I have found that a system using the same volatility entry and exit rule but a different time frame makes a huge difference in the results. For example, the results using a 10 minute time frame are much more stable than results using a 5 minute time frame. Has anyone else noticed this in their testing? I suspect there is a reason for this other than sheer chance. Would someone who is knowledgeable mind explaining? Thanks.
is your volatility based on so many bars back, like 20 bars or 10 bars, if so, it's because you taking in more data to look at.
yes it is looking at bars back. however, why would that necessarily make the results consistently better, or worse?
lets say you are using 5 bars back, on a 5 min chart that would look at 25 minutes, on a 10 min chart that would look at 50 mins. Your analyzing more data, which means maybe you are taking a better look at volatility.
let me rephrase the question. does anyone have a systematic way of determining which time frame to trade based on the volatility of that time frame? a wild idea (maybe for another thread)...has anyone tried pyramiding a position based on the same rules but different time frame?
Not true. You are looking at 5 bars in both cases. The corresponding time shouldn't matter. Maybe your answer lies to the fact that the 5min chart is so over-used that it's S/N is much lower making any known and naive strategy to fail.
I did some work regarding optimization of a strategy over time frame in Tradestation. I built psudo data bars in arrays and then did the backtesting on the psudo data with bar interval as an input. The particular strategy I was working on did not vary a whole lot over time frame but apparently yours did. Technically, the work was not that hard to do but in reality it was less work to just manually change the bar interval on the chart and rerun the test. Some people think that a succesful strategy has to work on any and all bar intervals, I don't agree with that much, if you find something that works on just a single bar interval what is wrong with that? If it works it works.