I am working on an automated system using TradeStation (Commodities/Indices). The basics of it are using a random number generator (coin toss) for entry direction and a set stop/target to attain a certain probabilistic outcome based on expected value. What I am having trouble with is that in order to get the probability expectation I want the stop gets hit much more frequently. Recent results have been that profit/loss always tends to hover around zero (back and forth between negative/positive) until a series of consecutive winners/losers shows up. Then it will hover around that level until the next set comes. This is good if those consecutive outcomes are positive, but bad when they are negative. I have looked at adjusting tighter targets, but that screws up the probabilities. However, if this gives me a better win/loss ratio, then I supposed that would make the difference.
I tried tossing the randomized direction, but found the results to be similar, if not worse due to "chasing" on my part. I truly believe in the randomization and prefer to use it, but am conflicted as to why I am having such a difficult time translating my system into automation. I must be missing something or viewing something from the wrong perspective.
Any suggestions? Get rid of automated target exits and exit discretionary whenever a decent profit is there? change the stop/target ratio? Have I just not given it enough time to have enough trades executed before the results can develop in the intended manner?
Would like to here what you all have to say.
Last edited by Jtrader342; March 13th, 2014 at 03:22 PM.
Broker/Data: Multiple broker + Multiple feed (for access to specific exchanges)
Favorite Futures: European Indices, Precious Metals, Energy
Posts: 423 since Dec 2013
Thanks: 204 given,
I'd say to anyone that to be successful they need to try to find and edge and mechanically exploit that edge to become consistent. However personally I'd not use a random number to decide my trade direction as your success and failures are going to be equally random. E.g., I think that you cannot discern as better result from not using a random direction because your discretionary decision on direction is equally random. Not that taking randomness is essentially any better.
I'd not advise that this is a good approach. I'd recommend going back to the drawing board and selecting and more quantitative decision making criteria.
Apologies no offence intended, the best of luck in your trading.