Exit strategy

[naifwonder]

I've been working on my efs code for some time now and have gotten it to the point where more or less every single entry takes you into profit (over 90 percent). Note how I said it takes you into profit and not gives profit. I say this because I feel that my exit strategy is off. Right now, I have preset profit targets which the trading system does reach just about every single time, but it really does suck when the entry is one tick away from your profit target than goes into loss (rare, but common enough) or you watch the contract soar into profits while you have just exited, meaning you missed your opportunity.

So much emphasis is placed on entry that exit is almost ignored. I'm not asking for anyone's strategy, but rather opinions on whether or not exits can be predicted to the same extent that entries can and if so, has anyone created a system that is capable of taking them into atleast 75 percent or so of the potential profit of a trade almost every time?

[demarcog]

I've been programming systems for several years now full time and the only time I ever achieved anything similiar to a results are seeing were when I had problems with the back tester. I hope your results are real, I know how much effort goes into creating a successful strategy.

There are alot of circumstances that can lead to "forward looking" conditions in a back tested strategy and you may want to check your code for the following. I was REALLY disappointed when my systems make a fortune in the backtester but not in real time.

First of all, make sure that you don't specify MARKET-THISBAR when entering a position. It really means market open, and you test will inflate the profit significantly. A quick test would be try MARKET-NEXTBAR and see what the profit looks like.

Also be very careful if you are using multiple interval, that is seriously forward looking.

I have alot of code I can send you if you are interested that manages stops. One technique is to use a fixed trailing stop that will capture a large portion of the profit.

A better way is to use a trailing stop based on the average true range, which will adjust based on the volatility of the market. I also have examples if you are interested.

There is also a way to use the graphs generated by the strategy analyzer to help you place stops based on the data plotted on the graphs. The idea is to place stops at a point to avoid the largest losses but still retain most of the profits. The graphs are the "maximun favorable excursions" and "maximun adverse excursions" I have some documentation I can look for if you are interested.

I hope your results hold up.....feel free to contact me here or offline at [email protected] if you prefer.

[naifwonder]

Thank you for the informative reply. I tested some trailing stop systems and so far haven't come up with anything truly significant yet except for one. The only problem is that it takes you into such great loss at times, you need an account size that is far larger than mine to tolerate such risk. On the brighter side, I did manage to come up with an exit strategy that focuses on creating profitable exit points based on the where the stop loss are placed. Using the entry strategy I have, it generates roughly 16 points every 10 trading days with an approximate success rate of 90 percent. It's not a lot of points given the timeframe, but the rate of accuracy is where I try to give most of my attention to. These are backtesting results however and I am currently testing it with real money.

By the way Demarcog, thank you very much for the offer to send over the code. I might take you up on that offer one day but for now if you could, I would greatly appreciate if you can send me over an example of your Average True Range system. Being that you were so kind as to offer me that information, I'll let you in on a summary of what I'm currently working on. I'm working out several factors that allow me to predict the open, high, low, and close of a candlestick. So far, what I have found gives a good forecast of the high and low of a candle stick with great accuracy. I figure if I can do that, I can create a system to automatically generate probable future chart patterns, and from that deduct other possible candle outcomes resulting in a good overall picture of where the price is headed. It might prove useful combined with what I already have for determing reversal points so I can better time my exits, and entries.

I guess I'm trying to get too specific in my speculation, but I think it might be possible. The idea is vague right now, but I should have a testable system fairly soon. I'll send you an example if I get it done.