Traders' Roundtable

Are EXIT POINTS more important than ENTRY POINTS ?


Eckhardt once said in New York that "If you initiate purely randomly, you do surprisingly well with a good liquidation criterion."


I didn't believe him at first until I did the testings on my own. I found that a fixed duration strategy is able to generate a good looking equity curve.

My back-testings were done in Nikkei and JGB. Just wondering if you guys have similar experience in S&P or other European markets ?

I have never tested for random entries but I am certainly learning that given the same entry points, it is the exits that make the difference.

Take a simple crossover of price over a 100 day moving average and compare it to an identical system which uses a 250 day moving average. And use an envelope of some sort around such averages.

By jiggling the size of the envelope you can reach situation where the entry points for each system are often almost identical for the more successful trades.

But the difference will be found when you look at the exits for a trend trade which actually takes off and keeps on going. During a good strong trend, the longer term average will keep you in there. The shorter term will have you in and out. So there will be more "entries", more trades for the shorter term system. To that extent the entries differ. But on the stonking great trades which win big for the longer term system, the entry date will often coincide almost exactly with an entry for the shorter term system. And it is the exit which determines the profitability of that particular trade.

Philosophically it is difficult to fit pure-reversal systems into an argument about entries vs. exits.

sluggo wrote:Philosophically it is difficult to fit pure-reversal systems into an argument about entries vs. exits.

AFJ, I think he got you there

Except, gentlemen, that I was not talking about a "reversal system".

My wording was no doubt clumsy, but note my reference to putting an "envelope" around the moving average!

I was referring to a sytem such as the ATR breakout. Or take a BB Breakout: look at 2 std bands around an 80 day average and then, say, 1 std bands around a 300 day or 0.75 std bands around a 400 day. Or whatever. The entries can often coincide - but on a winning trade, the exits will assuredly not.

It seems the point is the question. What is “better” for you. A system which takes relatively smaller risks and smaller profits, but which also stays with the “true” trend. OR, The system which has a massive initial risk, and sometimes takes a massive profit. Both have merits and limitations.

PTCM wrote:Are EXIT POINTS more important than ENTRY POINTS ?


Eckhardt once said in New York that "If you initiate purely randomly, you do surprisingly well with a good liquidation criterion."


I didn't believe him at first until I did the testings on my own. I found that a fixed duration strategy is able to generate a good looking equity curve.

My back-testings were done in Nikkei and JGB. Just wondering if you guys have similar experience in S&P or other European markets ?

Not sure I understand. Are you saying flip a coin go long or short and exit regardless after a fixed period of time, say two weeks. Doesent make sense it should be profitable except by luck. Are you saying enter with no trend determination whatsoever?

Deleted due to lack of sleep and brain overload.

A 2 std dev around an 80 day MA often equates to the same risk AT ENTRY as 1 std around a 300 day MA.

After that - we are into the territory of how you govern open trade risk.

RedRock wrote:

PTCM wrote:Are EXIT POINTS more important than ENTRY POINTS ?


Eckhardt once said in New York that "If you initiate purely randomly, you do surprisingly well with a good liquidation criterion."


I didn't believe him at first until I did the testings on my own. I found that a fixed duration strategy is able to generate a good looking equity curve.

My back-testings were done in Nikkei and JGB. Just wondering if you guys have similar experience in S&P or other European markets ?

Not sure I understand. Are you saying flip a coin go long or short and exit regardless after a fixed period of time, say two weeks. Doesent make sense it should be profitable except by luck. Are you saying enter with no trend determination whatsoever?

RedRock, I think that that's what PTCM meant...
Anyways, I wanted to get hands dirty in backtesting and see that's coming out of it. I was testing on Commodities... Random Walk Indicator -Originally Developed by Michael Poulos
Presented in Technical Analysis Of Stocks and Commodities by Michael Poulos (see TASC, January 1992 and September 1992).

RWI -Random Walk Indicator:


* NO STOPS !!!!!!

Short Term RWI uses a Min Period of 2 and a Max Period of 7
Long Term RWI uses a Min Period of 8 and a Max Period of 64

Interpretation:

Long Term RWI of Lows above 1 indicates a sustainable down trend
Long Term RWI of Highs above 1 indicates a sustainable up trend
Short Term RWI of Lows peaking above 1 indicates a price peak
Short Term RWI of Highs peaking above 1 indicates a price trough

Enter long or cover when

Long Term RWI of Highs is above 1 and Short Term RWI of Lows peaks above 1

Enter short or sell when

Long Term RWI of Lows is above 1 and Short Term RWI of Highs peaks above 1.


Results of the test:











Looks like for Random Walk that's not bad

BARLI,


That Equty Curve looks pretty good. what kind of slippage did you incur in the simulation?

I have recently shfited my research to focus on EXIT POINTS instead of ENTRY POINTS. Most of my Ph.Ds friends don't belive this concept. In fact, one guy from Goldman told me this is the wrong way to approach the problem. haha......nice try

Somtimes these PhDs just don't add up to dogshXX. I think that's what happened at LTCM. They tend to overfit historical data, underestimate slippage, and spend 90% of their time to research ENTRY POINTS.

PTCM nor slippage neither commissions were included into this simulation..

Here are results after I included 0.4 slippage

If it's purely random (with a random seed at the beginning of a run), you will see a different result every time. You should probably run your simulation many times and see what happens, e.g., how consistent are your results.

good thought, I'll do more simulations. Post some more thoughts zuijlen!