Your system could be right but it can still be hard to trade. After you shorted at 858.5, it went straight up till 866. That's a loss of 7.5 points in less than 30 minutes. It will be hard for many day-traders to take that kind of hit. In this case it obviously worked out at the end. I'm not sure if many people can handle 10-15 pts. swing and make 5 points at the end of the day. It will be better to start calculating the max intra-day drawdown too. This way, let's say, after "n" number of observations, the max drawdown is 10, then you have a system that will have 11 pts. as the stop-loss point an entry and an exit.
I have noticed that problem as well. In the past I have tried to limit risk while paper trading by using stop orders. This almost always resulted in less profit than doing it without a stop-loss. If you have the stomach to ride the ups and downs, the system is right far more often than wrong. If I was trading it right now I would trade one contract per $10,000 available cash, and set my orders to accept up to a 50 point swing. That should keep my risk per trade at around 25% and be able to ride out the swings. I still need to do more research on money/risk management. I will think about how I might be able to get it to output a max drawdown. Thanks for the advice...
Thursdays Prediction: -0.03 -Result- Open: 829.91 Close: 835.19 Change: +5.28 Prediction: Wrong Note: Wrong, but not traded since it was not above 1 or below -1. Obviously its up to the individual trader as to what certainty to trade. I believe you would still do quite well trading anything greater than 0.5 and less than -0.5. I encourage everyone to paper trade these signals in a their own way and let me know their results. Paper Trade (one ES contract, Market Orders): (Not Traded) Overall Profit/Loss: $1,775.00 Next Prediction: Friday (02/13/09): Prediction: -0.75 Signal: Wait This signal should not be traded because it is not more than 1.0 or less than -1.0.
See, I would call it a perfect prediction. The SPX eventually went up only 1.45 point and that is a no change, sideways and that is exactly what your -0.03 meant. So as I said earlier you are shortchanging the system by your evaluation. When the prediction is between -1 and 1, you should call it a sideways prediction and expect the index not to move more than +/- 5 points. Also this could be played this way: The prediction is sideways and the market swang downward by 20 points, just like today? Well, let's get long and trust the system, that we would end up close to the open. Just a thought...
Interesting idea... I marked it wrong simply because I don't want to look like I'm making exceptions or doctoring the results. Based on the rules I had set prior, it was wrong. But you make a very compelling case for the sideways indicator. Only thing I would worry about with calling it sideways is that it may not always be sideways with that prediction. However, that prediction does mean the software is conflicted about the direction. In which case the market in general is probably conflicted, resulting in a sideways market. I will definitely try to evaluate predictions that are close to zero and see if they routinely result in sideways action.
Although there is one more hour left, (not counting a 10-15 pts rally happening ) this has been an excellent call so far. A -0.75 indicates for me a sideways to the downside, when most of the sideways action takes place in the red side, and that is what has happened (so far). We went into green territory twice today, but after 5.5 hours SPX is still -2.5... There is 1 problem with sideways behaviour and the last 2 days are excellent examples. Today was a clearly boring a little down, a little up behaviour, yesterday was wild swings in both directions, but ending up close to the open. But those 2 behaviours need to be played differently, and your indicator even if correct, doesn't indicate what type of sideways we are going to get. Nevertheless, good calls in the last 2 days!
You are right... The last two days have shown that it may be possible to exploit the sideways signals. That's something I had never thought of. I know of a way it could possibly predict the days volatility, but it would require a good bit of work on my part. Still it is possible, and I will add that to the "to do" list.
Fridays Prediction: -0.75 -Result- Open: 833.95 Close: 826.84 Change: -7.11 Prediction: Right Note: This signal was not traded because it was not greater than 1 or less than -1. Paper Trade (one ES contract, Market Orders): (Not Traded) Overall Profit/Loss: $1,775.00 Next Prediction: Tuesday (02/17/09): Prediction: 3.28 Signal: Buy This signal should be traded by buying ES on Tuesday at Open (9:30am) and Exiting (selling) at Close (4:00pm). This signal is historically accurate over 80% of the time. Note: Weekends and holidays are bad things for this system. It is making it's prediction Friday night for Tuesday. Alot can happen in that time. In general, I trust predictions made before weekends or holidays less than others. Ideally predictions would come an hour or so before market open, but I didn't think about that when designing the software. I may try to change that.
Tuesdays Prediction: 3.28 -Result- Open: 818.61 Close: 789.17 Change: -29.44 Prediction: Wrong Note: N/A Paper Trade (one ES contract, Market Orders): Buy ES @ 797.00 (9:30am) Sell ES @ 787.25 (4:00pm) Change: -9.75 Profit/Loss: ($487.50) Overall Profit/Loss: $1,287.50 Next Prediction: Wednesday (02/18/09): Prediction: 5.48 Signal: Buy This signal should be traded by buying ES on Wednesday at Open (9:30am) and Exiting (selling) at Close (4:00pm). This signal is historically accurate over 80% of the time.