What is the message in this graph?

[RedRock]

Here is a graph showing step values for a breakout system. Notice how the std dev. bands contract tightly around 30. Curious what others think the implications of this constriction mean, if anything. I have a few guesses but am curious in other viewpoints. Are there implications as to robustness or lack of going forward, which could be read into this? Does this suggest curve fitting, at the 30 point? Or none of the above... Appreciate your thoughts! rr

[Forum Mgmnt]

We can only completely answer this question if we know what other parameters varied duing the test and what the range of those values was.

The literal meaning of the restriction is that at a 32 day breakout the other parameters did not affect that test.

- Forum Mgmnt

[RedRock]

Hello Forum Mgmnt,

The other parameter stepping was the ATR for a chandelier exit.(5-8 x1) Ill attach that graph as well as the scatter plot.

Other factors:
Very diverse global futures portfolio.
-.4 ATR entry offset
20 ATR days
Volatility MM .5%
MACD portfolio direction filter


Ok. Now I think I 'get it'. Looking at the ATR chandelier plot. The range about the average is the effect various breakout lengths has at that particular value.

So, looking at the atr plot I could 'see' that a value of 7 is probably the most stable in relation to other possibilities? And likewise on the Breakout plot see that 26 - 32ish is the most stable value range.

Can anything be read into this in terms of robustness. I presume that using '29' 7, could be infered to be the most stable of these two interrelated variables. Yes?

I'm going to re-run these tests using your Excursion Ratio for further illumination.

rr

[RedRock]

The following graphs show the E.R.

Interestingly the cagr is very constant at apx 50% +-2% across all results.

Drawdown is 50% but varies from 40-60. (this is a big portfolio in a no frills entry exit test...)

E.R. rises as expected with a longer entry period. The number of trades declines with the higher period also.

Yet, The returns are very consistent. Interesting. (It is accounting for rolls)

I suppose this again demonstrates what a good friend the trend truly is.


Anyway, I do have a grip on the bands on the graphs now which was the point of this whole post.

[Forum Mgmnt]

I first conceived of using bands like these when outlining my ideas for historical testing software in 2001 but it took me a while to implement the graphs in Trading Blox.

One of the interesting thing about these bands is that since starting to use them I find myself focusing on the bottom of the band, i.e. the worst case results as much as the top of the band.

All else being equal, I'd rather trade a parameter setting that is less likely to be bad or at a point where it is less sensitive to drift in the other parameters. I find this thinking useful for natural pairs like moving average lengths (fast, slow), breakout lengths (entry exit), etc. It is my hypothesis that these points lead to a more robust system. I haven't done sufficient analysis to be certain of this yet, however, so if anyone else does please share your results.

My never having thought to do this before just shows how much we traders conform our thinking to the tools we use. If all we have are the tops of these graphs, that is what we focus on. Once we have the bottoms and the middles they become more important, or at least we realize that we've been focusing on a narrow part of the problem.

- Forum Mgmnt