Assume you have a signal which with (on average) 60% probability predicts directional change in US treasury yields on a daily basis, i.e., every morning it signals buy/sell of a US treasury. Using this signal one trades a naked long/short position every day buying/selling in the morning and getting out of the position in the afternoon (flat at the end of the day). Now, this strategy is always profitable at longer horizons (>= 6months) but can experience long series of losses and is thus fairly volatile and probably unacceptable to investors (shows a Sharpe ratio of 0.7). I am looking for suggestions on how to reduce the volatility of such a strategy using very liquid instruments (treasury futures options?) with the condition that the "hedges" could be executed electronically (so that the whole strategy can be fully automated).

Brainstorming okay? Overlay a classification system to high-grade the trades, cut down on the total number and boost the win/loss ratio.

That's the same as saying, find a signal with higher success rate. That would be great but let's assume I don't know how to do that (or it is impossible).

fair enough. have you looked at maximum favorable and maximum unfavorable excursions on winning and losing trades? perhaps that could give you an idea where to exit (essentially) with a future or where to position the strike prices for an option.

Kelly is not a good idea as it makes things volatile even if you have a fixed probability of success (which I don't).

If you have a series of wins and a series of losses the Kelly would be pefect, note that you don't have to use the fully kelly amount, you could use half kelly. Your one key area which will definately decrease volatility is diversification, diversify as much as you can. No matter how good your strategy is if you don't have a decent level of diversification your vol levels will like like a gambling system and you would be focusing on risk of ruin stats and that's all.

jurowilmott1a series of questions here ...have you tried to identify the "phases" where your strategy is successful ? in short, what i am aiming at, perhaps, is that you might need to look a little bit closer at your strategy and perhaps a "key" indicator is missing -- or is the strategy perhaps completely random and your success is devoted to market fluctuations ? how often do you calibrate your strategy to the market?moreover, is there another way you could trade it ? you mention a 60% hit rate. that alone suggests to me that there is money to be made; perhaps not holding the trade for so long ?i am not saying that you are not on to something here, but there are so many things that could impact your performance. btw, do you have to be in the market all the time ? my point here is that, if you (your model) dont know what to do -- then dont do anything.nicolai

Thanks for suggestions Nicolain, I asked myself those questions too. The strategy seems to have phases where it works better than 60% and others where it's worse... Unfortunately, I can't determine exactly what makes it work (other than the original idea included in the constraction of the signal). Based on statistical test I have a good reason to believe that it's not by chance (note: I didn't try many different strategies one of which turned out to work...)It is definitely independent of the market going one way or the other (up/down). You are right, I might not have to hold the position all day, in fact, I am looking at different holding periods now but I am missing some crucial data that would allow me to do more research in that direction-but will give it a try to the extent possible.