Hey Hari - Could you please eyeball these hypothetical numbers and see if I have the idea? I would appreciate it!
mmm
12-05-14…278.4
12-05-15…276.0…278.16
12-05-16…275.2…278.08
12-05-17…276.4…278.20
Not quite
12-05-14…278.4
12-05-15…276.0…278.16 That’s correct. Subtract 10% of 278.4 - 276.0 or .24
12-05-16…275.2…278.08 That’s not correct. First, the rounded previous average is 278.2. 10% of the difference between that an 275.2 is .30, so that should be 277.86, rounded 277.9.
12-05-17…276.4…278.20 Continuing in the same way, 277.9 - 276.4 = 1.5 so you should subtract 10% of that or .15 from the previous average of 277.86 to get 277.71. For the next one, you would use 277.1. Note that the second decimal place is meaningless but I use it to control round-off errors. Technically, I believe it is called a guard digit.
If you are interested, I will tell you that this method produces an average of the current day’s weight, 90% of the previous day’s weight, 81% (.9^2 = .81) of the second previous day’s, 72.9% of the third previous day’s and so on, the influence of each day declining by 10% over its previous influence. By 20 days it is down to about 10% and is more-or-less negligible after that.