Message Board - BigPumpkins.com

SF Bay Area, California

Using basic statistical analysis tools on observed pumpkin weights from a particular seed line, it is possible to generate probability distributions that will estimate the probability of popping a big one from that seed line.

For the purposes of this exercise, I am assuming that pumpkin weights from a seed line approximately follow a normal distribution and that the major factors that could bias the results tend to cancel themselves out.

The methodology is pretty straight forward. Simply calculate the average and standard deviation of pumpkin weights for a seed line and then use normal distribution tables to calculate the probabilities of certain outcomes.

Those of you who are mathematically challenged can use my Improbable Pumpkin Probability Calculator to determine pumpkin weight probabilities at the following link:

http://home.pacbell.net/diana_do/probability.htm

For this to work with reasonable confidence about 20 known weights from a seed line are needed....

1/9/2006 1:33:28 PM

SF Bay Area, California

I analyzed three proven seeds, the 846 Calai, the 723 Bobier and the 1068 Wallace in this manner with the following results:

846 Calai
Average Weight = 801.5 lbs (YES - believe it or not)
St. Deviation = 199 lbs
Probability of popping 1,000+ = 16%
Maximum Predicted Weight = 1,398 lbs.

723 Bobier
Average Weight = 885 lbs
St. Deviation = 233 lbs
Probability of popping 1,000+ = 30%
Maximum Predicted Weight = 1,552.5 lbs.

1068 Wallace
Average Weight = 1,058 lbs
St. Deviation = 212 lbs
Probability of popping 1,000+ = 60.5% (18% for Russ)
Maximum Predicted Weight = 1,695 lbs.

vince

1/9/2006 1:34:26 PM

Andy W

Western NY

i used to be pretty good at stats, but here's my take on the numbers- the max. prediction is about 100 pounds too high for each. i think the only way that scenario works perfectly is if there's a single-seed silver bullet theory, being planted by one of the top 10 growers under perfect weather. just out of curiousity, n=? for each seed?

pretty interesting numbers, for sure.

1/9/2006 1:51:03 PM

BrianInOregon

Eugene, OR

I'm a math fan so I'm always interested when actual numbers come up. Andy makes a good point. I take it these numbers were based on the AGGC pumpkin data? If so, there's no question the average weight and probability predictions are also skewed as the AGGC only contains pumpkins over 500 pounds. I wonder how many 846's, 723's, and 1068's were planted with poor/no results to report.....not to mention the fact that most of the 846's and 723's grown in the past couple of years ended up in the hands of a select few "heavy hitters" who have the connections to get seeds like that and the soil/skills to grow them to their potential. The AGGC is a great database but basing something like a probability calculation on data that isn't complete doesn't tell the whole story.

It's interesting that the 1068, a relatively new "hot seed," has the highest probability of producing a 1000+ pumpkin. There's no question the 1068 is a good seed but a 60% chance of growing a 1000 pounder? Put a good seed in the right hands and you'll get those kind of results. As more people grow the seed, there's no question that number will go down quite a bit but hey....I'd still try one LOL.

1/9/2006 2:20:04 PM

SF Bay Area, California

At first glance, the maximums seemed a little high to me also but on some reflection I think they are not totally out of hand - just not very likely. A lot of planets would have to line up to achieve them. The best seed in the seed line, a top grower, perfect weather, great soil, no pests/pathogens, ect...

Alternatively, the fact that many smaller pumpkins drop out of the stats entirely may be biasing the numbers up somewhat. I'm counting on higher weights in the future due to the grower learning curve to pretty much cancel that bias out.

The number of samples for each pumpkin were as follows:

846 Calai; n=118
723 Bobier; n=111
1068 Wallace; n=19

1/9/2006 2:23:30 PM

Sioux Falls, SD garyboer@dakotalink.com

does this assume top growers and good conditions?
. .or average conditions of 20 entered weights?

1/9/2006 2:26:52 PM

Sioux Falls, SD garyboer@dakotalink.com

a 1695 after every 1068 has been grown seems possible

1/9/2006 2:28:19 PM

mshy

Nekoosa, Wisconsin

Miss Doucet (LOL). Are you including splits and unofficial weights in your calculations. If so, are you standardizing those weights so they would reflect the pumpkins final weight. For Example a pumpkin that splits and weighs 992# at day 46 is not the same as a pumpkin that weighs 992# at day 120.

Mark

1/9/2006 2:32:18 PM

Sioux Falls, SD garyboer@dakotalink.com

. . .at any rate. . .
if you have a below average record as a grower,
odds are you'll be on the left side of the bell.

. . but I think you're right in that the lowest end of
results will not be fully represented in your sample.

This will be less a factor in top-notch seeds. . .they are
not just given to everybody.

1/9/2006 2:34:14 PM

SF Bay Area, California

In my calculations above, I used the AGGC database for the sample points. I included every known fruit and made no corrections. I don't think corrections should be made for light weights due to splits or premature rotting that are present in the sample because that is a legitimate expected outcome that will be reflected in the whole population.

However, it is bothering me that most of the tail of the sample population under 500 pounds is removed from the data.
I do think this needs to be corrected to make the predictions more accurate. Does anybody have a sense of roughly what percentage of plants fail to push their fruit over 500 pounds or are culled?? In my experience this number is about 30% or so....

1/9/2006 3:26:32 PM

Eugene,OR

So this means that I have to grow 5.2 1068 Wallace plants this year to get one to break 1000 pounds??? I hope that there are four more being auctioned off in the next few weeks! Are you including the 947 Legoretta (which was the 1068 Wallace grown in my patch this year) as one of my data points?

1/9/2006 3:50:08 PM

Sioux Falls, SD garyboer@dakotalink.com

not much hope for that .2 of a plant. . .

if you have 60% chance with one,
then 84% with 2 and 93.6% with 3

1/9/2006 4:40:38 PM

SF Bay Area, California

Yes - one of my data points for the 1068 Wallace was the 947 Legoretta. This fruit lowered the overall average for the Wallace seed and also reduced the standard deviation somewhat lowering the expections for the entire seed line!

If you grow 5 1068 Wallace plants this year and only get 1 1,000 pounder, you should be shot and have your lifeless body drug through Half Moon Bay during the pumpkin festival...

By the way, I calculated your odds of 18% with the Improbable Pumpkin Probability Calculator by adding a handicap of 250 pounds to the starting weight. If you enter the Wallace seed's mean and std deviation into the caculator and use a starting weight of 1250 pounds instead of 1000, you get your grower-skill adjusted probability of growing a 1,000+. pounder.

http://home.pacbell.net/diana_do/probability.htm

1/9/2006 4:59:47 PM

CliffWarren

Pocatello (cliffwarren@yahoo.com)

The results must be skewed to the high side because none of
my results are on AGGC. No, seriously, interesting work Vince.

I would think that "grower ineptitude" and "poor climate" are
not linear functions.... not linear with the actual seed
selected... (not asking you to solve that, just pointing it
out). That is, if I grow 500 with my own seed, that doesn't
mean that I'm due for 1100 with the 1068.

And finally, a way for me to get in the parade at Half Moon Bay!!!
If you drag it behind a nice Mustang convertible, I
might agree to the idea.

1/9/2006 5:27:15 PM

gordon

Utah

I'd like to see the numbers on the 1068 in a year from now... after more average growers plant it.

anyone think the odds of growing a 1,000 lber will have gone up? I don't

I was going to say average Joe's ... but the Joe's that grow seem to be above average. :)

1/9/2006 5:31:49 PM

the gr8 pumpkin

Norton, MA

My 720.6 (from a 1068) isn't in the AGGC, I think this happens alot. AleX Noel.

1/9/2006 6:09:55 PM

SF Bay Area, California

aleX - Just for fun I added your 720.6 to the 1068 data set to see how it impacted the predictions. It lowered the average of course, but it also increased the standard deviation somewhat. The effect on the predicted probabilities are shown below:

1068 Wallace
Average Weight = 1,058 lbs - Updated: 1,041
St. Deviation = 212 lbs - Updated: 219.5
Probability of popping 1,000+ = 60.5% Updated: 60.1%
Maximum Predicted Weight = 1,695 lbs. Updated: 1,700

You can see that missing a few data points doesn't really impact the predicted probabilities that much even with the relatively small sample size in the 1068 data.

1/9/2006 7:11:36 PM

Tom B

Indiana

I had one....it was well below the curve...LOL

1/9/2006 9:42:21 PM

Eugene,OR

Tom, you should be shot too!

1/9/2006 10:39:58 PM

Tremor

Ctpumpkin@optonline.net

I think we're forgetting that Ronny & Dick "persuade" the lower end bell curve growers to not talk. It's a Rhode Island thing.

1/10/2006 10:26:32 AM

CliffWarren

Pocatello (cliffwarren@yahoo.com)

The Pumpkin Mafia???

1/10/2006 10:46:40 AM

Tremor

Ctpumpkin@optonline.net

Clarification post: Ron & Dick don't "persuade" growers to not report less than ideal results. Since RI politicians are the most corrupt in the US, it was a political joke taken out of context. RI *growers* are as solid as they come.

A point always worth considering but impossible to track is the under-500 crowd that just can't be tracked.

Another ace up the sleeve of the 1068 is the close-knit support mechanism that exists in the SNEPGA. Grower support has really influenced the progeny of the 1068 to the undeniable credit of the high class growers who have pushed it up where it is.

1/10/2006 10:52:37 AM

Eugene,OR

Before growing the 1068, how many of the SNEPGA guys had grown one over 1000? 1100?

1/10/2006 11:11:40 AM

Eugene,OR

That is, how many of the SNEPGA guys who grew the 1068 last year.

1/10/2006 11:12:38 AM

CliffWarren

Pocatello (cliffwarren@yahoo.com)

LOL, I hope you all know we're all totally joking about the Pumpkin Mafia... yes, I've heard that about RI. What do they call it, "Rouge's Island?"

At any rate, perhaps the under-500 ones should not be counted, as these could (most likely) be considered outliers anyway... if the reason for the low weight is because of the grower or the weather and climate.

1/10/2006 11:25:52 AM

gordon

Utah

so a typical pumpkin produces around 400 seeds right ?
where are the other 300 ?

... outliers ? planted with poor results ... grower cause ... seed cause ?
are there a couple dozen or more 723's sitting in garages and basements around North America ... but their owners don't grow anymore...or they have moved... or they passed away or something else...

1/10/2006 1:31:35 PM

huffspumpkins

canal winchester ohio

The pumpkin mafia...lol, Go against the family & you'll wake up with SVB's in your bed. Or worse...You'll get the Rhode Island message, a Jack-o-Lantern wrapped in newspaper that means your pumpkin patch sleeps with the fish's....

1/10/2006 1:39:37 PM

Eugene,OR

Yeha, but if you are in good with the pumpkin mafia, you will be rewarded!

1/10/2006 1:52:23 PM

Andy W

Western NY

ha ha, yep. shortly after 10:26am they got to him.

1/10/2006 1:52:58 PM

pgri

The RI pumpkin mafia is great support system and not only for the 1068 Wallace but also the 500 Wallace!! Our 2006 plannning meeting is Saturday to see who is NOT swimming LOL. PS Tremor where is the Governer of CT sleeping these days. LOL peter

1/10/2006 2:52:44 PM

pgri

Clarification EX GOVERNER

1/10/2006 2:55:35 PM