrote:
SharkFM wrote:
Sorry - this may have been talked about already did you guys test with rider only at 50 kph??

Bike only vs bike only will certainly see differences. But here a less aerodynamic bike could be better with a rider aboard as it is designed (or should be designed) to blow air around the rider's body parts.

We tested both at 30mph, so, yeah basically.

15 pages of ad hominem and speculation. Yawn. Wake me up next year or whenever you actually have data to present......,

My latest book: "Out of the Melting Pot, Into the Fire" is on sale on Amazon and at other online and local booksellers
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
Nicko wrote:
I think you should release all data now, just de-indentify the bikes.

I thought about this. I'm definitely going to. But not until a week or two before the report comes out.

If I left the values on the y axis but de-identified the bikes, that would give away a big part of the surprise, wouldn't it? Part of the idea is to find out how different these bikes are (or aren't)

Nicko wrote:
It can't hurt to have several minds looking into this, no?

There are some minds looking into it. Are you volunteering? I have de-identified the data for those who are looking into it, but maybe you have some unique thoughts on same

Nicko wrote:
Best case, you'll get the analysis protocol peer-agreed before you release the 'white paper' ... (hate that expression, write a report dammit).

Agree. It's a report...have slipped and called it a white paper a few times, but it isn't.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
Ok.
Release all data for one bike now, de-identified.
Let any nerd make whatever analysis and presentation they feel is best. Submit. Discuss. Without dogs in the fight..
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
Nicko wrote:
Release all data for one bike now, de-identified.
Let any nerd make whatever analysis and presentation they feel is best. Submit. Discuss. Without dogs in the fight..

I'll do one better. Here is my Felt. Baseline run was first of the day and control was second to last in the day (late afternoon). I don't think this gives anything away -- it does show fairly consistent data.

This is still preliminary.

Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
"It's a report...have slipped and called it a white paper a few times, but it isn't."

cervelo knows exactly how you feel.

Dan Empfield
aka Slowman
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
kileyay wrote:
Nicko wrote:
Release all data for one bike now, de-identified.
Let any nerd make whatever analysis and presentation they feel is best. Submit. Discuss. Without dogs in the fight..

I'll do one better. Here is my Felt. Baseline run was first of the day and control was second to last in the day (late afternoon). I don't think this gives anything away -- it does show fairly consistent data.

This is still preliminary.

Hmm, if I remeber correctly 0.001 CdA difference is roughly a watt so you have an approximately 2-6 watts difference between the first and last run, not to mention the positive difference on one side versus negative difference on the other.

I think I voted for a 6-10 watt difference across best to worst bike, so either there is a bigger spread between bikes or your report is going to be difficult to write with definitive conclusions. I am hoping for the former! Good luck.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
kileyay wrote:
Nicko wrote:
Release all data for one bike now, de-identified.
Let any nerd make whatever analysis and presentation they feel is best. Submit. Discuss. Without dogs in the fight..

I'll do one better. Here is my Felt. Baseline run was first of the day and control was second to last in the day (late afternoon). I don't think this gives anything away -- it does show fairly consistent data.

This is still preliminary.
Wow.

That is not data. At best, it's a sales persons picture of another knowledgeable persons data aquisition, filtering, pre-processing, averageing and presenting.

True "raw data" means IMO each and every individual, physical sensor recording. For every 10ms (or whatever sampling frequency used). Millions of data points.
Layman "raw data" could mean the calculated axial drag force (in 'grams', sic!) from a combination of sensor readings, snipped in time and averaged down to a single number data point.
The spreadsheet mentioned earlier (with hundreds of lines...) is an accumulation of such data points.

What you have in your 'picture' is another level of abstraction away from the raw data, with measured drag force scaled by measured dynamic pressure. To top it off, it's without beta-square correction and placed in 'yaw categories' instead of a proper, numerical x-axis... please...

Consider giving us "real raw data". Only then can the analysis protocol be scrutinzed in open air.
Nothing to hide, right?
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
These may be fair points but settle down...this is probably while Kiley didn't want to post "preliminarily"...because everyone will jump all over him based on information that may look nothing like the final product. Give him time; there will be ample opportunity to flame him later.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [DFW_Tri] [ In reply to ]
Quote | Reply
I asked him to publish "raw data" in some form, for one de-identified bike, without analysis, tweaking or pretty presentation. The kind of data that doesn't need justification, explanation or a 'preliminary' tag...

If the purpose of this massive effort was to present "the truth" in contrast to the "marketing" done by the evil forces (Cervelo et al), everything has to be public. The analysis protocol could/should be discussed before the brand-specific numbers are presented and the trenches gets dug.
Last edited by: Nicko: May 5, 17 9:00
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
Can a small amount raw sample data be released?

All we really need to know is which measurements were recorded and their units. From this a peer review of the normalization formulas can happen, and no results are given away.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
I understand what you are saying but at some point, there are too many cooks in the kitchen. Let them do their work and then you can blast all you want if you don't like how it is presented after receiving hopefully the raw data and summary.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
I think you're really going to be disappointed by the final product, even if I release all the 'raw' data I have available -- that which was given to me on a thumb drive from A2. Maybe I can get more, which I have asked for, without success thus far, but I think you are possible the only person who is going to be upset by what I put out -- at least, the only person this upset. Here's the data in the rawest form I have for the baseline run. I'm sorry you're so outraged. FFS man...give me a break. The intent is not to obfuscate -- it's to make it accessible to people who don't have your quantitative sophistication...which is to say, almost everyone. Talk about out of touch.



I'm happy to take your suggestions, input, and help. But not if you're going to be borderline abusive.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
Nicko wrote:
I asked him to publish "raw data" in some form, for one de-identified bike, without analysis, tweaking or pretty presentation. The kind of data that doesn't need justification, explanation or a 'preliminary' tag...

If the purpose of this massive effort was to present "the truth" in contrast to the "marketing" done by the evil forces (Cervelo et al), everything has to be public. The analysis protocol could/should be discussed before the brand-specific numbers are presented and the trenches gets dug.

Who made you, like, boss of the world?
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [Nicko] [ In reply to ]
Quote | Reply
Nicko wrote:
I asked him to publish "raw data" in some form, for one de-identified bike, without analysis, tweaking or pretty presentation. The kind of data that doesn't need justification, explanation or a 'preliminary' tag...

If the purpose of this massive effort was to present "the truth" in contrast to the "marketing" done by the evil forces (Cervelo et al), everything has to be public. The analysis protocol could/should be discussed before the brand-specific numbers are presented and the trenches gets dug.

Did you donate to the gofundme?

You know you can always go to the wind tunnel and do your own testing.

blog
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
kileyay wrote:
I'm happy to take your suggestions, input, and help. But not if you're going to be borderline abusive.

What about simply putting the contents of the thumb drive in a Dropbox, pictures and all ?
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [grumpier.mike] [ In reply to ]
Quote | Reply
grumpier.mike wrote:
Hmm, if I remeber correctly 0.001 CdA difference is roughly a watt so you have an approximately 2-6 watts difference between the first and last run, not to mention the positive difference on one side versus negative difference on the other.

I think you need to look at the difference on average and deviation from the mean, which, on a percentage basis, is between +/ 0.4 % and +/- 0.7 % depending on which method you use (of the two I've come up with, anyways). Besides all that, the baseline and control reflect taking the bike off and putting it back on, hours apart. Also, I wrote more about this here and here.

And hey, if these bikes are all within those error margins, that's a definitive enough conclusion for me that I should just pick the bike I find hawt or makes me unique or has the most storage or is most affordable or whatever.

The 'sources of error' section will be one of the longest sections. When you add up every potential source of error -- for instance, Chris from Flo will tell you the same brand/SKU of two brand new tires and differ by something like 7 watts, and this is just one potential source of error -- you can easily say none of this matters at all, that we'll never know which bike is fastest, etc. That's up to everyone else to determine from what we present. And whatever you conclude may be different from what I conclude, or what I've concluded already. Which is fine too.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
I'm not outraged, just slightly disappointed... You'd think PubliusValerius of all would understand trial-by-internet...
Like your tone in this outburst.

Your first try at presenting data ends up as poor as your public beheading of the TriRig Omni by the flawed method of "sideways bending by hand". "LOOK, I CAN SEE IT FLEX! FFS!!".
With your history of vile criticism of most everything, I expected a lot more from you when you were going to "do it right". Not Coggan level reporting, or RChung insightful, but at least factually correct and very open about shortcomings.

Look, I honestly tried to initiate a discussion about the quality and granularity of typical wind tunnel data sets and procedures. The data set sample you have is what it is. Not your fault. You could have presented that weeks ago.

Sorry I called you a 'sales person', that was mean...

edit:spelling
Last edited by: Nicko: May 5, 17 11:08
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
kileyay wrote:
grumpier.mike wrote:
Hmm, if I remeber correctly 0.001 CdA difference is roughly a watt so you have an approximately 2-6 watts difference between the first and last run, not to mention the positive difference on one side versus negative difference on the other.

I think you need to look at the difference on average and deviation from the mean, which, on a percentage basis, is between +/ 0.4 % and +/- 0.7 % depending on which method you use (of the two I've come up with, anyways). Besides all that, the baseline and control reflect taking the bike off and putting it back on, hours apart. Also, I wrote more about this here and here.

And hey, if these bikes are all within those error margins, that's a definitive enough conclusion for me that I should just pick the bike I find hawt or makes me unique or has the most storage or is most affordable or whatever.

The 'sources of error' section will be one of the longest sections. When you add up every potential source of error -- for instance, Chris from Flo will tell you the same brand/SKU of two brand new tires and differ by something like 7 watts, and this is just one potential source of error -- you can easily say none of this matters at all, that we'll never know which bike is fastest, etc. That's up to everyone else to determine from what we present. And whatever you conclude may be different from what I conclude, or what I've concluded already. Which is fine too.

Not criticizing, just commenting. I think the one thing I am learning is that this is really hard to do well because of many factor (variability, repeatability, fatigue,...) and it is not as easy as the casual observer believes it should be. We would all like to believe that if you go to a multi-million dollar facility and hop on your bike, you will get the same basic measurements one visit to the next. I guess this is why Cervelo alway took their "tare" bike with them and why Specialized decided to just build their own.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [grumpier.mike] [ In reply to ]
Quote | Reply
grumpier.mike wrote:
Not criticizing, just commenting. I think the one thing I am learning is that this is really hard to do well because of many factor (variability, repeatability, fatigue,...) and it is not as easy as the casual observer believes it should be. We would all like to believe that if you go to a multi-million dollar facility and hop on your bike, you will get the same basic measurements one visit to the next. I guess this is why Cervelo alway took their "tare" bike with them and why Specialized decided to just build their own.
+1. The error rate in these tests is definitely higher than people realize. Again, not criticizing, just commenting. It's a good lesson for those who like squinting at wind tunnel tests.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
Effect size with confidence intervals of the effect size!

If you have enough data to estimate variability in measurement (which depends mostly on how many repetitions of data you have that are comparable using the same bike), you can use this to estimate the confidence you have in the effect of changing to a different bike. If that CI (usually 95%CI in my field, but it would make sense to look at something more like 90% probably) doesn't include zero difference then you can be confident that the bike is different. He nice thing about this approach is that the effect size is the best estimate of how much the bike matters or differs regardless of how wide the confidence interval is.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [grumpier.mike] [ In reply to ]
Quote | Reply
I haven't been able to read the entire thread so pardon my laziness, but has anyone mentioned that some of the bikes have the extensions mounted underneath while others mount on top? Noticed that from some of the pictures. That is likely to skew results.

Trent Nix
Owned and operated Tri Shop
F.I.S.T. Advanced Certified Fitter | Retul Master Certified Fitter (back when those were things)
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [BudhaSlug] [ In reply to ]
Quote | Reply
BudhaSlug wrote:
Effect size with confidence intervals of the effect size!

If you have enough data to estimate variability in measurement (which depends mostly on how many repetitions of data you have that are comparable using the same bike), you can use this to estimate the confidence you have in the effect of changing to a different bike. If that CI (usually 95%CI in my field, but it would make sense to look at something more like 90% probably) doesn't include zero difference then you can be confident that the bike is different. He nice thing about this approach is that the effect size is the best estimate of how much the bike matters or differs regardless of how wide the confidence interval is.

Yup. The graph seems to indicate the between run variability can be pretty high (2-6 watts) compared to the within run variance. If setting the bike up on the jig and rider fatigue during the day,... have that high a variance, you need to do multiple runs to average across the between run source of error, which nobody does. Put it another way, you could have 0 within run variance, but you may need the replication of setting up the bike multiple times to determine a difference between bikes.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [trentnix] [ In reply to ]
Quote | Reply
Wow, lot of value to that post. About as much as this post.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [trentnix] [ In reply to ]
Quote | Reply
Did anyone mention that each bike has a different base bar. That's likely to skew results.

Did anyone mention that the P5 has hydraulic Magura brifters and the others are all Shimano? That's likely to skew results.

Did anyone mention that some bikes have full internal routing...

We could go on. And on. And on. All specs will be listed in a table in the report and you can make of it what you will and shit on it accordingly.
Quote Reply
Re: The Aero Bike Shootout TEST DAY THREAD [kileyay] [ In reply to ]
Quote | Reply
kileyay wrote:
Did anyone mention that each bike has a different base bar. That's likely to skew results.

Did anyone mention that the P5 has hydraulic Magura brifters and the others are all Shimano? That's likely to skew results.

Did anyone mention that some bikes have full internal routing...

We could go on. And on. And on. All specs will be listed in a table in the report and you can make of it what you will and shit on it accordingly.
But I thought you guys were installing aftermarket extensions in order to control the testing better? I think you went to a lot of trouble to match extensions from bike to bike, I was just surprised they weren't mounted similar. I'd love some reasoning as to why you did not, despite knowing it would likely significantly impact comparability of your tests.

You guys set the standards you're being evaluated on. You identified all of the questions you were trying to answer. You insisted your testing would be objective, implying that manufacturer testing is subjective.

You were extremely defensive when others expressed doubt you were prepared to test things in an objective, comparable fashion. And now you're extremely defensive when others are pointing out some fairly impactful things that might skew your results. Either you overlooked those things or you considered the cost-benefit and made a choice. Instead of taking the opportunity to explain those decisions, you're throwing rocks.

I'm sure plenty hope you fail at your stated mission, but some of us actually hope you succeed.

Trent Nix
Owned and operated Tri Shop
F.I.S.T. Advanced Certified Fitter | Retul Master Certified Fitter (back when those were things)
Last edited by: trentnix: May 6, 17 11:29
Quote Reply

Prev Next