Numbers trendlines not working properly

my experience is that the best fit line doesn't include enough significant digits.

Look up a web site where you can paste your data, then let it compute the coefficients.

like here:

https://arachnoid.com/polysolve/

I believe a polynomial trendline is far from what you really want anyway. It might "fit" okay in this small section of temperatures but it would get all squirrelly farther out and is most certainly not the correct fit scientifically from what I see. It looks more like an exponential or power fit but it seems to be outside of what Numbers (or Excel) can provide. That probably also explains why the polynomial coefficients are wonky. Do you know what the general relationship is between the temperature (X) and your Y data, what the formula is supposed to look like?

Also, that was not a 6th order poly equation. I noticed with my fake data that going from a 4th to a 5th changed all the X^0 through X^4 coefficients but did not give me an X^5 term. Going to a 6th order, I lost the X^4 too. Excel did not have that same problem but it did something just as bad. It seriously rounded off the coefficients, down to just a single digit of precision each.

My assumption is it is beyond the abilities of the algorithm Numbers uses, and beyond Excel's too, and maybe it is beyond the precision of floating point math.

look at all the decimal places! Wow! that's a lot