login
FTP/speed correlation (Page 2)
-
No new posts
|
 2016-06-20 3:20 PM in reply to: 3mar |
 812   Katy, Texas |  Subject: RE: FTP/speed correlationOriginally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. |
| 2016-06-20 3:55 PM in reply to: Jet Black |
1502 Katy, Texas | Subject: RE: FTP/speed correlation Originally posted by Jet Black Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. |
| 2016-06-20 4:03 PM in reply to: 3mar |
439 nashville, Tennessee | Subject: RE: FTP/speed correlation Wow this thread took quite a crazy turn! We are getting quite an engineering/physics lesson. |
| 2016-06-20 4:14 PM in reply to: mchadcota2 |
Not a Coach 11473 Media, PA | Subject: RE: FTP/speed correlation Originally posted by mchadcota2 Wow this thread took quite a crazy turn! We are getting quite an engineering/physics lesson. Actually, what you are mostly getting is lesson on data usage. A lesson even more valuable most of the time. |
| 2016-06-20 6:04 PM in reply to: 3mar |
Extreme Veteran 3025 Maryland | Subject: RE: FTP/speed correlation Originally posted by 3mar Originally posted by Jet Black I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. then you would be wrong. just look at the drag force equation. it takes a hugely great power increase to increase your speed when you start getting up to 24/25/26 mph play on bike calculator. your fit line just won't work |
| 2016-06-20 6:04 PM in reply to: Jet Black |
Extreme Veteran 3025 Maryland | Subject: RE: FTP/speed correlation Originally posted by Jet Black Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. yes as long as you remove outliers I could agree with this. but you would definitely have to keep the same gear, rider weight, etc |
| 2016-06-20 6:59 PM in reply to: dmiller5 |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation Originally posted by dmiller5 Originally posted by 3mar Originally posted by Jet Black I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. then you would be wrong. just look at the drag force equation. it takes a hugely great power increase to increase your speed when you start getting up to 24/25/26 mph play on bike calculator. your fit line just won't work Dude...bike calculator is a theoretical model. Seriously. I think this one is going over your head. For the range of data between 20-25 mph, within any reasonable range based on elevation and weather, a 1.5-2 mph error for any given power is probably giving it more credit than we can. Law of averages or not. That is simply the area of change from the variables that stay variable. At that bandwidth, in this range of speeds....no. it doesn't make a difference. A straight line is fine. |
| 2016-06-20 7:02 PM in reply to: 3mar |
Extreme Veteran 3025 Maryland | Subject: RE: FTP/speed correlation Originally posted by 3mar Originally posted by dmiller5 Dude...bike calculator is a theoretical model. Seriously. I think this one is going over your head. For the range of data between 20-25 mph, within any reasonable range based on elevation and weather, a 1.5-2 mph error for any given power is probably giving it more credit than we can. Law of averages or not. That is simply the area of change from the variables that stay variable. At that bandwidth, in this range of speeds....no. it doesn't make a difference. A straight line is fine. Originally posted by 3mar Originally posted by Jet Black I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. then you would be wrong. just look at the drag force equation. it takes a hugely great power increase to increase your speed when you start getting up to 24/25/26 mph play on bike calculator. your fit line just won't work yes, if you're fine with being off by 10% then sure, a straight line is fine |
| 2016-06-20 7:08 PM in reply to: dmiller5 |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation Originally posted by dmiller5 Originally posted by 3mar Originally posted by dmiller5 Dude...bike calculator is a theoretical model. Seriously. I think this one is going over your head. For the range of data between 20-25 mph, within any reasonable range based on elevation and weather, a 1.5-2 mph error for any given power is probably giving it more credit than we can. Law of averages or not. That is simply the area of change from the variables that stay variable. At that bandwidth, in this range of speeds....no. it doesn't make a difference. A straight line is fine. Originally posted by 3mar Originally posted by Jet Black I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. then you would be wrong. just look at the drag force equation. it takes a hugely great power increase to increase your speed when you start getting up to 24/25/26 mph play on bike calculator. your fit line just won't work yes, if you're fine with being off by 10% then sure, a straight line is fine I think you missed the one with the bar drawn over showing error. I'm pretty darn sure not a single point between those speeds will ever fall out of that box. So no. It won't be 10% off. It's about understanding your error and how you can apply data. Just because you can measure something to 0.000000 doesn't mean you can use it to 6th decimal. This is precisely why in A-Chem I always walked out of lab in half the time of anyone else and still had better results. It's about knowing when you can and should go to the 100th and when you can eye ball it because that's as good as it will get. |
| 2016-06-20 7:28 PM in reply to: 3mar |
Extreme Veteran 3025 Maryland | Subject: RE: FTP/speed correlation Originally posted by 3mar Originally posted by dmiller5 I think you missed the one with the bar drawn over showing error. I'm pretty darn sure not a single point between those speeds will ever fall out of that box. So no. It won't be 10% off. It's about understanding your error and how you can apply data. Just because you can measure something to 0.000000 doesn't mean you can use it to 6th decimal. This is precisely why in A-Chem I always walked out of lab in half the time of anyone else and still had better results. It's about knowing when you can and should go to the 100th and when you can eye ball it because that's as good as it will get. Originally posted by 3mar Originally posted by dmiller5 Dude...bike calculator is a theoretical model. Seriously. I think this one is going over your head. For the range of data between 20-25 mph, within any reasonable range based on elevation and weather, a 1.5-2 mph error for any given power is probably giving it more credit than we can. Law of averages or not. That is simply the area of change from the variables that stay variable. At that bandwidth, in this range of speeds....no. it doesn't make a difference. A straight line is fine. Originally posted by 3mar Originally posted by Jet Black I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. then you would be wrong. just look at the drag force equation. it takes a hugely great power increase to increase your speed when you start getting up to 24/25/26 mph play on bike calculator. your fit line just won't work yes, if you're fine with being off by 10% then sure, a straight line is fine your box has a range of 10+% |
| 2016-06-20 7:38 PM in reply to: 0 |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation Originally posted by dmiller5 Originally posted by 3mar Originally posted by dmiller5 I think you missed the one with the bar drawn over showing error. I'm pretty darn sure not a single point between those speeds will ever fall out of that box. So no. It won't be 10% off. It's about understanding your error and how you can apply data. Just because you can measure something to 0.000000 doesn't mean you can use it to 6th decimal. This is precisely why in A-Chem I always walked out of lab in half the time of anyone else and still had better results. It's about knowing when you can and should go to the 100th and when you can eye ball it because that's as good as it will get. Originally posted by 3mar Originally posted by dmiller5 Dude...bike calculator is a theoretical model. Seriously. I think this one is going over your head. For the range of data between 20-25 mph, within any reasonable range based on elevation and weather, a 1.5-2 mph error for any given power is probably giving it more credit than we can. Law of averages or not. That is simply the area of change from the variables that stay variable. At that bandwidth, in this range of speeds....no. it doesn't make a difference. A straight line is fine. Originally posted by 3mar Originally posted by Jet Black I'd still say the spread due to those variables would make no difference between a linear or curved fit in this range of data. I'd wager the spread is about 1.5 - 2 mph at any given power. Between 21-25 mph, a straight line vs curved won't make a difference. Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. If you kept all of your variables the same, then with enough data points the variables you can't control (wind, hills, etc) would become insignificant in the data due to the law of averages. then you would be wrong. just look at the drag force equation. it takes a hugely great power increase to increase your speed when you start getting up to 24/25/26 mph play on bike calculator. your fit line just won't work yes, if you're fine with being off by 10% then sure, a straight line is fine your box has a range of 10+% And that's the point. I'd argue that trying to get any more accurate (a prior) that can be used across multiple routes is as good as it gets. The variable variables will shift it that much in that range. It's not wrong. It's the opposite. I still don't think you're getting that. Understanding the uncertainty of data and extrapolation is a key to doing something right. That's why we have safety factors. Real life is never a line. ETA; note I said "line" not "linear" just want to throw that out there preemptively. Edited by 3mar 2016-06-20 7:42 PM |
| 2016-06-20 8:45 PM in reply to: 3mar |
| 439 nashville, Tennessee | Subject: RE: FTP/speed correlation You both are so far off. If you look at the P value of the significant outliers, it is so negligible that factoring any of the variables will not impact the velocity based on the vector. Judging by the inversely proportionate variables, taking the inverse tangent of the curve in question, you will likely end up with a range of 21-28 mph, with an acceleration of 9.8 m/squared. With that in mind, on a scale of 1 to 10, I will be biking very fast. |
| 2016-06-20 9:58 PM in reply to: 0 |
| 643 | Subject: RE: FTP/speed correlation FTP != speed. More so since as you go faster, aero dynamics come into play. Currently I'm at 294 FTP with a 20 min effort, which is about 20 watts higher that my previous FTP before a race. I can't hold 294 FTP for an hour, I know this. I can't hold any FTP I ever had for an hour. Just not happening. I just use it as a base measurement for my intervals. I am able to do 4x10, 3x15, and 2x20 @ FTP though. I'm nailing down my HIM pace and did a 2 hour ride at 230 watts and plan to do a 3 hour ride at 230 watts this week (on trainer). Anyway...back to the question: Sprints I get 21ish mph. All the sprints around here are very technical with maybe a 1-2 mile straits. The route has us going through schools, circles, etc. For Redman (HIM), I averaged 23.2 mph @ 210 watts. For Boulder half, I averaged 23.6 mph @ 205 watts due to the thinner air and less bike traffic/turns. for the KC Olympic I averaged ~22.4 mph @ 240 watts. I weight ~165 and am not fully aero. I could use a drop in my front and also I have my bottles in back but I do have an aero helmet and race wheels. Aero helmet helped the most to get "free" speed, after my tri bike/aero bars of course. I don't think there's going to be an easy way to correlate FTP + ? * weight -> to speed. The course, gear, air temp (warmer air is thinner), and aero dynamics play too much of a role. Edited by Blastman 2016-06-20 10:01 PM |
| 2016-06-21 12:44 AM in reply to: Blastman |
| Subject: RE: FTP/speed correlation Originally posted by Blastman I don't think there's going to be an easy way to correlate FTP + ? * weight -> to speed. The course, gear, air temp (warmer air is thinner), and aero dynamics play too much of a role. There actually is a way to correlate power to speed if you can identify the variables. The line in the first reply addresses most of them quite well based on my experiences. |
| 2016-06-21 6:03 AM in reply to: mchadcota2 |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation touche |
| 2016-06-21 6:37 AM in reply to: 3mar |
Extreme Veteran 5722 | Subject: RE: FTP/speed correlation Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. Are the points at the lower wattage truly rides at 170w or are they rides at 140-170 and some 230 to come to an average of 170 ? |
| 2016-06-21 7:05 AM in reply to: marcag |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation Originally posted by marcag Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. Are the points at the lower wattage truly rides at 170w or are they rides at 140-170 and some 230 to come to an average of 170 ? They're all steady state. Since I'm tapering I got a low end data point this morning. 19.1 mph @ 128 watts (111 BPM). By the way; it's really nice to just go out there and spin and enjoy the scenery every once in a while. |
| 2016-06-21 7:47 AM in reply to: 3mar |
Extreme Veteran 5722 | Subject: RE: FTP/speed correlation Originally posted by 3mar Originally posted by marcag Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. Are the points at the lower wattage truly rides at 170w or are they rides at 140-170 and some 230 to come to an average of 170 ? They're all steady state. Since I'm tapering I got a low end data point this morning. 19.1 mph @ 128 watts (111 BPM). By the way; it's really nice to just go out there and spin and enjoy the scenery every once in a while. Interesting. If the data is truly linear you may want to look into the reason why. |
| 2016-06-21 8:07 AM in reply to: mchadcota2 |
Expert 2355 Madison, Wisconsin | Subject: RE: FTP/speed correlation Originally posted by mchadcota2 I'm curious how other triathlete's FTP correlates with their actual speeds. Would anyone be willing to share their FTP(and method of testing), and their typical speeds in races? I know that speed varies a lot based on other factors but just a range would be good. I would just like to know how mine compares to others to see if I need to make any changes. Here's mine: FTP(8 min trainerroad test)- 272 Typical racing speed- 22-23 mph for sprint tris Average power is how fast you are going Normalized power is how hard you are riding Keep those 2 numbers as close as possible and you will ride as fast as you can without going too hard. |
| 2016-06-21 8:25 AM in reply to: marcag |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation Originally posted by marcag Originally posted by 3mar Originally posted by marcag Originally posted by 3mar That's kind of the point though. The data will never be perfect. It is what it is. When I look at it, I see it like this, with the shaded area being range of error, because that's all you'll ever get. In the range of error you get for outdoor rides, which is what we're trying to correlate to, you'll never see a curve, so why bother? It's exactly what happens when us engineers try to out engineer ourselves. A line is fine. Anything else would be outplaying the data and variables with outdoor rides. Perhaps I should have done that in the first place. The data can't be perfect, because it can't apply everywhere. For what I use it for, this is more than accurate enough. Also, I use it for interpolation rather than extrapolation and only use that anecdotally. Are the points at the lower wattage truly rides at 170w or are they rides at 140-170 and some 230 to come to an average of 170 ? They're all steady state. Since I'm tapering I got a low end data point this morning. 19.1 mph @ 128 watts (111 BPM). By the way; it's really nice to just go out there and spin and enjoy the scenery every once in a while. Interesting. If the data is truly linear you may want to look into the reason why. I know. I plotted this morning's ride and was like WTF? I was expecting to see at least some curve 130-250 watts. (Capture.JPG) Attachments ---------------- Capture.JPG (69KB - 5 downloads) |
| 2016-06-21 9:19 AM in reply to: 3mar |
Extreme Veteran 5722 | Subject: RE: FTP/speed correlation off topic, but what kind of tires do you use ? |
| 2016-06-21 9:33 AM in reply to: marcag |
| 1502 Katy, Texas | Subject: RE: FTP/speed correlation Conti 4000 SII's. They both have latex tubes as of yesterday (for this weekend's HIM). I actually still had one latex tube in there (I guess I never changed it out after my last race) so I had been riding one butyl one latex. The rest of my set up is a Felt B16 with no name Chinese carbon 80 mm wheels with a wheel builder cover on the back. I never take the cover off. I put it on over a year ago and it has never rattled or came loose so I've never wanted to mess with it. Also, it has a cool batman logo sticker on it. The only real difference between racing and training is that in training I wear separate tri shorts and a tri tank top (sleeveless) with a non-aero helmet, and for racing or TTs I wear a sleeved once piece tri suit and an aero helmet. Everything else stays largely the same on every ride. |
|
|
|
|
| |||
|
View profile
Add to friends
Go to training log
Go to race log
Send a message
View album
CONNECT WITH FACEBOOK