Very interesting market. I did some analysis using Price Action Lab. This is a good tool not only for finding patterns but also it helps to determine what R:R is best. This is what I got: R:R # of patterns # of Long # of Short 5:5 3 0 3 8:8 20 11 9 10:10 22 4 18 20:20 57 3 54 30:30 95 0 95 It is interesting you mentioned 8 points because this gives the best mix of long/short patterns. Actually, R:R = 8:8 gives more long than short signals, 11 long and 9 short. If it is increased to 10:10 the number of long patterns falls sharply and by 20:20 you get only 3 in 57. For R:R 30:30 you get no long patterns and this is basically a short only trading market. I find the results extremely interesting. Actually I may start trading this market. Around R:R = = 8:8 there is a window of opportunity. I used daily continuous data for Wheat from 1998 in the analyis.
If you have any parameters in the system, it's a rule fo thumb to have 30x mor etrades in the backtest compared to parameters. So, if you had a simple moving average crossover system with the length of 2 moving averages as parameters and no stop, you'd need at least 60 trades for the backtest to make any sense.
Intraday Where are you getting over 10 years of historicals for any commodity ? Thanks for any help.....
That's fine. Let's see... EMA - one parameter MACD-H - three parameters 4 * 30 = 120 I have about 400 trades for the first four and a half years of the test. I still have three and a half years left, so that's more than enough.
R:R = Risk:Reward or on this case equivalent to Profit target : stop-loss I don't know how it recognizes patterns but it works well for sure. There are some articles in their website that describe the program operation you can take a look at. I have attached a pic from the search for R:R = 8:8, profit factor > 1, Trades per pattern > 30. It would also be interesting to look for R:R 5:10, 8:16, etc. Now, since all those patterns are variations of one another you would treat them all as one system. Normally, you would have the program search for patterns in a subset of the data set and then forward test to eliminate data mining bias. P is win rate, PF is profit factor, CL is max consecutive losers
Interesting program. I looked at their website yesterday. They don't talk about pricing, so I assume you can't mention how much it costs? It'll be sometime before I can afford such a program and I'll have to look into how well it works.
You got me interested in grains (again). I am running the program with CBOT Corn data. I want to see if the same results apply. I think PAL goes for about 2K per 2 years and IMO a good price for what it can do.