I am designing a fund-switching strategy for my 401k account with fidelity.I have a 30 day rule for holding funds, so i am trying to capture those longer moves in the funds. I can switch from one fund to another no more frequently than the 30 day limit. My system is fairly simple.Just a combination of price rate of change and a indicator that measure's how smoothly price is trending.I am unable to really measure what the optimal lookback period should be for the time frames i am trying to capture. I am thinking somewhere between 3-6 months, but would be curious if others have done this and have some feedback they would like to share. Thanks for listening.
I have not done that but in statistics 101 I was told that 30 samples can be considered to be a "large" sample. I always start any investigation into the best parameter with 30 and work from there.
http://www.formularesearch.com/ I know Freeburg has done some work on mutual fund sector timing that is good because I've seen him speak on it and it was solid. However I don't know how or if you could access it. There is one other trader that does real well with it that he mentioned but I forget his name. Freeburg's stuff probably is more indicative of how you would do it your self the other guy I'm thinking of would be more of a "service" black box type approach which may not appeal to you. Either way, the above link has Freeburg's email and phone number. If your intersted in the other guy, I think he would know who I'm talking about. Far from a vendor hawking his wares, he's the real deal having been at this for god only knows how many years, doing real quantative research not bollingerband BS or candlestick witchcraft.
Presumably those 30 days are calendar days, which translate to roughly 21 trading days per month. This is a lookback period of 5 weeks (current week plus the previous four). I would double this to a lookback of 9 weeks for a margin of safety, but any more than that and you probably won't be as responsive to fund-switch signals as you would like. So a lookback of 9 for weekly data or 41 for daily data would be a good place to start until you come up with a better lookback. Truly optimal lookback calls for application of adaptive techniques that you may not have the time to research or the resources (like advanced trading software) to apply. But if you're determined, look at Jurik's CFB indicator for auto-adjusting lookback periods. I'm not saying you should get it, just saying it is one method of obtaining (near-) optimal lookback periods. http://www.jurikres.com/faq/faq_cfb.htm#results hth and good trading, kut2k2