From

Location

Message

Date Posted

Netherlands

Some time ago I mentioned it is impossible to have a linear relationship between size and weight, making the 2013 OTT chart incorrect. A pumpkin taping 1200 lbs that grows one inch will put on about 8 lbs. A pumpkin taping 2000 lbs that grows one inch will put on much more weight, though according to the new OTT chart it will also be 8 lbs. Science says this is impossible. Several people tried to explain why it was correct, but it remains impossible.

To see if I made the wrong assumption or not I crunched the data from all 1200 lb to 2100 lb pumpkins using both the linear 2013 OTT chart and the more logarithmic chart supplied by Team Pumpkin.

If I am correct, then the new 2013 OTT chart should make pumpkins over 1200lbs go heavier and heavier while the Team Pumpkin chart should remain in balance with as many pumpkins going light as heavy. The charts meet at about 1200 lbs, so the 2013 OTT chart should be correct at 1200 lbs.

Here are the results.

10/13/2014 5:02:03 AM

Netherlands

2013 OTT chart

1900-2100 lbs – 92% heavy / 8% light
1800-1899 lbs – 80% heavy / 20% light
1700-1799 lbs – 79% heavy / 19% light
1600-1699 lbs – 74% heavy / 26% light
1500-1599 lbs – 57% heavy / 43% light
1400-1499 lbs – 76% heavy / 24% light
1300-1399 lbs – 63% heavy / 37% light
1200-1299 lbs – 51% heavy / 49% light

10/13/2014 5:02:14 AM

Netherlands

Here a few comparisons at the top end.

Team Pumpkin chart

1900-2100 lbs – 50% heavy / 50% light
1800-1899 lbs – 40% heavy / 60% light
1700-1799 lbs – 45% heavy / 55% light
1600-1699 lbs – 53% heavy / 47% light

10/13/2014 5:02:28 AM

Netherlands

If you have a pumpkin over 1300 lbs and you want to feel good at the weigh-off, go for the 2013 chart. For a more accurate estimate, I'd go for the Team Pumpkin chart.

10/13/2014 5:06:54 AM

Willamette Valley, Oregon

The short answer is that sorting them by weight isn't an appropriate way to assess chart accuracy. See this old post for more info:

http://www.bigpumpkins.com/msgboard/ViewThread.asp?b=3&p=514667

Skip down to 8/21/2014 4:23:58pm
Also be sure to check out the graphs I posted links to.

As of the GPC data I just downloaded for 2014:

Top 10 by OTT (459"+): 2013 +3.2%, TP -6.5%
Top 50 by OTT (435"+): 2013 +2.2%, TP -4.6%
Top 100 by OTT (419"+): 2013 +1.5%, TP -3.7%

Below about 410" to 420" the two charts are reasonably close in the estimates (within 1-2% usually).
Under 410": 2013 -0.5%, TP +1.0%

10/13/2014 6:17:53 PM

Willamette Valley, Oregon

A telling point from your last post, Brad: at weigh off time you don't know you have a 1300+lb pumpkin (unless you pre-weighed). You only know you have a, say, 394" pumpkin. That's why the chart is best assessed by OTT, the thing you know before it's weighed.

From this year's data:
1290-1310 lb pumpkins range in OTT from 369" to 404"

10/13/2014 6:29:40 PM

Willamette Valley, Oregon

As far as scientifically impossible...it's only impossible if you assume a sphere or another fixed shape with a volume proportional to the radius cubed (or another fixed power function).

If the assumed shape is allowed to vary somewhat with OTT measurement (especially over 410-420") a non-cubic (non-power function) relationship emerges empirically. This fits the data much better than old style charts including the team-pumpkin chart.

I'm not sure what this other shape is, just that it isn't the same as smaller pumpkins.

So even though it may seem strange, the 2013 chart fits past data substantially better than other charts at high OTT's. And seems to predict weights of high OTT's better as well.

The trick is to assess it properly...by OTT not weight.

Again, see that old post for more info.

10/13/2014 6:45:34 PM

Netherlands

"I'm not sure what this other shape is, just that it isn't the same as smaller pumpkins."

Most likely it is the bulging at the bottom as the weight increases.

10/14/2014 5:56:43 AM

Netherlands

I understand your way of assessing the chart, but choose to approach it from the other way, being, how accurate is it as the weight of the pumpkin increases? At the end of the day growers are going to choose their seed based on weight and % heavy, not OTT.
The top 25 this year averaged 6.2% heavy on your chart and you would expect this to be much closer to 0% like the TP chart is.

10/14/2014 6:22:03 AM

Netherlands

Shape has less to do with it than you might think. If you add one inch to a little ball or one inch to a huge square, the added weight will not be the same.

There will never be a perfect chart as there are so many other factors involved (especially flesh thickness). Perhaps it is good to have both charts side by side and average them.

10/14/2014 6:24:47 AM

East Jordan, MI

I'm no scientist, but the statement that "it is impossible to have a linear relationship between size and weight" made me think.

I think the reason that the statement is false is because you are forgetting that you aren't measuring something that is solid. You are measuring something that is hollow. And the cavity increases just as fast as the shell (maybe faster in the larger pumpkins). So the weight of an inch in OTT WILL be very similar to smaller pumpkins (with less cavity but same wall thickness). The percent heavy comes from getting that shell to be thicker and more dense (less cavity). That is why it is MUCH easier to get a 1073 lb est to be 1312 actual lbs (22.227%) than it would be to get a 1684 lb est to be 2066 actual lbs (22.227%) -- you are measuring less air in the smaller pumpkins.

At the end of the day the chart is an estimate (with a plus or minus 5% error rate). And if everyone uses the same chart it doesn't really matter because everyone is compared against the same constant. In other words, a 2008 that is 462 ott compared to a 2058 that is 462 ott the percentage number doesn't really matter, it is easy to see which is heavier to any chart.

At least that is my take.

10/14/2014 6:43:56 AM

Netherlands

Thanks for your reply. Of course you assume the wall thickness in small pumpkins is the same as big pumpkins. Is this true? In order for there to be a linear relationship, the wall thickness would have to decrease as the pumpkin gets larger. Does this happen?

You write: "That is why it is MUCH easier to get a 1073 lb est to be 1312 actual lbs (22.227%) than it would be to get a 1684 lb est to be 2066 actual lbs (22.227%) -- you are measuring less air in the smaller pumpkins."

The point is that the 2013 chart DOES have 1750 lb+ going 20% heavy. The TP chart doesn't.

10/14/2014 8:16:28 AM

Sw Wi

Im my humble opomion, i don't care what chart is used as long as the chart doesn't change every year.

10/14/2014 1:32:58 PM

Willamette Valley, Oregon

My point is that you shouldn't expect a properly functioning chart to perform well on a weight based cut-off. The weight based cut off should make a chart look heavy for reasons I've discussed before (i.e. it's a biased sample).

The reason the TP chart comes out looking ok in the weight based cut-off approach is counteracting errors:

The TP chart makes large OTT pumpkins appear light by about ~4-6% on average (depending on the exact OTT) - That's what the data show.
Just by luck the incorrect weigh-based cut-off approach makes charts look ~4-6% heavier than they would be using a proper OTT approach. Add those two errors together and they cancel, but that doesn't make the weight cut-off method or the chart right.

A broken clock it right twice a day. That's all that's happening here. (I should say that it's the weight based approach that broken (i.e. wrong). To be fair the TP chart is just off at high OTTs, I wouldn't say it's "broken" altogether.)

The old thread explains a lot of this in more detail, too, with data and graphs etc.

It's a tough concept, because prior to the 2013 chart, the weight-based approach has appeared to work, but it's because all past charts have had the bias toward below-chart pumpkins at high OTT, in the same way the TP chart does.

10/14/2014 4:10:55 PM

Germany

Ok, different charts can make a pumpkin appear heavy and light, but even with the same chart a pumpkin can be heavy or light.

http://www.bigpumpkins.com/Diary/DiaryViewOne.asp?eid=230682

Shape and, that´s even more important, the relative orientation of this shape with respect to the ground matters.

10/14/2014 4:20:41 PM

Everywhere there are growers

For anyone who wants an easy to follow explanation for why the Team-Pumpkin charts ‘work’ I suggest you take a few minutes to read the article in the ‘How To…’ section here on Bigpumpkins.com

10/14/2014 8:51:14 PM

Netherlands

http://www.bigpumpkins.com/Diary/DiaryViewOne.asp?eid=230804

Thanks for posting this Bubba. Certainly proves my point, and corresponds with the TP chart.

10/17/2014 4:27:51 AM

Netherlands

Jörg (Pumpking), thanks for your input. I tried that too some time ago to prove the entire OTT system isn't correct. That is why I didn't use the OTT system in my marrow chart.

With huge pumpkins now going 20% heavy, the charts aren't really saying much, which makes a weigh-off more exciting. Time to introduce echoscopy and 3D-scans to determine the thickness of the walls...lol

10/17/2014 4:37:42 AM

Germany

...and here´s another example, even people who don´t have a pumpkin can easily do this experiment:

A CD record serves as your pumpkin model. Now you can "grow" it as an upright wheel, or you can grow it as a flat millstone (both can happen in the world of pumpkins, you know).
Now you determine the OTT of your CD record. You don´t need to measure, you simply use the standard values of
120 mm diameter and 1.2 mm thick.

For an upright wheel you will obtain the following OTT data:
CC 2 x diameter + 2 x thickness is 242.4 mm
EE 2 x diameter + 1 x thickness is 241.2 mm
SS 2 x half diameter + 1 x half circumference is 308.5 mmm

OTT 792.1 mm

For a flat millstone you obtain the following OTT data:

CC circumference of the disc is 377 mm
EE 1 x diameter + 2 x thickness is 122.4 mm
SS (same as EE) is 122.4 mm

OTT 621.8 mm

10/17/2014 5:21:00 AM

Germany

In the world of true pumpkins you could obtain these OTT values in cm (instead of mm), the upright wheel with 792 cm OTT and the flat millstone with 622 cm OTT could basically be the same pumpkins, with the same volume and the same wall thickness, they simply have a different orientation on the patch.

The flat to the ground ones produces smaller OTT values, hence they will be guesstimated at lower weight and will thus be more likely to surprise the grower with heavier actual weight, whereas the tall ones (like the upright wheels) most likely behave in an opposite manner.

Therefore, a more reliable OTT' chart (which more precisely accounts for wall thickness, i.e., the heavy ones usually are the ones with thick walls, the light ones are the thin-walled fruit) needs to account for the influence of fruit shape. In case of very symmetric fruit (ellipsoid-like fruit, regular discs etc.) this could perhaps be done by individual weighting parameters for CC, EE, SS (then the OTT' value is just a number obtained from a more complex equation, but it isn´t the sum of CC, EE and SS any longer). Such an equation should, for example in case of the CD record, produce the same OTT' value no matter whether you have the CD as the "upright wheel pumpkin" or the "flat millstone pumpkin".

This could be a step into the right direction, although there will still be other problems/difficulties, such as deep ribs, highly irregular shapes, "birdbath pumpkins" where the tape runs across nothing for a long distance (when measuring the EE) etc. etc., and therefore I´m pretty sure there will be new versions of OTT charts every couple years.

10/17/2014 5:21:06 AM

Willamette Valley, Oregon

I think the confusion here could be in accuracy versus precision:

http://climatica.org.uk/wp-content/uploads/2012/02/Accuracy-vs-precision1.jpg

The question of chart accuracy (the tendency for the chart to hit the center or average of the weights for all pumpkins at a given OTT) and chart precision (the spread around the average) are two very different questions.

Both the GPC and Team-Pumpkin chart have similar overall precision as measured by the standard deviation: right around 8.5 to 9.0 percentage points depending on the year going back a few years. It's a toss up as to which chart is better in terms of overall precision. A tie really. That's because the precision is influenced by the variation not explained by the charts and that's linked closely to the OTT measurement method itself, which both charts use.

Brad could be right that a system other than OTT might work better to improve our precision. I'm not sure.

So given the similar precision in both models, a 428" OTT / +20% heavy pumpkin on the GPC chart (which gives heavier pumpkins in that range) would come in at +15% on the Team-Pumpkin chart (which gives lighter pumpkins on average in that range). Likewise a -20% on the TP chart might be -15% on the GPC chart. Those chart differences simply reflect the underlying bias in each chart and not the overall precision. As a chart accuracy assessments, those individual pumpkin numbers have little meaning given the known biases in accuracy. You may not see the -20% when you use the inappropriate weight cut-off and that's a good reason not to use the weight cut-off approach to assess model fits.

10/17/2014 7:18:02 PM

Willamette Valley, Oregon

I think pumpking's examples of OTT error are fun and interesting, but the model uses OTTs measured on pumpkins laying on their flat side (I'd guess 95+% of the time.) so that's probably not a major issue.

I'd strongly suspect even a full displacement volume of a pumpkin would still have a fair amount of variation vs its weight. Wall thickness and density, internal air volume, other variations related to shape, and who knows what else will still cause error.

10/17/2014 7:25:13 PM

Willamette Valley, Oregon

And one more for today, ha ha....

Remember that the OTT to weight conversion charts are really only known to be accurate at the end of the season at the scales. That's where the data we fit the charts to comes from. We generally don't know precisely what's happening with weight mid-season vs OTT. We can guess, but without a lot more mid-season weights we can't be sure.

It's not surprising that some pumpkins that have been weighed early put on more lbs per inch than the chart might indicate. The chart isn't made to assess that, as I just indicated, so any deviation from the chart weights per inch through the season don't really prove or disprove the appropriateness of the chart to estimate end of season OTT vs weight relationships.

Some early speculation...It seemed to me that my tally of the average %chart got heavier as more data was posted through the season. I intend to look into that more, for fun (LOL), once all the data are in. If that preliminary assessment is true then, on average, pumpkins might put on more weight per inch OTT near the end. This could indicate a %chart vs days since pollination relationship perhaps. Not sure if that's true yet, but that could be yet another source of variation. (Would this vary by genetics as well?)

10/17/2014 7:42:45 PM

Willamette Valley, Oregon

Ok last one... When I start getting to serious about this stuff, I have to stop and laugh/be happy that our lives are so good that we can argue about the ridiculousness that is the relationship between the size of a pumpkin and its weight. LOL. No wonder my wife thinks I'm nuts. LOL!

10/17/2014 7:46:02 PM

Willamette Valley, Oregon

A math-minded friend of mine sent me this article, unsolicited, this afternoon - good timing!:

"The growth of giant pumpkins: How extreme weight influences shape"

http://www-old.me.gatech.edu/hu/Publications/Hu11.pdf

Some familiar names are cited in there. And, Brad, your photo is in there with your huge squash.

I haven't read it all yet, but thought it would be interesting to share.

10/18/2014 1:03:22 AM

Netherlands

Thanks for your input again Scott. The major concern I have is the constant 8lb gain, regardless of size. Many, many moons ago there were a few growers that grew the entire year on scales and I have grown marrows and field pumpkins on scales for years (in order to make the marrow chart). In all cases, no exceptions, the larger the fruit became, the more lbs it grew per inch.

Thanks for the link to Hu's article. I somehow always seem to get involved in studies, publications and projects regarding all things giant, but never actually read Hu's article, though I was aware of the contents. Somehow I managed to get into the newspaper with him a few years ago.

PS. I think many of our wives think we are nuts. I am, so I'm not too worried about that.

10/18/2014 3:10:43 AM