Attached is a variation that multiplies the array by -1 instead of reversing it. This has some advantages. For example, if you are mostly interested in the distance between thw two points (e.g. peak width), errors due to noise will tent to cancel, while with array reversal, you get a systematic increase in width with increasing noise.
Is the peak a well known mathematical shape (Lorentzian, Gaussian, ?) Maybe you could fit it using Levenberg-Marquardt?