Is there a rule of thumb for the amount of permutations a robust system should include during  optimization? Given 5 years of data, is 1,000 too small and 250,000 too large?

Here is the question , what does the optimization surface look like ?. You should use our 3D optimizer and see what the optimization surface looks like.

Another question what is the premise of the system ?. Is the premise sound.

I'm not sure what the surface looks like. I can kinda eyeball it for local maxima and global maxima, but not precisely.

On the question of the validity of the premise, the answer is no. My first Traders Studio system is a dual moving average crossover with a noise filter. Obviously I don't need permutations where the fast moving average is of larger length than the slow moving average, but I'm burdened with those results unless I uncheck each undesirable permutation. Is there a quicker way of ensuring that only permutations where the fast moving average is a lower length than the slow moving average?

The 3D moving average tool let's you filter out invalid combinations like short term moving average longer than long term. This is part of gold membership , or is on sale on site by itself for $69.95, and about $56.00 with our current 20% off sale.