Going back to the original post: "Seriously, my trading experience has shown me that trying to predict the direction is worse than flipping a coin" Question: for stocks in secular straight-line uptrends, what is the statistical risk of the stock not continuing to go up? Yes, all uptrends must end -- but to jump on and sell 10 or 20 bars later, is not there a better than 50-50 chance of making a profit? http://stockcharts.com/candleglance/?NDAQ,SKYW,ASF,QDEL,ASGN,bbsi,bbd,tie|C|M21 Does price persistence exist, and does it influence the statistical bias of short term trading outcomes?
Yes, price persistance exists. simply do a lag 1 regression. further, back to the coin flipping example. it is helpful to think of it in bayesian prob. take a large sample of coin flips. what is the prob of a flipping a head, given the previous 5 flips are heads? if the prob is greater than 0.5, then either it means that the experiment is biased, or the flips are not independant.
This is no theory, and I'm not paranoid. I stay well clear of stocks such as INTC for the very reason that they are so heavily manipulated intraday. Better to trade a stock where I'm the big cheese with 4,000 shares, than where I'm just a fly waiting to be swatted.
The question is: are you merely correlating the coefficent of the price and volume movement that you've just observed in the past, or are you predicting a short-term future move? I believe it's the former. If I am watching the tape and a Level II screen and I see volume and price moving in a direction, and I get in quick and take a short ride for a profit, that's reasonably doable. But, to just blindly follow a simple moving average trendline based upon price changes doesn't work any better than closing my eyes and randomly pushing "buy" or "sell." At least, it's never worked any better for me. Which reminds me of a funny incident. I'm lying in bed one morning, watching the market on my laptop, and my cat jumps up on the end table. I reach over to scratch him, and of course, he steps right into my lap and onto the computer keyboard -- and executes a 1,000 buy order for a stock that I had already input, but hadn't actually decided to buy. I thought I was gonna have a stroke, but damned if the stock didn't start to climb, and about 60 seconds later I sold for $180 profit. And, no, you cannot borrow my cat.
NOISE By the way: have you ever posted the solution to your test? PS: Sorry, I didn't mean to imply you're paranoid. Sorry.
I did a Monte Carlo simulation and I applied MACD on it. Can you see divergences that seem to predict the price? If you do, you can use technical analysis to break the casino's house.
Fair, but maybe you should include some element of stochastic volatility like some model of volatility like GARCH, etc.
I like to think the flips are not independent because traders are influenced by yesterday's price. Prices change because of actual or anticipated events. But in momentum stocks the chart builds in a sort of inflation (expectation of higher prices). If one considers this chart: The linear trend says over $16 by the end of December. I'm wondering if it's possible to estimate the probability of the stock hitting that target. I would think that there is a higher chance of capturing profit in a chart like this than would be a for a "typical" chart, but how to quantify the risk?
To evaluate the probability of the stock hitting a target you can use Hoadley, but this won't take into account the bias you suggest.