Optimization run time longer for subset of instruments?

[darmbk]

Hi all,

I have run an optimization over an instrument list that consistently takes about 3 hours. I later decided that I wanted to run the same optimization over a subset of these instruments so I created an new instrument list. However, this smaller instrument list suggested it was going to take 18 hours! I aborted this and selected the original list which again forecast a 3 hour completion time. I tried the subset a second time and now it forecasts a 6 hour completion time? What could be going on here?

Thanks,
darmbk.

[NinjaTrader_ChelseaB]

Hi darmbk,

One possibility is the estimation from the first test may be including the time in needs to download historical data needed for the test. After being downloaded once it shouldn't be needed again. (Possibly on a longer test the data may be downloaded and NinjaTrader needs to sort through it before re-estimating the time left)

Another possibility is the results from some of the tests in the optimization are cached without the code being changed and the same set of parameters used for that particular test.

And last, the estimation is recalculated as the test goes on. I wouldn't rely too heavily on the estimation being extremely accurate.