How To Calculate Your Patt Percentages

[old] Porkchop · Post by **[old] Porkchop** » August 30th, 2005, 11:01 am

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(bmoore @ Aug 29 2005, 08:17 PM)</b></div></td></tr><tr><td class='quote'><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 29 2005, 08:51 PM)</b></div></td></tr><tr><td class='quote'>dense <br />adj. dens·er, dens·est<br /><br />1. Having relatively high density.<br />2. Crowded closely together; compact: a dense population.<br />3. Hard to penetrate; thick: a dense jungle.<br /><br />1. Permitting little light to pass through, because of compactness of matter: dense glass; a dense fog.<br />2. Opaque, with good contrast between light and dark areas. Used of a photographic negative.<br />3. Difficult to understand because of complexity or obscurity: a dense novel.<br />4. Slow to apprehend; thickheaded.<br /><br />PC,<br /><br />I was being facetious. <br /> </td></tr></table><br /><br />Some people can't take a joke. Apparently, John, you can't make a joke (at least on this forum). I got your linear/geometric progression quip. People take you too seriously sometimes... <br /> </td></tr></table><br />Okay, so I don't have psychic abilities.<br /><br />Does the fact that your only response is to deflect my criticism with humor indicate that you concede the points I made?

[old] Porkchop · Post by **[old] Porkchop** » August 30th, 2005, 11:31 am

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 29 2005, 08:52 PM)</b></div></td></tr><tr><td class='quote'>Interesting figures.<br /><br />How did you get all 1433 times?<br /><br />Age 44.5 hwt has a PATT time of 5:56.8.<br /><br />Thus the mean time of 7:34 has a PATT percentage of 78.6.<br /><br />Probably including all rowers, not just those who entered their times in the rankings, would result in a lower PATT percentage than this. As it has with running, the curve will likely keep expanding. <br /> </td></tr></table><br />This is just wishful thinking on your part, John. The reason that the PATT percentage of the mean time is not 50% is the very flaw in the PATT calculation that I pointed out above. You seem to think that the world record time should dictate the distribution of performance among the entire universe of rowers (or perhaps all potential rowers). In order to attain a PATT percentage of 50, a rower would have to have a time of 11:53.6. You apparently think that if more people in that AWG would report their times, we would find that 50% of them row a 2K in 11:53.6 or slower and that 25% of them row a 2k at 23:47.2 or slower. Your theory does not reflect reality.

[old] John Rupp · Post by **[old] John Rupp** » August 30th, 2005, 2:15 pm

PC,<br /><br />Reread my post about density. <br /><br />I already have told you that PATT, being directly related to WR quality times AWG, has nothing to do with percentiles. <br /><br />What about that don't you understand? <br />

[old] John Rupp · Post by **[old] John Rupp** » August 30th, 2005, 2:16 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(Druss @ Aug 30 2005, 12:58 AM)</b></div></td></tr><tr><td class='quote'>John go to <a href='http://www.openoffice.org/' target='_blank'>Open Office</a> and download the free suite that includes a programme compatible with Excel.<br /><br />George <br /> </td></tr></table><br />George,<br /><br />I will do that.<br /><br />Thanks very much!

[old] Porkchop · Post by **[old] Porkchop** » August 30th, 2005, 4:11 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 30 2005, 01:15 PM)</b></div></td></tr><tr><td class='quote'>PC,<br /><br />Reread my post about density. <br /><br />I already have told you that PATT, being directly related to WR quality times AWG, has nothing to do with percentiles. <br /><br />What about that don't you understand? <br /> </td></tr></table><br /><br />That seems rather inconsistent with this previous post:<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 29 2005, 03:40 PM)</b></div></td></tr><tr><td class='quote'><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(bmoore @ Aug 26 2005, 11:55 AM)</b></div></td></tr><tr><td class='quote'>The 90th percentile time is 6:44.8, which gives you a 87.0 PATT. <br /><br />The 75th percentile time is 6:59.2, which gives you a 84.0 PATT. <br /><br />The 50th percentile time is 7:25.7, which gives you a 79.0 PATT. PATT is now higher by 29%. [right] </td></tr></table><br />I got the 84 and 87% from this, it appearing that in this case anyway a greater number of rowers in the rankings are between 84 and 87 PATT percentages....<br /><br />which indicates the middle point on a bell curve, i.e. 50% ranking percentile would be an 84 to 87 PATT percentage.<br /><br /><i>However, I really think that is nowhere the case and, again, that PATT and the percentiles in the rankings are not related, <u>at least <b>not yet</b></u>. <u>This is the case as there is <b>not yet enough participation </b>and <b>not enough of a range of performances in the rankings</b> to get a <b>true percentile</b> of abilities and performances</u>.</i><br /><br /><i>Once there is enough of a range and much greater participation in the rankings, I think we will see that a 50% on the bell curve will be considerably less than an 84 to 87 PATT percentage.</i><br />[Emphasis added.] <br /> </td></tr></table><br />You argue there that the incongruity between PATT percentages and percentile rankings will decrease as the sample size increases. The point you seem to be missing is that there will never be a relationship between them under any circumstances, because your formula is an unsound analytical tool both mathematically and logically. Now, go back and read your own post on density. <br /><br />

[old] John Rupp · Post by **[old] John Rupp** » August 30th, 2005, 4:39 pm

PATT, being directly related to WR quality times AWG, has nothing to do with percentiles. <br /><br />1) Given a certain 100% PATT, and a certain time to compare, the PATT percentage will always be the same. <br /><br />A percentile can and does change quite often, depending on who enters their times in the rankings. <br /><br />2) If only those with PATT percentages of 87 entered their times in the rankings, then all of the percentiles would be at an 87 PATT percentage. <br /><br />Thus, the ranking percentiles are highly DEPENDENT on who enters their times in the rankings and, even more, are at the whim of those who enter their times.<br /><br />3) None of that affects PATT whatsoever. Given the same 2 scores, the PATT percentage will always be the same.<br /><br />Given a large enough and/or representative enough sample size, the percentiles for a certain AWG will begin to stabilize, but will never do this entirely.<br /><br />4) Thus if you have a highly representative sample or population of rowers in a certain AWG, then it becomes possible to see that a certain percentile would be close to a certain PATT percentage.<br /><br />For example, I think that a true representative sample and/or large enough representative population, would show those at the 50th percentile would have less than a 78 percent PATT.<br /><br />This does not illustrate any relationship and/or cause and effect between one or the other as, again, there is no relationship in between them.<br /><br />5) PATT is and will always be an exact calculation. <br /><br />The ranking percentiles are not exact, and will never be exact.

[old] slo_boat · Post by **[old] slo_boat** » August 30th, 2005, 5:05 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 30 2005, 03:39 PM)</b></div></td></tr><tr><td class='quote'>The ranking percentiles are not exact, and will never be exact. <br /> </td></tr></table><br /><br />Actually, the percentiles are exact at the time they are calculated. However, they are subject to change.

[old] slo_boat · Post by **[old] slo_boat** » August 30th, 2005, 5:15 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 30 2005, 03:39 PM)</b></div></td></tr><tr><td class='quote'>Given a large enough and/or representative enough sample size, the percentiles for a certain AWG will begin to stabilize, but will never do this entirely.[right] <br /> </td></tr></table><br /><br />The only problem with this is that you are working with the <i>population</i> of people who have entered their times, not a <i>sample of the population </i>of the people who have entered their times.<br /><br />You don't have any other options though, because the only population you can use is the population of people who use an erg and who have submitted times to the rankings. If you want to do a study of the population of people who have used an erg, you need to do a better study.<br /><br />If you understood statistics and sampling, then you would understand that the sample size does not have to be very large at all. If you are curious, there are ways to handle the long tail distribution. That is going to take something more than <i>Statistics for Dummies</i> though. You might want to brush up on integrals. You could use the brute force method and take repeated samples and then use the mean of the samples. That way you could assume a typical bell shaped distribution.

[old] slo_boat · Post by **[old] slo_boat** » August 30th, 2005, 5:18 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 30 2005, 03:39 PM)</b></div></td></tr><tr><td class='quote'>. . . would show those at the 50th percentile would have less than a 78 percent PATT.[right] <br /> </td></tr></table><br /><br />This is brilliant. You have defined a long tail distribution!

[old] bmoore · Post by **[old] bmoore** » August 30th, 2005, 6:13 pm

John,<br /><br />I think you've mangled the statistics a bit. The median time will always be the exact middle time in the population of posted times. This is the 50th percentile. It's different than the mean, or average time, because there is quite a smattering of times beyond 9 minutes that seem to trail off forever (well at least until about 12 minutes, with one very big exception). Here's the 50th percentile for the last 6 full years with the PATT%:<br /><br />2000 - 7:18.0 (80%)<br />2001 - 7:12.7 (81%)<br />2002 - 7:18.2 (80%)<br />2003 - 7:20.3 (80%)<br />2004 - 7:20.0 (80%)<br />2005 - 7:25.7 (79%)<br /><br />Although the times change, there appears to be a strong correlation between the 50th percentile and 80% PATT.<br /><br />Also, a sample of 1,433 times should be large enough to statistically represent the entire population, where any variances would be statistically insignificant. There is the potential problem of self-selection bias, but I don't think this creates a significant variance while you think it destroys the credibility of the percentiles. That's quite a significance variance in our thinking. I think it's probably not a big deal, and you think the sky is falling. I wish you were handling the statistical terms accurately to be able to prove something, but I'm afraid you're dismissing the opposing argument by mis-interpreting it.<br /><br />I think I've faithfully handled the PATT system's calculations and logic, but it's only simple algebra and not statistics. It's in the interpretation of the usefulness of the data that we have a difference of opinion on. Again, I find PATT system interesting and even used it as a gauge for pacing on my recent half marathon and 6k attempts. But you haven't convinced me that it has usefulness beyond a rookie training season. Once a rower has established enough fitness, say after 6 months of good quality training, and has established a rookie season set of personal bests, then the incremental changes in the PATT system don't seem to be of any use. The changes are simply too small. The absolute change in ranking, time, and even percentiles would seem to be more meaningful to use for training.<br /><br />You're doing a great job in explaining the exactness of the PATT system, but I'm questioning it's broad application across the ranking divisions and it's usefulness for an established rower's training. (Quite honestly, I realized that all my statistical modeling did was explain a simple distribution, but it's how I use the distribution that really matters. For example, compare two companies that earn 5% net income. One has revenues of $3 billion, the other has $3 million. One is the #1 company in the industry, the other is a small market niche provider. But comparing both on the 5% number doesn't really do much for comparing the companies.)<br /><br />SIDEBAR: I couldn't even find you in the rankings! Talk about self-selection bias. You're an example of a fast rower that doesn't enter times. This tilts the scale toward the rankings representing the population.<br /><br />Anyway, John, I'm still waiting for compelling evidence to use the PATT across the ranking divisions and how it can be used for an established rower's training.

[old] bmoore · Post by **[old] bmoore** » August 30th, 2005, 6:16 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(slo_boat @ Aug 30 2005, 05:15 PM)</b></div></td></tr><tr><td class='quote'>That is going to take something more than <i>Statistics for Dummies</i> though. <br /> </td></tr></table><br /><br />Hey, that book got me through my MBA program's statistics class! (Or was it Forgotten Statistics?) Either way, it applies.<br /><br />The best "Dummies" book I was was "Homeschooling for Dummies". Scary, very scary.

[old] John Rupp · Post by **[old] John Rupp** » August 30th, 2005, 6:56 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(bmoore @ Aug 30 2005, 03:13 PM)</b></div></td></tr><tr><td class='quote'>The median time will always be the exact middle time in the population of posted times. </td></tr></table><br />Let's say you have 1400 rowers. The 700th rower does 7:17.9 and the 701st rower does 7:18.1. These two are your middle two rowers and there is no one in between them. What is your median time?<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>there appears to be a strong correlation between the 50th percentile and 80% PATT. </td></tr></table><br />Interesting!<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>a sample of 1,433 times should be large enough to statistically represent the entire population, where any variances would be statistically insignificant. </td></tr></table><br />Yes if the sample was representative, but it isn't. There is a certain arrogance in the sport and also that permeates the rankings -- note PorkChop's comments about those slower than him -- that precludes all rowers feeling comfortable by putting their times in the rankings. Also the fact that the percentile method, vs a percentage of WR, is used. No one wants to be less than the 50th percentile, or even below the 80th for that matter. In my opinion the percentiles are not encouraging and are not a good method to use.

[old] John Rupp · Post by **[old] John Rupp** » August 30th, 2005, 7:06 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>There is the potential problem of self-selection bias, but I don't think this creates a significant variance while you think it destroys the credibility of the percentiles. </td></tr></table><br />Yes.<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>That's quite a significance variance in our thinking. I think it's probably not a big deal, and you think the sky is falling. </td></tr></table><br />Not the sky falling -- just not as relevant and inclusive as it could be.<br /><br />What do you feel is being mishandled or interpreted?<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>I think I've faithfully handled the PATT system's calculations and logic, but it's only simple algebra and not statistics. </td></tr></table><br />Right.<br /><br />I have said a number of times that they are not related and are not the same thing.<br /><br />I have found the ranking percentiles to be totally not relevant, nor useful and have never used them. The rankings are quite useful, however, seeing what others have been doing, taking aim, moving up (or down) and/or back and forth competitions, motivation for improvements etc.

[old] John Rupp · Post by **[old] John Rupp** » August 30th, 2005, 7:12 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>SIDEBAR: I couldn't even find you in the rankings! Talk about self-selection bias. You're an example of a fast rower that doesn't enter times. This tilts the scale toward the rankings representing the population. </td></tr></table><br /><br />haha.... well I am there... did you look in 50-59 lightweights?<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>Anyway, John, I'm still waiting for compelling evidence to use the PATT across the ranking divisions and how it can be used for an established rower's training. <br /> </td></tr></table><br />You did just give me an idea.<br /><br />Based on your investigations, the 50th percentile came out to approximately an 80 percent PATT. This could be extrapolated to the other percentiles, based on the world records, i.e. PATT, vs the fastest time entered, which might only be an 80% PATT itself.<br /><br />In this sense, even though they are not related in an absolute computational way, one could be used to facilitate the other in the rankings.<br /><br />However, I am still against the philosophy of percentiles.

[old] Porkchop · Post by **[old] Porkchop** » August 30th, 2005, 9:17 pm

<table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(John Rupp @ Aug 30 2005, 05:56 PM)</b></div></td></tr><tr><td class='quote'><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE(bmoore @ Aug 30 2005, 03:13 PM)</b></div></td></tr><tr><td class='quote'>The median time will always be the exact middle time in the population of posted times. </td></tr></table><br />Let's say you have 1400 rowers. The 700th rower does 7:17.9 and the 701st rower does 7:18.1. These two are your middle two rowers and there is no one in between them. What is your median time?<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>there appears to be a strong correlation between the 50th percentile and 80% PATT. </td></tr></table><br />Interesting!<br /><br /><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><div class='genmed'><b>QUOTE</b></div></td></tr><tr><td class='quote'>a sample of 1,433 times should be large enough to statistically represent the entire population, where any variances would be statistically insignificant. </td></tr></table><br />Yes if the sample was representative, but it isn't. There is a certain arrogance in the sport and also that permeates the rankings -- note PorkChop's comments about those slower than him -- that precludes all rowers feeling comfortable by putting their times in the rankings. Also the fact that the percentile method, vs a percentage of WR, is used. No one wants to be less than the 50th percentile, or even below the 80th for that matter. In my opinion the percentiles are not encouraging and are not a good method to use. <br /> </td></tr></table><br />I don't think you really read or understood my remarks, John, but that's the primary problem in any discussion with you.

Concept2 Forum

How To Calculate Your Patt Percentages

Training

Training

Training

Training

Training

Training

Training

Training

Training

Training

Training

Training

Training

Training

Training