How much edge do I need? (and other Edge-Ratio questions)

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
AgeKay
Contributing Member
Contributing Member
Posts: 5
Joined: Thu Jun 14, 2007 4:11 pm

How much edge do I need? (and other Edge-Ratio questions)

Post by AgeKay »

These questions are for c.f. and everybody who has read and implemented or used the Edge Ratio measurement described in Faith's "Way of the Turtle" (this book is excellent, btw).

I have implemented the method for measuring the edge ratio of an entry setup described in the book mentioned above (using an EMA of 14 bars for the ATR) and I have a few questions regarding the application and the meaning of the e-ratios determined by this method. I will number my questions so that it's easier for you to answer them specifically:

1. How much edge should I be looking for? That is, what is the minimum edge ratio that I need to be able to trade profitably. Is 1.2 or 1.6 enough or do I need 2 or 3?
If I understand correctly, to relate the edge ratio to an actual trade it represents the best case for exiting (i.e. exiting exactly at the Maximum Favorable Excursion, which is the the high or low, respectively) and not being stopped out at the Maximum Adverse Excursion. c.f. explains also that the e-ratio will be a lot lower if exits are added. Then there are also transaction costs. Is there a general value that I should substract from the e-ratio to get an e-ratio that is adjusted for likely lower performance in real trading?


2. Can I use low (e.g. 0.7, 0.5) e-ratios to trade the opposite direction of the signal? Can I just divide 1 by the low e-ratio and use that value to compare it then with high e-ratios?

3. I would also like to determine the optimal time frame for the setup (short, medium or long-term). I am averaging the e-ratio at every point with the previous 3 and next 3 e-ratios to smooth the e-ratio and reduce randomness (make it more robust) for this purpose. Does this make sense or is there a better method?

4. I am also calculating the e-ratio seperately for long and short signals and I have noticed in my (preliminary) tests that long and short e-ratios are often inversely correlated. That is, if the long e-ratio is high, the short e-ratio is often low. First I thought that this is due to a bug in my code, but examining my code reversing the signals confirm that my code seems to be ok. So might this be due to the nature of the market or the market being in a general trend during the test window that causes most long trades to be profitable and most short trades to be unprofitable? I am testing 5min bars on 3 months of forex. Could there be other explanations for this phenomon? I have attached a screenshot as an example. The transparent lines show the actual e-ratio values and the non-transparent lines show the smoothed values. The labels on the graph highlight the highest and lowest smoothed e-ratio (e-ratio@bar number | signal ).

Thanks in advance for your replies!

Regards,

Hermann
Attachments
E-Ratio measurement of GBP/USD, 5 min bars on 3 months of recorded prices
E-Ratio measurement of GBP/USD, 5 min bars on 3 months of recorded prices
e-ratio measurement, GBP-USD, 5min.png (32.93 KiB) Viewed 10620 times
explorer
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Thu Jun 07, 2007 8:45 pm
Location: Bay Area, CA

Post by explorer »

AgeKay,

I would like to have an in-depth discussion, but unfortunately I am unable to run the system yet. (No trades)

I have the MACD Reversable PM in the Portfolio Manager
Donchian Entry in the Entry block
Donchian Entry and Entry Edge Tester in the Exit block
Multi Money Manager in the Money Mgr block

If you still interested and can give a little help to start I would be glad to discuss as we proceed with testing.

Thanks your kind help, explorer
AgeKay
Contributing Member
Contributing Member
Posts: 5
Joined: Thu Jun 14, 2007 4:11 pm

Post by AgeKay »

Hi explorer,

thanks for your interest. I don't know if I can help you since I don't own or use the Trading Blox Software.
But I have done some testing with the edge ratio as described in the book and found it rather meaningless without considering stops. What I mean by that, is that the position might first go against you a great deal and then turn around and go even further in your direction producing a positive ratio. Another time, it might go just a little against you and then go the same distance in your direction like above producing an outstanding ratio. But since the maximum adverse excursion might be very different every time you wouldn't be able to profit from each of these moves, because you would have been stopped out in actual trading. Also, large adverse excursions with little favorable excursion would also skew the edge ratio to be a lot worse and not really representative of real trading. For example, in real trading you would have been stopped out pretty early and so it doesn't really matter how far the position went against you.

So in my implementation I build upon the concept of excursion and extended it to take initial stops into account. I still measure maximum adverse excursion and maximum favorable excursion, but I also take a range of stops (e.g. 1 to 10 in steps of 0.5) into accont. I use the maximum adverse excursion to check if I would have been stopped out and I devide the maximum favorable excursion minus a small penalty (right now 1ATR, because I won't be able to call the exact high or low of the move) by the volatility stop multiplier to get the theoretical return as R-Multiple. I also include spreads and commissions to make it a little more realistic. So now it's more like a simple backtest without the trailing stops. Because I assume that I got out at the maximum favorable excursion, this represents the best case scenario. For this reason, I don't use this to pick good signals, e.g. high edge ratio, (which you can't anyway if you ignore stops), but instead as a quick setup filter to weed out the signals that don't perform well even under best case conditions and I still get an idea of the trading duration.
explorer
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Thu Jun 07, 2007 8:45 pm
Location: Bay Area, CA

Post by explorer »

AgeKey,

c.f. and Tim generously helped me out with the installation in another post, so let me run a bunch of tests this weekend then get back to you to continue the discussion.

explorer
explorer
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Thu Jun 07, 2007 8:45 pm
Location: Bay Area, CA

Post by explorer »

AgeKay,

Are you still monitoring this thread? If yes, I am ready to talk, but you asking so many questions, that I don't know where to start. So I start here:
1. How much edge should I be looking for? That is, what is the minimum edge ratio that I need to be able to trade profitably. Is 1.2 or 1.6 enough or do I need 2 or 3?
As far as I am concerned, there is no magic ratio. On one side, you want it as high as possible (to tilt the odds to your side), on the other side as long as you have any edge (after slippage and trading cost) you can crank it up as high as you want with your money management rules as long as you are not violating your max drawdown limits. In other words it is not the size of your edge, it is your money management rules that makes the money. (As long as you have positive expectation/edge.)

As I have limited time, please let discuss 1 question at a time, so we (or at least I) don't get overwhelmed.

My question, if anyone have experience with it: What is the difference between the edge testing and the MAE, MFE testing? Can one use any of these techniques to set stops? What is exactly that one can do with the information obtained from displaying the edge?

explorer
AgeKay
Contributing Member
Contributing Member
Posts: 5
Joined: Thu Jun 14, 2007 4:11 pm

Post by AgeKay »

Yes I am still monitoring the thread, but I have lost interest in this topic for three reasons.

1) I realized through extensive testing that edge as defined in the book mentioned above without considering stops is nonsense. Thus, I developed my own implementation that includes stops.

2) Lack of interest in this thread by forum members.

3) I am looking at something else now (no, I am not searching for another Holy Grail, in case you wondered ;-)).

I'll probably still be reading the posts out of curiosity, but not necessarily comment.
explorer
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Thu Jun 07, 2007 8:45 pm
Location: Bay Area, CA

Post by explorer »

AgeKey,

I am sorry that you lost interest, although by now, you obviously know that the "edge" has nothing to do with stops. It simply will tell you, do you have a statistical adventage at any given time. Of course shorter term system like I think you work with have to have stops, so it was natural that you moved on to technics that deal with more scientific stop placement.

Since there are a post in this forum that greatly discounts the value of MAE based stop placement, I wonder what is your experience.

In other words, do you use MAE to optimize stops, or you just use TB stepping feature to optimize them, or you have your own method for stop optimization?

Thanks, explorer
AgeKay
Contributing Member
Contributing Member
Posts: 5
Joined: Thu Jun 14, 2007 4:11 pm

Post by AgeKay »

Thank you for your interest, Explorer.

yes, you are right I was testing a short-term trading system where inclusion of stops was necessary.

My stops were volatility based (ATR) and I was stepping through the multiplier and I calculated the MAE to determine if I was stopped out. It did definitely work. I could compare the performance visually and see immediately if it was curve-fitted (spikes in the performance measures) or a genuine edge.

No, I don't own or use TradingBlox, I programmed it myself.

I have recently turned discretionary to test out scalping in the us index futures that is based on "tape reading" where I am just looking at price & volume (time & sales), you could call it extreme technical analysis. It's for real, just demand & supply, unlike most technical analysis and indicators. I plan to automate this technique eventually though (very hard right now, need to more expierence and need to formalize it). See MarketDelta.com which visualizes time & sales in a unique way.

This might be off-topic, but going down that path of researching what truly moves price (i.e. demand & supply), I realized that time driven charts (min, hour, day, etc.) are probably the worst way to visualize price movements. If you look at the same price move in an activity driven chart or price driven chart, you'll see that trends are a lot clearer there, and most indicators (and trading systems) would probably have a lot less false signals. Volume driven charts make most sense to me since they are not arbitrary like price charts. See this periodicity tutorial for more info on different chart types. That list is not even complete. I found the most complete list of chart types in the MarketDelta help manual (2.9MB PDF, page 132ff).

P.S.: I am not affiliated with MarketDelta or Investor/RT, just trying to provide some info. The moderator can remove the links if it's not allowed to link to 3rd party sites.
explorer
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Thu Jun 07, 2007 8:45 pm
Location: Bay Area, CA

Post by explorer »

AgeKey,

Thanks for your in depth reply. I will explore the MAE based stop settings to put the technique in bed one way or another.

I agree with your research path. When I traded on shorter term, I only used tick charts.

The MarketDelta based approach is working (I know someone who using it sucessfully) because supply demand based trading supposed to work. My question here, both you and my "friend" using that system for scalping. At this point I am interested in 1-20days swing systems and longer term trend following systems. My question is:

Does this supply/demand MarketDelta based approach works on longer timeframe in your opinion? In other words, does the supply demand based approach has a theorethical or practical time limit? (Just wondering)

As far as all those different charts, you for sure got my attention. That is something that I always wanted to explore but never did. It will go on my to do list.

Thanks, explorer
AgeKay
Contributing Member
Contributing Member
Posts: 5
Joined: Thu Jun 14, 2007 4:11 pm

Post by AgeKay »

I don't think I have enough experience with MarketDelta yet to be able to give you an answer on whether it works on larger time frames (haven't even tried long-term or "swing-term" with MarketDelta since I am only interested in scalping), but the developer of MarketDelta (Trevor Harnett) says in the documentation/videos that most use it for short-term trading and scalping, but I have also read that a few people use it for entries and exits of their swing trades and gives traders in general more confidence when placing trades since you can see the price action.
So yes, it I think it would make sense to use it prior to placing orders, but I don't know if it is helpful in determining the longer term trend of a move. I think you should just ask Trevor directly regarding the time limit to get an authoritive answer.

I would like to know if your trading friend uses MarketDelta exclusively for scalping or if he is also using other tools for confirmation. Thanks!
Post Reply