Someone recently posted Control Charts/Nelson Rules - finding non random data points within a random environment. I like this approach a lot. Does anyone want to share methods on how they determine probability parameters for entering a trade? For instance, could one calculate the average range of a opening 30 minute bar every day, and further slice it by day of the week, etc. and say ok, it's Tuesday, the down bar is 1.5 X it's average range. With a minimal gap, I know there's a 60% chance the market will move this distance/direction... OR The 30 minute opening down bar is 1.5 X it's average. Because it formed in this specific way, I anticipate a 60% chance the next bar to only move 50% it's average range with a normal distribution around the opening bar's low. A follow up question is then are past statistics significant enough to enter trades or just a supplement to manage trades. As usual it's worth fighting off anyone chiming in to ridicule me in hopes of finding someone that genuinely wants to help. Thanks ET nation. BD
Get a PhD in one of the top U, then got a job in GS or MS. You will learn when you are part of the team.
In long run, pure statistics trading doesn't work. You have to continuously gauge the market's strength in the direction of the current trend, then make your decision accordingly. If you do it on several time frames is better. Think of an analogy with surviving in a land where you don't understand the spoken language.
A person considering taking the offer of the market does not overstep his bounds. I trade with 100% probability because I use the mathematics the markets dictate. I am not ridiculing you. I enter the first bar of the open as a consequence of the sentiment of the prior day's calculations using the mathematical requirements dictated by the market. As the day progresses I continue to take profit segments, segment by segment. Markets have no noise, no anomalies and no flaws.
If I notice a pattern that seems to repeat, first I eyeball it to get an idea for favorable and adverse price excursions. During this phase I also look at the bigger picture to see if there are any immediate filters that might be useful. Based on the "eyeball" phase results, I develop a few sets of rules for entry and trade management and apply these rules to every appearance of the pattern, entering all the results in a spreadsheet, for a period of time that encompasses varying market conditions (range, trend, strong trend, and chop), and provides a reasonable sample size. As an intraday scalper who uses a 5min/1min chart combo, a reasonable sample size for me is the number of trades presented over a three to four week period. If any of the data combinations produce a win rate/R:R combo that offers consistent profits throughout the varying market conditions, I test in real time (usually on paper) to make sure I recognize everything at the hard right edge. It's time-consuming and tedious work, but the reward for the effort has been excellent.
i too trade with 100% accuracy when using the mathematical dictates of yesterdays data. I am not ridiculing you. Papertrading yesterday offers excellent risk reward opportunities when you look at the chart first.
Go get yourself very well trained in the area of statistics. Then go learn a high end stats package like S-Plus. Very popular with the quants. One of many good ones. Then learn how to data mine. Just that simple. Should take you maybe four or five years...
The probabilities and statistics are not enough in and of themselves because big quant. players are running the same statistical studies and making the same conclusions and making trades that arbitrage away those inefficiencies, so the opportunities do not last long. They have lower transaction costs than you do, so with your costs, there may be little left. So you have to find pricing inefficiencies that they cannot exploit. Maybe non-scalable ones, for example, since they want trades that will scale. There are other possibilities too, such as riding their waves, but these require enormously intensive analysis. Still, Jeff Augen has done some of this on his own on a small scale, using just Excel and TradeStation data, and I would suggest his books and webinars.
Thanks for the responses. Some day I'll crawl back out of theoretical mode. Anyone think there is merit in measuring total price movement on a given day as opposed to focusing on average daily range (high to low)? Big money doesn't profit every day, dealers do...and they have some control/insight into order flow. Perhaps price travels approximately the same distance on a trend day as it does on a rotational day (iow everyday). The only difference is minimal retracement moves on the trend day as opposed to big retracement waves when price isn't going anywhere. I need to think this one over a little more...to think of the best way to measure the segments - maybe, adding up moves that are least 2 points (high to low) in the ES. Then I may be able to say, ok the ES has an ADR of 10 points, but oscillates 40 points on average... BD