I experimented with this extensively and found it made no difference. If you are finding it is making a big difference, perhaps: Check you're calculating the volatility of daily returns, not the volatility of price. Check you've shifted your returns vector backwards a day before you calculate today's volatility, otherwise you introduce lookahead bias, which will be inversely proportional to the lookback time. To be honest any time I found something that 'didn't make sense', I just had to go through my code very carefully, and I would almost always find a mistake somewhere. If you have a Sharpe ratio of 0.8 for a single instrument over 30+ years, that's too high- something's wrong. 0.3 is more likely.