Great Edge Ratio but poor system results

[blinkybill]

I have been testing and entry blox which seems to give one of the best edge ratios I have seen on a fixed exit bar scale yet when I replicate it into a system and use a time based exit matching the optimal edge ratio result I get quite poor results. I have made no other changes to the system. Any thoughts as to why?

[sluggo]

Edge Ratio attempts to measure the "goodness" of an entry signal, separate and apart from the exit. You have an entry that seems to be "good" according to Edge Ratio. When you test that entry with a "Hold for exactly N days, regardless of price action, and then get out" type of exit, the trading results are poor. Maybe the goodness of your entry is being diluted by a substandard, edge-dulling, low quality exit?

Heaven knows there are a large number of well-meaning people who firmly believe, and loudly proclaim, that entries are irrelevant; they preach that all profits accrue from good exits, and there is no such thing as a "good" entry. Maybe a watered down version of their dogma might contain a germ of truth: maybe good exits are not totally irrelevant. Maybe a good entry plus a good exit, is better than a good entry plus a mediocre exit?

You might enjoy this blog post from someone who is using Edge Ratio on a different backtester program: http://theasxgorilla.blogspot.com/2007/ ... pdate.html

[blinkybill]

Thanks Sluggo for your reply. It is always appreciated.

Yes I understand the debate with respect to entries and exits and I am aware that a bad exit protocol can offset a good entry protocol.

However as I pointed out the edge ratio is calculated using fixed exit points so if I am replicating that through my exit protocol shouldnt it give a similar result? Put another way the exit I am using is the same as that for calculating the edge ratio and hence the argument that it is a poor exit offsetting a good entry is nullified...

Or am I missing something...

thanks again

[Roger Rines]

However as I pointed out the edge ratio is calculated using fixed exit points so if I am replicating that through my exit protocol shouldnt it give a similar result?

Maybe not, if the E_Ratio you are using is like the one posted in the Marketplace. Its a maybe not because that calculation uses averages to determine a ratio as an attempted measure of "Goodness".

How broad, and other measures that describe the distribution surrounding the average location is not handled in the posted blox. To get distribution and other more meaningful information you will need to use raw ratios of all the individual elements that make up the average, and probably should consider looking at the raw excursion numbers as well. This more in-depth analysis would allow you to create a bin-distribution, and other stat math test so you can get a more meaningful indication of what your testing indicates.

[blinkybill]

Roger

You are correct and I have begun to dissect the individual data because as you suggest averages can be misleading.

Thanks Jim