I tend to agree with the last post over the previous answers.But, as i have given though to this question, i would like to add.
The Sharpe ratio depends on the timeframe used. If you want to use it to eveluate high-frequency strategies, forget about the textbook formula and think yourself. First of all the risk-free return is irrelevant in evaluating your return, so drop it. Next, your return depends on leverage used, so forget about using the margin requirement as a measure of how much capital was used.
Now that we are thoroughly lost, lets remember what is it that Sharpe ratio is trying to measure: it is a return to variability measure. So compute it per trade, simply as
average return
S = -------------------
stdev of return
Then you are supposed to maximize the ratio by changing the parameters of the model.
What i'm trying to say dont use it to compare completely different strategies in different markets, compare same strategy for different values of parameters.
Obviously a bigger value is better, but i would *not* say if it's less than S < 1 thats too bad. You could make a ton of money with a S= 0.01 strategy. For example, if you computed the daily Sharpe ratio for MSFT in the 1990's you would get something like 0.1. That didnt make MSFT a bad bet did it?
However, the S will affect the choice of position size. For that look up the literature on the Kelly criterion. A small S will force you to make very small bets, which may not cover fixed costs. (depends also on whether the trades are serially uncorrelated)
The small bets issue is very serious by the way, if your optimization tells you to bet on a 1/3 of an SP contract given your capital of say 100K, what are you gonna do? Trade SPY instead???? I suspect some naive people here, inquiring about automated systems possess trading capitals much less than that.