Er'cana pellets

belford · **Joined:** Thu Jun 08, 2006 7:01 pm **Posts:** 1890

I don't mean to be disparaging... out of character. (In-character, I'm still shocked at your data handling practices. :-)

I was tripped up because I was expecting one thing from your teaser, and got another. I apologize for getting sharp about it.

I have some experiments I want to try, but I'm gone all next week. (And preparing to be gone for the rest of this week.) I'll post when I have my ducks in a row.

BrettM · **Posted:** Thu Jul 19, 2007 1:14 am

belford wrote:

I don't mean to be disparaging... out of character. (In-character, I'm still shocked at your data handling practices. :-)

IC you're shocked that a writeup intended for a general audience makes its presentation in a way that would make any journal editor howl in derision and throw the paper across the room? Now I'M shocked.

Quote:

I was tripped up because I was expecting one thing from your teaser, and got another. I apologize for getting sharp about it.

No need to apologize. I didn't take it personally, and you are, of course, correct about the flaws. I'm sorry if I did a poor job of managing expectations, though. The graphs do have some value in showing that the oven-controlling functions are pretty funky, with some surprising discontinuities, as well as showing which parameters are most responsible for achieving those boundry conditions and discontinuities. Those are the places where pellet behavior suddenly changes radically, which was what the "research" was trying to get at, rather than the DRC-provided cheese of the KI points.

Quote:

I have some experiments I want to try

I'm looking forward to seeing those. Ever consider applying for a position as an FPI Research Associate?

Seriously, that is the kind of thing I live to do in the D'niverse, and I just invented FPI some time ago as a way to put an in-character gloss on my normal activities, though this is the first time I've used it to support actual story. I have been thinking about expanding operations to bring in other people of like mind. Perhaps someday FPI will be the equivalent of the D'ni Guild of Analysts. For some projects it would be much easier to have a team collecting the data so that it doesn't take 30 years to gather enough to be worth analyzing.

johnsojc · **Posted:** Thu Jul 19, 2007 1:21 am

Brett,

Is it my imagination or are all those oven response curves identical except for the scaling along the temperature and nutrient axes? It looks like ovens 1, 2, and 3 are just slices taken from the beginning of the response curve for oven 4.

(I would really, really like to see the raw data for these curves

)

oyferder · **Posted:** Thu Jul 19, 2007 10:38 am

I have a question about those pellet nutrient graphs by FPI

Is the goal to have as high pellet nutrient value as possible(is it that which gives us as many pellet points as possible), or is it too high nutrient value that gives us explosions?

Thanakar · **Joined:** Thu Jun 15, 2006 8:45 pm **Posts:** 620

Gelwyn wrote:

This recipe seems to give the best results -

[spoiler]Super oven (when facing the pellet machine, 1st on the left) - 19.5/50/19.5

Other ovens - 19.5/50/50

Takes 2 hours to cook and averages about 960/pellet.

There are recipes that score higher and take longer to cook which are better if you only have time to dump the pellets once per day. Two hours is just about right for me to do at least 2 batches per day.[/spoiler]

I've actually found nothing better than this recipe, even the three and four hours receipes give me the same average yield per pellet. Ingis' recipe averages me 970-980 per pellet (over 300k now), maybe you need to verify the your ovens, that 19.5 can be tricky, becuase you should be scoring higher than 960 avg.

BrettM · **Posted:** Thu Jul 19, 2007 11:28 am

johnsojc wrote:

Is it my imagination or are all those oven response curves identical except for the scaling along the temperature and nutrient axes? It looks like ovens 1, 2, and 3 are just slices taken from the beginning of the response curve for oven 4.

It's your imagination. Each oven does have an individual response curve, and they are different. The differences among the first 3 ovens are clearer at higher cooking times, but you can see them even in the first set at time 20. (Count the bands, for example, which show that some of the curves are steeper than others.) Each oven also has a different set of caps on each setting, beyond which the values remain constant. I cut off the chart data at those caps, which is the main reason for the scaling differences Belford is complaining about.

oyferder wrote:

Is the goal to have as high pellet nutrient value as possible(is it that which gives us as many pellet points as possible), or is it too high nutrient value that gives us explosions?

In character, FPI refuses to answer any questions about how exploders are produced. But, OOC, you are correct that the exploders occur at the highest levels of nutrient values, so the optimum orange-glow recipe is below the peak of the combined oven response.

I have no idea how the points are determined, except that they aren't simply equivalent to the nutrient value. The points seem to go up until you hit a combined value of around 200, then start going down again until they reach zero at the exploder level. I don't have enough data to tell me if the points just follow a bell curve or if it is skewed somehow. (Though I suspect skewing, where the points fall off on the far side faster than they climbed.)

The steamers occur at the negative values. It is interesting to note that oven 2 is incapable of producing a negative value at any cooking time. (Another example of the difference in the curves for johnsojc.) So the best strategy for steamers is to use low settings on oven 2, such as 5-5-5, forcing the response to zero so it doesn't counteract the others.

johnsojc · **Posted:** Thu Jul 19, 2007 12:33 pm

Quote:

It's your imagination.

I wasn't saying the curves on the FPI graphs look the same, just that they all seem to be an inverted parabola (or part of one) of the same basic shape. If one of the lesser ovens is 25% as efficient as say oven 4, then the curve for the same time interval would approximate the first 25% of oven 4's curve stretched out over that time interval. Of course the efficiency differences reflect in the different values of the nutrient level and the settings of amount and temperature settings required. Then it seems that perhaps some constant is added (or subtracted) to move the parabola around on the graph.

My tests confirm that orange glow pellets occur with values of 1 to 200. The white glow (40-30-20) is a point about 237 IIRC. The lowest value I have produced was -299 for a high sizzler. The highest was in the 350-360 (edit: looked it up.. value was 323) range which was a high exploder.

Points seem to have no rhyme or reason based on nutrient value. For the orange glow range of 1-200 the points (theoretical) go from 5 to 1000 (less 0 to 5% randomly calculated). Yet the white glow with a nutrient score of 237 rates only about 430. Outside the orange glow range (with exception of the white glowers), the points are all zero.

zander_nyrond · **Posted:** Thu Jul 19, 2007 1:04 pm

Jahura and i seem to have found yet another variable...the 30/30/30 recipe yields significantly higher point averages (980-990 for her--i've only done one batch, but scored appreciably higher) if you drain the water from the top two tanks on the display (but leave the paddles turning) and stop the paddles on the lower two tanks (but leave the water in). This, as she pointed out, produces a reasonable facsimile of the Minkata glyph, with the broken bridge between forming the crooked line.

i've noticed that different people get different results with the same recipe, so if anyone would care to try this and let us know how it goes for you, we'd bee grateful.

johnsojc · **Posted:** Thu Jul 19, 2007 1:41 pm

zander_nyrond wrote:

Jahura and i seem to have found yet another variable...the 30/30/30 recipe yields significantly higher point averages (980-990 for her--i've only done one batch, but scored appreciably higher) if you drain the water from the top two tanks on the display (but leave the paddles turning) and stop the paddles on the lower two tanks (but leave the water in). This, as she pointed out, produces a reasonable facsimile of the Minkata glyph, with the broken bridge between forming the crooked line.

i've noticed that different people get different results with the same recipe, so if anyone would care to try this and let us know how it goes for you, we'd bee grateful.

I think you are only the victim of a random number generator. 30-30-30 always produces a nutrient value of 200 and a max score of 1000. The pellet points displayed are are reduced by a random amount from 0% to about 5% of the total point score and any decimal points are truncated. This is what produces the variance you see in point scores for a single batch.

IngisKahn · **Posted:** Thu Jul 19, 2007 2:21 pm

I'll add that pellet points linearly decrease back to 0 from 200 to 266 nutrient points.

And if you think the nutrient curve looks familiar it's A sin(B) sin(C).

belford · **Joined:** Thu Jun 08, 2006 7:01 pm **Posts:** 1890

"OOC, you are correct that the exploders occur at the highest levels of nutrient values"

I think I missed a step in the logic. When you say "nutrient values", you are referring to the meter in the silo?

How did you get that value? Just eyeballing it? And how did you derive the negative values?

johnsojc · **Posted:** Thu Jul 19, 2007 11:29 pm

IngisKahn wrote:

I'll add that pellet points linearly decrease back to 0 from 200 to 266 nutrient points.

And if you think the nutrient curve looks familiar it's A sin(B) sin(C).

Hmmm... that would explain why orange glow pellets and white glow pellets can produce the same pellet scores but register a different amount on the tester (70% for white vs 40-60% for orange) and produce a different reponse when dropped. Everything is based on the so-called nutrient score.

belford wrote:

When you say "nutrient values", you are referring to the meter in the silo?

How did you get that value? Just eyeballing it? And how did you derive the negative values?

The numbers bandied about are confusing since some are not easily detemined.

There are really 3 numbers (4 if you count the oven sliders):
1) oven sliders - 100 positions, either referred to as 0-50 or 0-100
2) nutrient score - seems to range from about -300 to something in the +300's (I haven't found a score higher than 323 but I haven't tried to find one once I got that score)
3) pellet score on the KI - variable depending on the nutrient score. For nutrient scores from 0-200, the max pellet (KI) score is 5x the nutrient score less a random value from 0 - ~5% of the max pellet score. From comments above, nutrient scores above 266 and less than zero provide zero pellet points.
4) the silo testing meter - I have not figured out how this is determined (based on nutrient scores) except for the orange glow pellets where it is well behaved and ranges from 40-60%. White glows are 70%, exploders read 80-100%, steamers are from 10 to about 30% and the duds come in just above 30%.

The numbers come from a detailed analysis of the oven data produced and a little jiggery-pokery with the numbers to make certain calculations come out consistently. Did I mention a lot of guessing?

belford · **Joined:** Thu Jun 08, 2006 7:01 pm **Posts:** 1890

I understand guessing and jiggery-pokery. :)

Brett: "Each oven also has a different set of caps on each setting, beyond which the values remain constant. I cut off the chart data at those caps, which is the main reason for the scaling differences Belford is complaining about."

Ah, then let me refine my rapier-like criticism...

(Really, just a comment on data visualization.)

Each of your charts shows a slice through a particular oven's response curve for a particular time setting. That curve shows a bunch of information, but it has two notable features (landmarks): the peak of the curve, and the border where it flattens out.

It sounds like you cut out the flattened-out region because it was boring. But that actually obscures one of the interesting features. As a student, the fact that it flattens out is something that I want to learn from the graph.

BrettM · **Posted:** Fri Jul 20, 2007 12:16 pm

belford wrote:

It sounds like you cut out the flattened-out region because it was boring.

Nah. Mostly because MS Excel kept complaining about the amount of data I was trying to shove down its limited throat, and I didn't feel like hauling out the big guns (SAS:Graph) and figuring out how to produce the kind of graph I wanted.

I would like to try combining time and temperature into one category and produce a long graph covering the full time scale, though. (I would leave amount on the short axis since the response to that parameter is linear for a given time and temp.) Maybe I'll get around to that one day.

oyferder · **Posted:** Sat Jul 21, 2007 3:17 pm

belford wrote:

I think I missed a step in the logic. When you say "nutrient values", you are referring to the meter in the silo?

Yes the nutrient values(from the FPI graphs) must refer to the meter in the silo, I think, or, more proper said, the meter in the silo shows us the nutrient value.

Explanation:
It is easy to see from the graphs that the 50/50/50-combination(or 100/100/100 as someone may prefer to call it) gives a very negative nutrient value, and we know that this combination gives us 10-20% percent on the meter and steam and bubbles.

0.5/0.5/0.5(on all scopes)(or 1/1/1) has a nutrient value on zero, and this gives us about 30% on the meter and duds.

Someone tried the combination that gives the max point on the graphs and got 90% and an exploder

I think that should be enough proof

I am not sure if the combination that gives us an average pellet score(not an individual pellet score) on 1000 is on the tick mark on the meter(that lies on 50%), or if it is on 60%(where I have got all my best), or if it is possible to get a batch withe an average on 1000 at all.

Er'cana pellets

Who is online