What about if you could somehow get a statistical basis for what happens the next day in the Dow based on movement in the last hour of the previous day? Aren't stocks very likely to move in same direction to start the next day based on whether they close very near their high or low the previous day? So you can use this movement, along with your indicator and your money management expertise you shared with us a year ago and play the odds if they are in your favor. Just a thought.
The model is only 55% accurate in predicting direction, and the average magnitude of the move goes way down. I think if I want to play a directional model, I'd be better off doing an intraday volatility-breakout method ala Larry Williams. I've also looked at combining some traditional TA methods on when the model predicts a 1.2% or greater move, such as volatility breakouts, channel breakouts, some standard trendfollowing type methods to take advantage of a wide range day, but results were only marginal. Currently, in order to predict a minimum of a 2.0% move, the % success is only 62%. But the standard deviation is higher than I'd like quarter-to-quarter testing for past 6 years. There are two quarters with < 50%. With the 1.2% model I maintain a minimum accuracy of 76% in any given quarter going back 6 years.
Try S&P e-minis, and use inandlong's "line in the sand" system. Be long above the line, and short below it. You are looking at a 10+ point move. Start with an epsilon of say 0.75. Use P+epsilon and P-epsilon as entry triggers, and P as a stop. If you get stopped out say 3 times then either call it quits, or increase the epsilon and continue. Once your target is reached, close the position. I think you said a 60% chance of a 2% move, so maybe close 40% of the position and let 60% ride. You could approximate the % by using 2 contracts (or multiple of 2) and close half of them. Obviously you can play with these numbers and find something appropriate.
I've tried testing something very similar to this, as well as a variety of other classic intraday break-out and volatility methods designed to take advantage of a large move. The problem is that many times the large move is caused by the overnight gap. My model predicts the magnitude of the next day's close, but not the direction. For example, this past week the model predicting at least 1.2% move (close-to-close) for 10/15 and 10/17. We ended up getting a 4.8% and 2.9% move. But notice that the majority of the move was due to the opening gap. I would've been whipsawed trading any sort of intraday breakout or trend-following method, but would've done very well being long delta. Unfortunately not all days result in large moves like this. At only 1.2% move, it's a little harder to make money on a straight straddle. I could look at filtering out all gap moves, say more than 0.5%, and look at the remaining days to see if an intraday breakout technique would be profitable. I have another arb model that may work well well the aforementioned inandlong method (this model can be traded by retail trader using IB). But it requires following the overnight action b/c gap moves will crush the arb model. I'll post more on that after this thread is complete.