Hi Forrestang I cannot understand your performance data. Is Expectancy = rate of return for the period ? What is the starting equity to end up with a P&L of $331
Expectancy is basically the amount of money I can 'expect' to win on each trade, assuming the probabilities and average R:R hold. The formula for that is: Expectancy = (Probability of Win * Average Win) â (Probability of Loss * Average Loss) The starting equity is no value for my calculations. What is important is the risk, and for this testing I set the RISK of every trade to $50.00. Whether a trade had a stop of 10 pips or 30 pips the position was sized to meet that risk. The $331 is just how much money the system made over that period given that risk per trade.
Based on some of the posts from ntw, it seems like you would get the distances price has moved from some fixed anchor point. IN this example, I've picked the hourly open as discussed earlier in the thread. These two pictures are just made up, but should show the example. And that is that you would take the distances traveled from the hourly anchor for every bar. Purple being distance traveled. These would then be put into some distribution(the second pic). The question is, would you come up with only a few distinct ranges of distances that price has traveled, which would allow you to have an idea of the distribution, similiar to the SD and LD for the bar formation type trading? If this is the case, how would you then figure out the probabilities?
Forrestang, Thanks for the reply. I have often wondered if NUKE's strategy has merit, and your data has come closest to visualizing that for me. However, I am still missing an important piece of reference. For the period 1/18/11 to 2/11/11, SPY gained 3.0%. How can I use your performance data for comparison?
I guess the only thing I could say to that is that if you look at my charts, most of the trades taken are Longs. Which makes sense as for the testing of the bar formation, I was using the 20 period EMA as a filter. So during that rise we had on the EurUsd which was the instrument I was testing, most of the triggers where buy signals.
flyingdutchmen, Thanks for looking at my code and commenting. I'll be completely reevaluating it soon and will keep these tips in mind. Sorry the late reply. I took a long break from all of this and just focused on basic programming skills for a while. I had some fun studying GWBasic that runs on an interpreter i found online. I think what happened is i wasn't having fun with this and needed a break. Now i'm looking at the code again and it appears at first glance i may have made a significant error in setting it up. In Metatrader, the "start" function handles incoming ticks. So, this is where most of my processing should have been located. However, it appears that the code posted does not properly process data. It looks like the code processes data in the "init" function, which is fine, but the init function processes data only once, when the Advisor is called. So i hope i can find the motivation to completely rework this and come back with a more hopeful report.
Here's a question to anyone trading this method using Tick data. How do you deal with the possibility of multiple solutions, IE, there could be theoretically an infinite amount of combinations of PNR and the subsequent run after that which equal 90%. One idea is to choose the possibility with the greatest overall profit expressed as r/r. Another, is to choose the largest segment. Nuke's example in this thread on page 4 when he introduces the method, the targets are very small compared to the overall range of the bars, and if you go bar by bar you can see most trades were executed within the first 10 minutes of each bar. I have many other questions in mind, but this is my main one I'd like to address since I understand the initial method very well and have already front tested it with success. So, how to solve this challenge and make it into an opportunity... Also, did anyone save the videos from this thread by any chance? I would be very interested in seeing them. Thanks.
I guess I will update to anyone who may be still struggling trading the method. Working with a friend of mine, we are able to take high probability scalps on open using Nuke's method for most high priced stocks or stocks with big ATR (a prerequisite due to comissions). The limitations of the strategy in stocks from a daytrading perspective is leverage. In order to grab the high probability trades as noted by nuke, he was grabbing 5 pips in his example in the euro. When you break down what 5 pips is comparatively on stocks, you see you are grabbing small moves especially since the strategy works better on larger priced stocks. To those trying to figure out the tick method, it would benefit you to rethink what Nuke said about sorting the distances from open. Once you sort ask yourself if you can break down the data any other way instead of days. Now we're working on extending profit targets and using the same concept for mean reversion. From my experience, applying this method to DAILY charts, we are taking most of our trades the first 15 minutes and holding usually no more than a few minutes. Hope that gives some good clues. Also even with the tick method, we are not yet near 90% accuracy. Best we were doing right now is around 70-80% and most of the plays, the R/R is inverse. I have not figured out how to have positive risk to reward while keeping that accuracy.
Wow this thread's still alive. I haven't heard from OP in a long time though. Not sure what happened to him.