I'm not a data scientist. Spent quite some time on this thing. Ended up using a thing called "backward expanding weighted moving average", but the trick was to calculate the weights iteratively (i.e. amount/number_of_intervals). Might be able to refactor and use np.convolve instead which should be more performant but there's no need for that at the moment.
Cool stuff! I guess the results feels intuitive somehow, considering Bitcoin has mostly been up in price between in the last 5 years.
How I see it: DCA buys you peace of mind, in exchange for less profitability.
I'm not a data scientist. Spent quite some time on this thing. Ended up using a thing called "backward expanding weighted moving average", but the trick was to calculate the weights iteratively (i.e.
amount/number_of_intervals). Might be able to refactor and usenp.convolveinstead which should be more performant but there's no need for that at the moment.If that data is correct, as long as you don't lump sum at the top, you'll outperform DCA
Good time to lump sum in a dip
Now would be a good time too
So lump sum beats DCA 80% of the time.
By how much?