Backtesting

Quote from traderich:

I am reading several posts on here about backtesting set-ups and so forth.

I need to do more research on this, but was curious what others on ET think about this, first hand or otherwise.

I wonder if you had a heart attack and needed an electronic pacemaker do you choose the untested pacemaker or the tested pacemaker?

The hospital might have a new robotic surgeon. Robot is completely automated. Robot is scheduled to install your new pacemaker. I wonder if you choose the tested robot or the untested robot to perform your heart surgery?

Perhaps you wish a second opinion and ask a cardiologist to examine your case. Do you choose a cardiologist that is tested by medical school or an untested cardiologist?

The cardiologist may prescribe drugs. Do you choose the tested drug or the untested drug?

Some people test things to avoid experiencing a crash.

40image.jpg


Others might avoid testing things to experience a crash.

The choice is yours.
 
Signs of spring. racing season starts soon, F1 and nascar. (notice nascar is NOT in caps :D ) I like racing but watching nascar is like watching paint dry.

F1 rules. With that said i thought this was fitting for a backtesting thread.

Jeff Burton said: "you can't simulate a race no matter how much testing you do "
 
Speaking about racing i read back a couple pages and noticed a comment about cheating. i mean jack hershey is so good he claims his way is CHEATING.

Now that is taking being a charlatan to a new level, claiming to be so good that he is cheating. Sorry Jack, could not resist.
 
Quote from bighog:

Speaking about racing i read back a couple pages and noticed a comment about cheating. i mean jack hershey is so good he claims his way is CHEATING.

Now that is taking being a charlatan to a new level, claiming to be so good that he is cheating. Sorry Jack, could not resist.

Fine with me. your comment helps make a point about how it feels to use the SCT.

It is not the "so good" orientation but the viewpoint of seeing what is there for the taking and working to extract it.

It is simply there and I can't affect its flow but on the other hand I can apply cars to it according to the money velocity represented and just take as much of the flow as possible as it comes down the pipe.

I feel that people are monitoring; they see the size and pace of the flow and yet they are not participating except at occasional times when their strategy is "workable".

81 bars a day on 5 min chart. All bars have a given volatility. and the bars are forming trends and patterns, all tradable in sequences which repeat.

It is well worth considering. It is like it is there: what is there to backtest??
 
Jack, have a good weekend. Tell spydertrader i am pulling for the Toyota cars. But truth be known, just like trading, Toyota has a learning curve to go through. Good Day
 
Quote from jack hershey:

81 bars a day on 5 min chart. All bars have a given volatility. and the bars are forming trends and patterns, all tradable in sequences which repeat.

It is well worth considering. It is like it is there: what is there to backtest??

Jack, I feel a little responsible for describing one aspect of your methods as "almost cheating" on another thread. LOL.

Ed Seykota, who I know you admire, strongly advocates backtesting. He also expects drawdowns in his systems (up to 40%, say) and considers day-trading to be inherently unviable due to increased transaction costs relative to potential profits and the risk of over-trading. I think he would also say words to the effect that if you can't code a system (automatic signal generation) then you don't really have a system at all (but you may well have something else that fulfills some other need).

One reason your methods are difficult to backtest is that they rely on being present in the NOW of each bar formation: the signals are often fleeting and contextual (sequences). Has anyone come close to coding / backtesting your methods successfully i.e. Stanford?

I wonder what it would take to convince Ed that your approach has merit - he seems to think the intraday wiggles are merely noise.

Maybe the theory of the large will one day be unified with the theory of the small. :cool:
 
Jack,

I know this is getting off topic here, but could you answer a few more questions

-------------------------------------
Quote from jack hershey:

I attribute periodicity to three independent variables of the market. It turns out that the lowest frequency is price. the period of volume is half that of price (Twice the frequency.

I used this relationship to depict the P,V relationship in boolean algebra and a combination of two if then statements.

----------------------------------------

This is where sometimes the gaussians get confusing, where in one price cycle you have two volume cycles (which would seem to indicate two volume gaussians). So instead of black vol rising, red falling (one gaussian), you can get black rising then falling - and then red rising and red falling (two gaussians, one all black and one all red); finishing the the price cycle. I guess it's not really confusing, it just seems to indicate a different mode.

--------------------------------
snip...
I felt that the cycle of P, V could be extended to another degree of freedom for increased trading sensitivity. So I added another market independnent variable. It may be considered to be sentiment and I express it as A/D where A is accumulation and D is distribution. Both are broadly understood.

It turned out that there is a periodic relationship that is part of the P, V relationship. the frequency of A/D is twice the volume and Four times the price fundamental frequency.

Scoring was the resulting boolean concept that resulted. All trading universes can be sorted by using the three variable frquency concept. thus the cycle of price is divided into 8 parts and the combinations of the variables define the sequence of the cycle. 0 to 7 is the trough and 4 to 3 is the peak. All three variables go through change as expected at these end effects.
crossing the axis of the cycle (acceleration to deceleration ) involves two of the three variables.

The four other dividing lines occur with only one variable changing (A/D)

-----------------------------
While A/D is a broadly understood concept, it was mentioned at some point how different measures can be used, BOP for example. That particular measure might stay in accumulation mode while price and volume cycle several times. That would cause some missing numbers in the scoring sequence going from trough to peak.

So to have an A/D measure that would have twice the frequency of volume, well the only thing I can think of might be the DOM showing a shift in sentiment.

-----------------------------------------------------
snip...
Price is the dominant wave form and the volume Gaussian I use and you see on charts is functioning at twice the frequency of the price trending.

Thus for starters, you always know where you are, what is next and how fast things are changing.

A channel setting up involves point 1, 2, and 3. By this time you see the "trend" and the volume as a harmonic of price. To get to pint three you go through a full gaussian of volume (dom then non dom). The trend continues and the volume oscillates. The A/D is shown as well. This is the first tough to understand paragraph.

------------------------------

Maybe this is where I wonder a bit. (I do understand the usual gaussian - uptrend black rising red falling, and the r2r shifts,etc.) When going point 1 to 2 to 3, that's one price cycle and one volume gaussian, where I would think with strict P/V scoring there should be two volume cycles (in my mind that would look like B/\B R/\R). Unless the two volume cycles would only show up in an even harmonic. That might make sense, or am I way off here?

----------------------------------
For me I am always cognizant of the perice waveform components as a function of their harmonics amplitude and phase angle. There are no missing harmonics usually but there is always an emphasis on the odd or even harmonics as a result of the amplitudes. This is tough for most people to get straight.

Market pace is the most significant aspect of this. The slower the pace the more the odd harmonics dominate. The faster the pace the more the even harmonics dominate. Think of a psotagestamp curve, for example, where trading the nondominant is not usually done with two exceptions.

------------------------------------------

There is a definite change of character in the waves during slow pace. What peaked my interest here was the idea that recognizing when the character changed you're switching harmonics and could know what to expect next.
------------------------------------------

I am not going to be able to easily show you how the head and shoulders is odd harmonic unless I get a scope and a few signal generators. The same for even harmonic double tops. Just inmagine that because I have seen so many patterns on scopes, etc, I can also look at those scopes and see which components of circuits are messing up.
----------------------------------------

If the opportunity presents itself, I'll bring the scope and signal generators :-)

One thing that has stood out at times is the waveforms' peak shift, just looking at the period. For a reference point: http://en.wikipedia.org/wiki/Waveform

Imagine the triangle wave peak shifting to the right to look more like the sawtooth wave - a shift from lateral to uptrend. Then the sawtooth shifts from a right to left translation (example: 5 periods rising, 2 falling, then 2 periods rising 5 falling) Seeing the cycle shift like that would indicate a trend change and most likely have the red to red gaussian shift. It would probably be clearer if I showed it on a chart
-----------------------------

Human psychology closely follows rational natural functions. the theory of large numbers is in effect and the HERD is there all of the time. the HERD follows the smart money and I front run the smart money simply by the "human nature" of price and volume movement. It is like the waveforms of psychology are "continuous functions". It is stats to be sure but for making money, the operating point of the market can only "migrate" and not anything else.
----------------------------

I always appreciate hearing/reading your viewpoint. Thanks for taking the time review this.

- EZ
 
Here is the best way to backtest.

Review your mistakes. Ok, Ok, if you really want to see how many mistakes you (and all of us) make day in and day out just review your chart at the end of the day.

Maybe it is not so much a review of mistakes but rather a review of where you and all of us could heve done better.

Backtesting is overrated. The proof of that is proven over and over when the real bullets fly. I am not a Doctor but i think the emotional level of doing heart surgery on a frog is quite different than doing the same operation on a member of the local high school. Point being, backtesting in the hands of an inexperienced trader is silly, at best.

In backtesting traders are looking for something that works, something that will work in the future (how hard can that quest be?). Truth be known the fallacy of backtesting is it does not highlight what can go wrong. Thats what you are looking for, HOW TO AVOID. Finding what works is out there, has been out there, will stay out there. Use backtesting as a crutch, not as the answer.

Backtesting is saying you never had an opportunity to make noney management decisions as the battle unfolded, that is not trading, that is gambling. As traders we reserve the right to change our minds, gamblers are locked in once the bet has been placed and the "WINDOW" is closed. Which are you?
 
The Hershey Methods are outdated. Anyone who has been studying Quantum physics for any amount of time knows how the theories of single universe interactions are hopelessly incomplete. Market data shows this.

Anyone with half a brain knows that you cannot make simultaneous predictions of conjugate variables which would be required for Jask's theories to hold. This has been known science for more than 100 years. The writings of Bohr and Dirac address these issues specifically.

A more modern approach than Jacks antiquated and pre-newtoninan view of market dynamics is the Chladni vibrational models applied to a multiverse. These achieve much greater accuracy in the third and fourth potentiates. The correlation of angular momentums to volume is much greater as well.

These truths should be apparent to anyone who has studied anything of use in the last 10 years. If we stuck to Jack's science we would not have been able to transcence the 100,000 transistor on a chip mark. A mature understanding of the true interactions at the subatomic level of waves is required for any sort of real correlation to the universal harmonics.

New research into superconducting is also directly applicable.

But we all digress. Why has this thread turned into a Jack Hershey pontification? Was this not supposed to be about back testing?

Why is this man with obviously old science allowed to continue with his drivel? He clearly is using buzzwords to try and confuse those who do not understand the subjects he espouses.

*Max*
 
Very true Ezzy!

All I was looking for was some information about backtesting. Could be articles or books or first hand knowledge. I was asking if backtesting relating to stocks was any different than someone who goes to the casino and writes down the numbers that come out on a roulette wheel. We all know that writing down the past numbers on a roulette wheel is completely useless for helping to find the next number that will come out, but folks still do it.

Peace.
 
Back
Top