Mechanical ES system

[jboydston]

Quote from jack hershey:




The ratio of trend duration is 27:8. You are using 64:6.

With a signal lag of about 2 days plus 64 bars divided by 15 to see the number of days data you average (arith) puts the average about 1/2 the duration behind the current bar.

And the other factor is that you enter on bars that precede signals. Either you didn't write it down correctly or you need to syuggest to us how you take on a trade prior to the signal.

I'm not trying to be sarcastic. I seriously have no idea what you are talking about. Must have been absent from class that day. ;-) I appreciate you patience with me, but can't offer any answers until I understand this further.

regarding the "precede" signals, I tried to clarify that I meant to say "following " not preceding.

Long = open of first bar following price breaking up through 63MA. First bars penetrates the MA, then I buy the open of the next bar.

 

[EliteThink]

Quote from jboydston:



Again, Why do you say this, can you offer a suggestion? Maybe you have some experience you can share with us?

MA systems work at best very short term. They typically win 33% of the time and have a 1.2 profit factor. You could instead of having different inputs for the long and short signal merely have a trend filter.

I know it has been asked several time, do you have a backtest report?

 

[jboydston]

Quote from AAAintheBeltway:

One thing that worries me are the parameters, which are highly suggestive of curve fitting. I would pose two questions: How much data did you backtest this on? Did you test the sensitivity of the MA lengths and entry bar parameters?

Yes, it is optimized.

I started with 6-month back to present. Then took out the top few trades by narrowing down the optimization range. Increased to 1 year back, re-optimized, etc. ANd again at 2 years back.

How would one "test the sensitivity of the MA lengths and entry bar parameters?"

 

[traderkay]

ignore Jack. you'll never understand what he means lol.

 

[jboydston]

Quote from EliteThink:



MA systems work at best very short term. They typically win 33% of the time and have a 1.2 profit factor. You could instead of having different inputs for the long and short signal merely have a trend filter.

I know it has been asked several time, do you have a backtest report?

If anyone wants some documentation, please just ask. I have covered what I think to be my bases. But your objective opinions are why I am here ;-)

I use Tradestation, so give me some time to clean up the report. Default report is huge.

EliteThink I realize that many different strategies may improve this system, including trend filters. However, with all due respect I am not trying to build a perfect system. I want to know if a system developed via backtesting that shows positive results can make it in the real world.

I specifically picked a system that I was not too comfortable with. I expect it to have flaws. regardless, it it profitable on paper. Will it work????

 

[gms]

The ratio of trend duration is 27:8. You are using 64:6.

With a signal lag of about 2 days plus 64 bars divided by 15 to see the number of days data you average (arith) puts the average about 1/2 the duration behind the current bar.

I don't mean to put words in his mouth (not that he'd let anyone do that regardless), but I think Jack is telling you that the lookback period of the MAs does not correspond approriately with the range over which your average trade lasts. Or something like that. Let me know if I'm correct in that assessment so I know if I'm getting closer to understanding Mr. Hershey.

Beyond that, another factor about BTs is that it has to also do with what equity you're BTing. It may look wonderful on one group, and disastrous on another, and has nothing to do with the system, but has all to do with that particular equity, equities or one trade out of a 100, in the history. So, watch out - BT can become BS!

 

[MYDemaray]

Because of the optimization...I fear this system may be too curve-fitted to perform well in the future. I like to see systems that still work well over time with symmetrical Entries/Exits. That's just my humble opinion. Good luck!

 

[jboydston]

Quote from MYDemaray:

Because of the optimization...I fear this system may be too curve-fitted to perform well in the future. I like to see systems that still work well over time with symmetrical Entries/Exits. That's just my humble opinion. Good luck!

Good point. Most of the systems I've come up with short more aggressively than long. Is this because of market trend or maybe markets fall faster than they rise?

 

[trend456]

Quote from jboydston:



Yes, it is optimized.

I started with 6-month back to present. Then took out the top few trades by narrowing down the optimization range. Increased to 1 year back, re-optimized, etc. ANd again at 2 years back.

How would one "test the sensitivity of the MA lengths and entry bar parameters?"

I don't know about optimizing parameters but I think it's pretty hard to come up with a optimized system to work in real life. I'm sure it can be done.

Good Trade.

Trend

 

[AAAintheBeltway]

Quote from jboydston:



Yes, it is optimized.

I started with 6-month back to present. Then took out the top few trades by narrowing down the optimization range. Increased to 1 year back, re-optimized, etc. ANd again at 2 years back.

How would one "test the sensitivity of the MA lengths and entry bar parameters?"

A robust system should function approximately the same if the parameter values are changed within a fairly broad range. For example, a 60 bar MA system would be expected to work about the same if the value used was 50 bars or 70 bars. If a system is highly sensitive to changes in values, it is a sure sign of curve fitting. If I were you, I would test the system on sample chunks of data going back 10 years or so. You said you optimized over the last two years, which was a rather unusual period in the stockmarket. Also, you didn't say if you tested over other periods, but I'm assuming you did not.