So the model gave me no advantage over anyone who can read a free chart when the index was overbought or oversold. And that is a lot of the time. The model did give me a big edge when the stochastic turned around in the range between 20 and 80. I could predict that. I tended to get very jittery in these overbought and oversold zones and ended up losing most of the profits I made the rest of the time.
Today, I came up with a rule which uses one of the forecasts I was already generating from the model to give predictions for turning points in these overbought and oversold zones. I would say it works about 1/2 the time. Its performance is halfway between a trading algorithm that reverses from long to short or vice versa at the end of days when the stochastic crossed its moving average during that day and a model which shows perfect foresight and reverses position at the beginning of days that see stochastic crossovers. Let's see if it adds to my performance.
For what it's worth the model is not formally predicting that the stochastics will cross on Friday using the rule I found to be optimal. But unless there is another strong rally it's going to be hard for them not too and there are clear sell signals in palce on the S&P 500 and Dow indices, so I'm skeptical about that strong rally. The market has, however, continued to be stronger than I expected.
4 comments:
What's the point of having a model/indicator that works half the time? Isn't that the same as flipping a coin?
Don't take this the wrong way but it seems you need a rule about stop loss and profit taking more than modeling entry/exit. BTW, are you using the web interface of IB or your own program interface?
This model works rather differently in concept to most people's TA systems. I treat each day as a separate trade. An implicit or explicit decision is made each day on the direction of trading. Each day there is a stop loss of a 1.25% movement in the index (this isn't including my "overnight trades"). This is only triggered on dramatically badly wrong calls.
To understand why a "right only half the time" indicator could be useful: The default is to maintain yesterday's position. If the stochastics actually cross over in the overbought or oversold zone we want to change direction. Assuming we can't predict the crossover at all is the base case. In that case we'd wait till the end of the day to make the reverse trade after we observe the stochastics crossing. The other extreme is if we knew ahead of time that there would be a crossover and made the trade at the beginning of the day. The new indicator adds about half the performance addition that would come from perfect foresight to the base case. That's a huge gain.
Up till now in these situations I have been trying to make a trade intraday. But often the market reverses itself again intraday and negates the signal. So that could get really screwed up.
The model consists of 3 time series models estimated in an econometrics package and an Excel workbook. I just need high, low, and close data each day which I enter by hand in the spreadsheet. I don't know how to do any of that fancy stuff and am sure it wouldn't work with my econometrics software.
Moom,
Thanks for that explanation. I get it now.
Yes, so it comes down to you can you get in a day early on the move. And about 1/2 the time you can. It's never going to say to go in the wrong direction in this situation either hold the position or change it.
Post a Comment