PPS.... I wrote:
So for short sales, the annualized IRR is
(1 + (s-p)/s)^(365/days) - 1, where "s" is the initial
net sales proceeds, "p" is the later net purchase proceeds, and "days" is
the hold time. For your example,
that would be about 615.56% annualized IRR.

Needless to say, the __annualized__ IRR is misleading, IMHO.

I don't know what conventional practice is for short sellers. But IMHO, for
hold times of less than 1 year, I would use the simple rate, namely (s-p)/p
formatted as Percentage.

Granted, that does not take "time value" into account. But I think any
attempt to do so is misleading.

Consider your example, but with a hold time of 14 days. The annualized IRR
would be about 167.50%. Do you really __feel__ (subjectively) that the
"rate of return" is more than 3.5 times better when the hold time is only 7
days?