I've ran 10,000 of simulations using years of historical data. I'ved used SMA and Exponential MA and among others. Use too small of a period and trading fees eat you alive, use too long of a period and you miss the swings. Calm markets that move sideways eat you alive, you need large market swings to make money. So, I've looked multiple years back seeing if a once in 5 year swing makes it all worth while. Nope. There is no sweet spot.