RobtF
16 posts
msg #109215
- Ignore RobtF |
12/8/2012 4:02:21 PM
Kevin I'm new and trying to understand. You're saying that by modifying and optimizing the ADX you're basically replacing the need for his proprietary indicator?
|
Kevin_in_GA
4,599 posts
msg #109216
- Ignore Kevin_in_GA |
12/8/2012 4:10:57 PM
No problem. A WFA requires some means to select which system will be traded forward - I had 32 variations of the basic connors entry evaluated against a equal weighting of the Monte Carlo Annual Return and the Sharpe ratio, with a look back of 500 days, re-evaluated at the end of each month. The system with the best composite performance was traded for the next month, and the process repeated. This way the system results are always based on out-of-sample data, and are dynamically adjusted each month.
The end result was as follows:
STATS FOR 12/31/1999 THROUGH 11/16/2012
NUMBER OF POTENTIAL TRADES: 7,262
PORTFOLIO SIZE OF 5 - INITIAL EQUITY OF $100,000, FIXED TRADE SIZE OF $20,000
NUMBER OF TRADES: 1,549
MONTE CARLO AVERAGE PERCENT PROFITABLE: 6.49%
MONTE CARLO AVERAGE TRADE RETURN: 4.44%
MONTE CARLO AVERAGE ANNUAL RETURN: 106.74%
MONTE CARLO AVERAGE DRAWDOWN: 10.60%
PERCENT IN MARKET: 49.27%
SHARPE RATIO: 1.255
PORTFOLIO SIZE OF 10 - INITIAL EQUITY OF $100,000, FIXED TRADE SIZE OF $10,000
NUMBER OF TRADES: 2,365
MONTE CARLO AVERAGE PERCENT PROFITABLE: 69.58%
MONTE CARLO AVERAGE TRADE RETURN: 4.47%
MONTE CARLO AVERAGE ANNUAL RETURN: 81.97%
MONTE CARLO AVERAGE DRAWDOWN: 6.31%
PERCENT IN MARKET: 37.65%
SHARPE RATIO: 0.9257
Outstanding results given the entire test was on out-of-sample data. Only marginally less than the optimized backtest results, which indicates a very robust system.
|
Kevin_in_GA
4,599 posts
msg #109217
- Ignore Kevin_in_GA |
12/8/2012 5:32:23 PM
Kevin I'm new and trying to understand. You're saying that by modifying and optimizing the ADX you're basically replacing the need for his proprietary indicator?
++++++++++++
Not quite - I just noticed that one could easily optimize his basic entry conditions to produce a trading system that is very good - no proprietary oscillators needed.
|
RobtF
16 posts
msg #109222
- Ignore RobtF |
12/9/2012 9:20:18 AM
Thanks, I downloaded the book. Guess I'll now be getting be getting a call from Connor's people to sell me a service. We'll see whether I can resist or not.
|
cebucher
3 posts
msg #109224
- Ignore cebucher |
12/9/2012 4:19:59 PM
Hello Kevin - I was wondering how you are running all the permutations of your trading strategies. Is it with SF or are you using statasearch as I have seen you reference in prior posts?
|
Kevin_in_GA
4,599 posts
msg #109225
- Ignore Kevin_in_GA |
12/9/2012 6:51:26 PM
Stratasearch. SF's backtesting capabilities pale to nothing next to what SS can do. Make no mistake - I use Stratasearch to develop and validate all of my trading systems. I post them here because I think sharing good ideas is the right thing to do, and most can be easily re-coded into SF syntax.
|
cebucher
3 posts
msg #109226
- Ignore cebucher |
12/9/2012 8:26:37 PM
Kevin - Thanks for your contributions and insight. I have not used SS yet, but was wondering if it would be capable (with reasonable effort) of duplicating the ConnorsRSI including his "duration of up/down trend" and "relative magnitude of price change"? Is SS as easy to learn and use as SF?
|
Kevin_in_GA
4,599 posts
msg #109227
- Ignore Kevin_in_GA |
12/9/2012 9:18:45 PM
Kevin - Thanks for your contributions and insight. I have not used SS yet, but was wondering if it would be capable (with reasonable effort) of duplicating the ConnorsRSI including his "duration of up/down trend" and "relative magnitude of price change"? Is SS as easy to learn and use as SF?
1. You probably could do this is SS, but I'm not sure it is even needed given the results I have posted already. All it will do is limit the number of trades you can take, which ultimately limits any system. Looking for the perfect entry is great unless you only get to trade it a few times a year.
2. Not initially, but I learned both SS and SF code pretty quickly. There is a lot more you can do in SS, and it allows you access to data going back to 1980. On the other hand, it is more "codey" in language - SF uses very intuitive statements in its search strings which is pretty nice.
|
cebucher
3 posts
msg #109228
- Ignore cebucher |
12/9/2012 9:23:50 PM
Thanks, I'll give ss a try.
|
sohailmithani
192 posts
msg #109255
- Ignore sohailmithani |
12/10/2012 2:32:25 PM
Hi Kevin,
Did you do a trade by trade comparison of SS and SF to see if they both pick the same stocks and at same or similar prices and exit too at same or similar prices?
I have recently got AmiBroker. I coded one of my existing filters in SF and compared trade by trade results. Was surprised to see that there were stocks picked by SF but not AB and vice versa. Not sure which one to trust with backtest results.
I even used a very basic code in both SF and AB to say buy stocks close > 5 and when rsi(2) < 10 and exit when rsi(2) > 75 but results were not very close in terms of trade picking.
Anyone has done and such comaprison?
|