Since both are very rough (llinear and exponential extrapolations) I was hoping to find more skilled people in this forum being able to fit a function usable for (better) predictions.
So, where are the R and Python magicians?
I would say, itâs like predicting the weather : to much unknown variables and chaos to predict more than short term possibilities
I enjoy the idea here tho
Base on what was annonced publicly by Purism, the first one seems much more accurate, but as @eugenr noticed, may be a bit under the prevision made by Purism
An exponential curve like the second one, seems improbable in such short term
Data science is not some of my specialities but my few cents to this topic are:
How good is the data quality? I havenât had a look on how the information is gathered - just out of posts in your linked forum topic? Is it all valid, no doubles, etc.?
The timeframe is maybe a bit too short to make a valid prediction
Keep in mind such things as CPU shortage, Chinese New Year, etc.
My personal opinion, it looks linear.
Edit: I think it could also be interesting to plot the monthly shippings (not cumulated)!
Dates in the âshippingâ column have been reduced to a single date, if doubled or trippled entries have been available.
What is a valid prediction in that case? My approach, or more question to community, is more like âbetter-than-nothingâ. At least for all the non-early-early backers, non-early backers and non-backers of the original crowdfunding campaign.
At this point nothing is âbetter than nothingâ because you cannot possibly have enough data to make an accurate prediction since the people building and shipping the phone do not have enough information to make an accurate prediction. You all keep feeding this impatience monster you keep as a pet and it is impossible that it will turn out well. Youâll either guess that shipping will happen sooner than reality and be angry when it doesnât prove true or youâll guess that shipping will happen later than reality and be angry because âitâs taking too long.â
No good will come of this, I promise you, because there is ample evidence on this forum that exercising impatience has made people angry ever since Aspen.
Yes but I meant that you would have to keep in mind what happended on that dates from your historical data and not only what could happen in the future. E.g. when there would have been no shippings due to some crisis for some months, nobody would assume there wonât be any shippings after that crisis.
More seriously though, I still wish that Purism would release some accurate âwe shippedâ data, or make some predictions.
Dealing with future uncertainty is not beyond the wit of man.
Prism has already committed to ONLY estimating a certain distance out due to availability uncertainty in the supply chain. I think it is reasonable for Purism to stick to this.
To expand on what Gavaudan was saying; my personal observation is that much of the frustration from past estimates stem from purism setting and then not meeting expectations and/or not resetting expectations promptly. (The details and examples of this have been discussed ad nauseam in other threads and if further discussion is desired, that is best suited for other threads; not to derail this thread)
I would like to see a bit more transparency from purism on the shipping progress this far, but that is a personal preference; partially stemming from how often I have seen purism tout itâs transparency. Iâm not by any means saying what they should/shouldnât do on that front just expressing my preference.
I also think estimates/predictions should have upper and lower bounds with confidence bands and beyond 6 months out should probably fade to grey as that is so much uncertainty as to be pure guesswork. A single line estimate often gets taken as having near certain probability/high confidence and in this case I donât believe that is reasonable.
I am not asking Purism to make estimates. They, as you mentioned in your post, have made their statements on shipping predictions. I am just curious how a more skilled person with methods of statistics / stochastics / data science / you name it, would fit a function in here to have a simple data perspective.
Agree. That would be good to see in a plot as well.