Avoiding Curve fitting

As you can see liar and conman, this system generated a sample of 56 trades during the same testing period and 50% were winners.

Now, you really wasted my time with this nonsense, I hate proving to someone he is an idiot, but you deserved it. Here is the list of trades., if you still have any doubts:

Trade # Trade Type Entry Date Entry Time Close Date Close Time Profit MAE Reason For Close
--- Out 5/7/2002 0:00:000 5/8/2002 0:00:000 0.0000 0.0000 Enter long signal
1 Long 5/8/2002 0:00:000 5/14/2002 0:00:000 2.1091 0.6300 Profit target stop
--- Out 5/14/2002 0:00:000 5/16/2002 0:00:000 0.0000 0.0000 Enter long signal
2 Long 5/16/2002 0:00:000 5/29/2002 0:00:000 -2.2778 2.3000 Maximum loss stop
--- Out 5/29/2002 0:00:000 5/31/2002 0:00:000 0.0000 0.0000 Enter long signal
3 Long 5/31/2002 0:00:000 6/7/2002 0:00:000 -3.1300 3.2400 Maximum loss stop
--- Out 6/7/2002 0:00:000 6/11/2002 0:00:000 0.0000 0.0000 Enter long signal
4 Long 6/11/2002 0:00:000 6/14/2002 0:00:000 -2.0020 2.4000 Maximum loss stop
--- Out 6/14/2002 0:00:000 6/18/2002 0:00:000 0.0000 0.0000 Enter long signal
5 Long 6/18/2002 0:00:000 6/20/2002 0:00:000 -1.9817 2.0100 Maximum loss stop
--- Out 6/20/2002 0:00:000 6/24/2002 0:00:000 0.0000 0.0000 Enter long signal
6 Long 6/24/2002 0:00:000 7/3/2002 0:00:000 -1.7815 1.8300 Maximum loss stop
--- Out 7/3/2002 0:00:000 7/8/2002 0:00:000 0.0000 0.0000 Enter long signal
7 Long 7/8/2002 0:00:000 7/10/2002 0:00:000 -1.8340 2.3700 Maximum loss stop
--- Out 7/10/2002 0:00:000 7/12/2002 0:00:000 0.0000 0.0000 Enter long signal
8 Long 7/12/2002 0:00:000 7/22/2002 0:00:000 -1.7675 2.1600 Maximum loss stop
--- Out 7/22/2002 0:00:000 7/24/2002 0:00:000 0.0000 0.0000 Enter long signal
9 Long 7/24/2002 0:00:000 7/25/2002 0:00:000 1.5267 0.0100 Profit target stop
--- Out 7/25/2002 0:00:000 7/29/2002 0:00:000 0.0000 0.0000 Enter long signal
10 Long 7/29/2002 0:00:000 8/5/2002 0:00:000 -1.6289 1.9700 Maximum loss stop
--- Out 8/5/2002 0:00:000 8/7/2002 0:00:000 0.0000 0.0000 Enter long signal
11 Long 8/7/2002 0:00:000 8/16/2002 0:00:000 1.6240 0.9000 Profit target stop
--- Out 8/16/2002 0:00:000 8/20/2002 0:00:000 0.0000 0.0000 Enter long signal
12 Long 8/20/2002 0:00:000 8/28/2002 0:00:000 -1.7633 1.8000 Maximum loss stop
--- Out 8/28/2002 0:00:000 8/30/2002 0:00:000 0.0000 0.0000 Enter long signal
13 Long 8/30/2002 0:00:000 9/5/2002 0:00:000 -1.6534 1.6900 Maximum loss stop
--- Out 9/5/2002 0:00:000 9/9/2002 0:00:000 0.0000 0.0000 Enter long signal
14 Long 9/9/2002 0:00:000 9/11/2002 0:00:000 1.5869 0.0000 Profit target stop
--- Out 9/11/2002 0:00:000 9/13/2002 0:00:000 0.0000 0.0000 Enter long signal
15 Long 9/13/2002 0:00:000 9/23/2002 0:00:000 -1.5820 1.7900 Maximum loss stop
--- Out 9/23/2002 0:00:000 9/25/2002 0:00:000 0.0000 0.0000 Enter long signal
16 Long 9/25/2002 0:00:000 10/7/2002 0:00:000 -1.4945 1.5000 Maximum loss stop
--- Out 10/7/2002 0:00:000 10/9/2002 0:00:000 0.0000 0.0000 Enter long signal
17 Long 10/9/2002 0:00:000 10/11/2002 0:00:000 1.6100 0.0300 Profit target stop
--- Out 10/11/2002 0:00:000 10/15/2002 0:00:000 0.0000 0.0000 Enter long signal
18 Long 10/15/2002 0:00:000 11/1/2002 0:00:000 1.6408 0.9200 Profit target stop
--- Out 11/1/2002 0:00:000 11/5/2002 0:00:000 0.0000 0.0000 Enter long signal
19 Long 11/5/2002 0:00:000 11/21/2002 0:00:000 1.7997 1.5600 Profit target stop
--- Out 11/21/2002 0:00:000 11/25/2002 0:00:000 0.0000 0.0000 Enter long signal
20 Long 11/25/2002 0:00:000 12/6/2002 0:00:000 -1.9500 2.0200 Maximum loss stop
--- Out 12/6/2002 0:00:000 12/10/2002 0:00:000 0.0000 0.0000 Enter long signal
21 Long 12/10/2002 0:00:000 1/10/2003 0:00:000 1.7794 1.1400 Profit target stop
--- Out 1/10/2003 0:00:000 1/14/2003 0:00:000 0.0000 0.0000 Enter long signal
22 Long 1/14/2003 0:00:000 1/21/2003 0:00:000 -1.8837 1.8900 Maximum loss stop
--- Out 1/21/2003 0:00:000 1/23/2003 0:00:000 0.0000 0.0000 Enter long signal
23 Long 1/23/2003 0:00:000 2/10/2003 0:00:000 -1.7780 1.7800 Maximum loss stop
--- Out 2/10/2003 0:00:000 2/12/2003 0:00:000 0.0000 0.0000 Enter long signal
24 Long 2/12/2003 0:00:000 3/14/2003 0:00:000 1.6849 0.7500 Profit target stop
--- Out 3/14/2003 0:00:000 3/18/2003 0:00:000 0.0000 0.0000 Enter long signal
25 Long 3/18/2003 0:00:000 5/5/2003 0:00:000 1.8760 1.5900 Profit target stop
--- Out 5/5/2003 0:00:000 5/7/2003 0:00:000 0.0000 0.0000 Enter long signal
26 Long 5/7/2003 0:00:000 6/4/2003 0:00:000 1.9887 1.0000 Profit target stop
--- Out 6/4/2003 0:00:000 6/6/2003 0:00:000 0.0000 0.0000 Enter long signal
27 Long 6/6/2003 0:00:000 8/29/2003 0:00:000 2.1819 1.9100 Profit target stop
--- Out 8/29/2003 0:00:000 9/3/2003 0:00:000 0.0000 0.0000 Enter long signal
28 Long 9/3/2003 0:00:000 12/29/2003 0:00:000 2.3870 1.7500 Profit target stop
--- Out 12/29/2003 0:00:000 12/31/2003 0:00:000 0.0000 0.0000 Enter long signal
29 Long 12/31/2003 0:00:000 3/22/2004 0:00:000 -2.5634 2.5900 Maximum loss stop
--- Out 3/22/2004 0:00:000 3/24/2004 0:00:000 0.0000 0.0000 Enter long signal
30 Long 3/24/2004 0:00:000 4/2/2004 0:00:000 2.7400 0.0000 Profit target stop
--- Out 4/2/2004 0:00:000 4/6/2004 0:00:000 0.0000 0.0000 Enter long signal
31 Long 4/6/2004 0:00:000 5/10/2004 0:00:000 -2.6012 2.7400 Maximum loss stop
--- Out 5/10/2004 0:00:000 5/12/2004 0:00:000 0.0000 0.0000 Enter long signal
32 Long 5/12/2004 0:00:000 6/29/2004 0:00:000 2.4584 1.0100 Profit target stop
--- Out 6/29/2004 0:00:000 7/1/2004 0:00:000 0.0000 0.0000 Enter long signal
33 Long 7/1/2004 0:00:000 7/16/2004 0:00:000 -2.6362 3.0400 Maximum loss stop
--- Out 7/16/2004 0:00:000 7/20/2004 0:00:000 0.0000 0.0000 Enter long signal
34 Long 7/20/2004 0:00:000 10/29/2004 0:00:000 2.4325 2.4000 Profit target stop
--- Out 10/29/2004 0:00:000 11/2/2004 0:00:000 0.0000 0.0000 Enter long signal
35 Long 11/2/2004 0:00:000 12/1/2004 0:00:000 2.5963 0.0000 Profit target stop
--- Out 12/1/2004 0:00:000 12/3/2004 0:00:000 0.0000 0.0000 Enter long signal
36 Long 12/3/2004 0:00:000 1/20/2005 0:00:000 -2.8217 2.9900 Maximum loss stop
--- Out 1/20/2005 0:00:000 1/24/2005 0:00:000 0.0000 0.0000 Enter long signal
37 Long 1/24/2005 0:00:000 4/29/2005 0:00:000 -2.5900 2.6500 Maximum loss stop
--- Out 4/29/2005 0:00:000 5/3/2005 0:00:000 0.0000 0.0000 Enter long signal
38 Long 5/3/2005 0:00:000 5/19/2005 0:00:000 2.4535 0.0000 Profit target stop
--- Out 5/19/2005 0:00:000 5/23/2005 0:00:000 0.0000 0.0000 Enter long signal
39 Long 5/23/2005 0:00:000 11/8/2005 0:00:000 2.6362 1.1100 Profit target stop
--- Out 11/8/2005 0:00:000 11/10/2005 0:00:000 0.0000 0.0000 Enter long signal
40 Long 11/10/2005 0:00:000 1/11/2006 0:00:000 2.8300 0.0200 Profit target stop
--- Out 1/11/2006 0:00:000 1/13/2006 0:00:000 0.0000 0.0000 Enter long signal
41 Long 1/13/2006 0:00:000 5/15/2006 0:00:000 -3.0058 3.1500 Maximum loss stop
--- Out 5/15/2006 0:00:000 5/17/2006 0:00:000 0.0000 0.0000 Enter long signal
42 Long 5/17/2006 0:00:000 7/12/2006 0:00:000 -2.7776 2.8400 Maximum loss stop
--- Out 7/12/2006 0:00:000 7/14/2006 0:00:000 0.0000 0.0000 Enter long signal
43 Long 7/14/2006 0:00:000 8/17/2006 0:00:000 2.5410 0.7600 Profit target stop
--- Out 8/17/2006 0:00:000 8/21/2006 0:00:000 0.0000 0.0000 Enter long signal
44 Long 8/21/2006 0:00:000 10/4/2006 0:00:000 2.6929 0.5500 Profit target stop
--- Out 10/4/2006 0:00:000 10/6/2006 0:00:000 0.0000 0.0000 Enter long signal
45 Long 10/6/2006 0:00:000 11/15/2006 0:00:000 2.8966 0.2100 Profit target stop
--- Out 11/15/2006 0:00:000 11/17/2006 0:00:000 0.0000 0.0000 Enter long signal
46 Long 11/17/2006 0:00:000 5/21/2007 0:00:000 3.0842 2.0000 Profit target stop
--- Out 5/21/2007 0:00:000 5/23/2007 0:00:000 0.0000 0.0000 Enter long signal
47 Long 5/23/2007 0:00:000 7/19/2007 0:00:000 3.3033 1.2200 Profit target stop
--- Out 7/19/2007 0:00:000 7/23/2007 0:00:000 0.0000 0.0000 Enter long signal
48 Long 7/23/2007 0:00:000 8/10/2007 0:00:000 -3.5147 3.5800 Maximum loss stop
--- Out 8/10/2007 0:00:000 8/14/2007 0:00:000 0.0000 0.0000 Enter long signal
49 Long 8/14/2007 0:00:000 9/25/2007 0:00:000 3.3383 3.3000 Profit target stop
--- Out 9/25/2007 0:00:000 9/27/2007 0:00:000 0.0000 0.0000 Enter long signal
50 Long 9/27/2007 0:00:000 1/7/2008 0:00:000 -3.6134 4.1900 Maximum loss stop
--- Out 1/7/2008 0:00:000 1/9/2008 0:00:000 0.0000 0.0000 Enter long signal
51 Long 1/9/2008 0:00:000 1/22/2008 0:00:000 -4.2600 4.3600 Maximum loss stop
--- Out 1/22/2008 0:00:000 1/24/2008 0:00:000 0.0000 0.0000 Enter long signal
52 Long 1/24/2008 0:00:000 3/17/2008 0:00:000 -3.0933 3.1400 Maximum loss stop
--- Out 3/17/2008 0:00:000 3/19/2008 0:00:000 0.0000 0.0000 Enter long signal
53 Long 3/19/2008 0:00:000 4/18/2008 0:00:000 3.0457 1.4100 Profit target stop
--- Out 4/18/2008 0:00:000 4/22/2008 0:00:000 0.0000 0.0000 Enter long signal
54 Long 4/22/2008 0:00:000 5/15/2008 0:00:000 3.2711 0.4600 Profit target stop
--- Out 5/15/2008 0:00:000 5/19/2008 0:00:000 0.0000 0.0000 Enter long signal
55 Long 5/19/2008 0:00:000 6/24/2008 0:00:000 -3.5035 3.6400 Maximum loss stop
--- Out 6/24/2008 0:00:000 6/26/2008 0:00:000 0.0000 0.0000 Enter long signal
56 Long 6/26/2008 0:00:000 7/15/2008 0:00:000 -3.2690 3.4000 Maximum loss stop
--- Out 7/15/2008 0:00:000 7/17/2008 0:00:000 0.0000 0.0000 Enter long signal
57 Open Long 7/17/2008 0:00:000 8/22/2008 0:00:000 2.1200 1.7000

----------------------------------------------------------------------------
 
Contunued

Now, if N = 3. you get 54 trades, 28 winners and 26 losers, Success rate is 52%.

If N = 4, you get 46 trades, 26 winners and 20 losers, SR = 56%
N = 5, SR = 57%
N = 6, SR= 56%
N = 7, trades = 40, SR = 57%

you get the idea idiot.

At the same time, the Harris system produced 128 trades in total - if I recall correctly. This means that the system did some type of timing to fit all these trades in the same time period given the exit objectives.

This is what analysis means idiot. Your misconceptions, errors, gross misunderstandings of trading and languages are your own. You are hopeless.

I know, I know, you will come back and cry spewing nonsense. You got your answers. Better hide away. Come back with a different name.
 
Quote from bashatrader:

Contunued

Now, if N = 3. you get 54 trades, 28 winners and 26 losers, Success rate is 52%.

If N = 4, you get 46 trades, 26 winners and 20 losers, SR = 56%
N = 5, SR = 57%
N = 6, SR= 56%
N = 7, trades = 40, SR = 57%

you get the idea idiot.

At the same time, the Harris system produced 128 trades in total - if I recall correctly. This means that the system did some type of timing to fit all these trades in the same time period given the exit objectives.

This is what analysis means idiot. Your misconceptions, errors, gross misunderstandings of trading and languages are your own. You are hopeless.

I know, I know, you will come back and cry spewing nonsense. You got your answers. Better hide away. Come back with a different name.
More...

Now tell me how did you feel after pressing the enter key to post?
How long did the feeling last?
What images or thoughts come to mind as you think about these feelings?
 
No, Jack's concept of the price, volume relationship as described in the attached paper, "Catch Up with Tomorrow's Paper Today" is BROKEN. In it, Jack scores stocks using price, volume and A/D and the scores can range from 0 to 7.

I used Spydertrader's code for the scoring and tested buying the "0 to 7 turn" that's shown in the diagram on the last page. I sold each stock 5 days later and got the equity curve below. (5 days worked better than 1,2,3, or 4 days).

I had to exit after 5 days instead of at the "4 to 3 turn" because the model is so broken that the transitions from 4 to 3 are VASTLY outnumbered by the 0 to 7 transitions... enough to make the average trade last for YEARS. Which means that this concept is at odds with reality.

Of course Jack and his followers attacked my methodology and said it should have been applied to a prescreened "universe" of stocks but there's NO MENTION of that in the paper.... see for yourself. Jack also said I should have sold at the "4 to 3 turn" but I already explained why that won't work.

<img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=2014607>
Quote from intradaybill:

I believe jack is correct about volume and A/D
More...
 

Attachments

Question. Is there more than one way to calculate accumulation/distribution, or is this pretty much the standard?

( ( (C - L) - (H - C) ) / (H - L) ) = CLV

The CLV is then multiplied by the corresponding period's volume, and the cumulative total forms the Accumulation/Distribution Line.

http://stockcharts.com/help/doku.php?id=chart_school:technical_indicators:accumulation_distrib

Also see at the bottom of the page: It sometimes difficult to detect subtle changes in volume flows. The rate of change in a downtrend could be slowing, but it may be impossible to detect until the Accumulation/Distribution Line turns up. This drawback has been addressed in the form of the Chaikin Oscillator or Chaikin Money Flow
 
Quote from Stoxtrader:

Question. Is there more than one way to calculate accumulation/distribution, or is this pretty much the standard?

( ( (C - L) - (H - C) ) / (H - L) ) = CLV

The CLV is then multiplied by the corresponding period's volume, and the cumulative total forms the Accumulation/Distribution Line.

More...
I think that's the right basic formula, but the time span people use and how they immediately compound it with other stuff implies to me that there isn't really a standard.

I wasn't initially successful with that indicator so I haven't reviewed what I tried for months, but I had tried ways of filling in the formula. The first looked at the most recent bar only, and the second looked at it over a period. The most recent bar is simple enough since you just plug in the bar's close, high, and low and you got it. Over a period, it would take the latest close, and then the maximum and minimums to get the high and low values.

IIRC the link you gave is something like the top link on Google. If you click down the next 10 results or so you'll probably find a different way each time on how CLV is used, and that's just as an indicator--not even how one would use it in a strategy.
 
Quote from Trader666:

No, Jack's concept of the price, volume relationship as described in the attached paper, "Catch Up with Tomorrow's Paper Today" is BROKEN. In it, Jack scores stocks using price, volume and A/D and the scores can range from 0 to 7.
More...
I don't know about that scoring system, but maybe there's something to the relationship.

I guess what to expect is as a trend continues, the volume will start to dwindle. Now to be so metaphorical as to be corny, that is like the ocean receding before a tsunami. Then there's a reversal. At least that's what I saw on one chart I was obsessing over yesterday.* I was only looking at long positions.

Anyways that's the theory I'd test out, disregarding that score system.

A problem I'm having is I haven't found decent trend indicators yet. I've used MACD for mean reversion, but I haven't been able to put something together that identifies trends particularly well. Yes one can draw them on paper fine and dandy but I'm not doing that 8,000 times a night. :p

Another problem is when do I know that the volume has completely receded? I'd know for 100% certainty when volume spiked back up and all hell broke loose, but that's too late. All I can think there is getting something like a volume oscillator that would raise a flag when it drops to the low side.

Anyways I'll try to scribble something out to handle that and we can play the curve fitting game with it.

* NYSE:ALX 2008-2009. It's an odd chart but consider I'm looking across all equities I can get my mittens on. I focused on it because the stochastics strategies I was trying liked that symbol, yet it would screw it up each time.
 
Mostly everyone's intentions are to help others out. My orientation to this matter is to use my past learnings and applications to iteratively refine where I find a another person is operating.

Some snippets that respond to things that are on the table in no order of importance.

A/D is applied in many ways. All have their intrinsic values. We use it in the DAS as a way to keep track of a stock in its cycle/ (Daily Analysis Sheet, page 23 of BMBW,v 2.2). we currently use a scale of -5 to +5 and the values 5, 4, and 3, make up the score of 1 in the binary raw score).

A caveat on trading stocks is that trading the natural cycle is what is envisioned. This cycle is "seen" on daily charts. there is no way that a person can harden and constain stocks to "do" what he wants them to do. The current Universe Excel sheet illustrates this. We use it each Sunday AM to plan our week's trading. Currently 20K of capital is returning 4K of profits weekly with those who have been doing the weekly meeting for a matter of months. This is not a standard nor is it proving anything. It is a status report of using the PVT method. The carrier tend of stocks in the Universe has been doing about 50% on average in 6 months (A record is kept in Excel of the period's start finish marginal shift for the period in the IAS). (Initial Analysis Sheet, page 16, IBID).

The 8 point scoring cycle (7 through 0 in chronological sequence, trough to peak to trough) logically, goes through all part of the cycle. If it diesn't for anyone, then they need to reason further; into the meaning of "cycle".

Trading quality stocks to make half their long run in position trading is a good way to learn how markets work. As may be seen on the Universe Excel, stocks have three characterisitics for trading: rank, duration of run, and frquency of the cycle. the respective units of measure are %/day, Days and Days. The range used is 3 to 6, 2 to 7, and 3 to 18, repsectively. trading open to open as a poster does will only take part of the run and timing that portion using indicators is very possible and profitable. Only use A of A/D and only use FRV to Peaking volume values.

As a followon for repeating the the statistically insignificant results of the posted backtest, years ago, I posted the one page rule set for the PVT signal generator. (See page30 of PTPT). This page is a chart entitled Unusual Volume Correlation Table). The 30 minute volume values were verified by testing (dealing with curve fitting) using a 400,000 element sample. This thread is important for curve fitting and it may be poissible if there is interest to have that followon performed. There is a year or more of daily Universe data available on ET as was noted in the past.

The benefit of the followon test is that the stock selection will not be arbitrary, the cycle may be used instead of a time out poorly chosen, and a reasonable amount of initial capital and trading duration can be used.

If done, the result will resemble another time period where it WAS done as a foward test using an ATS. The results published were 11.1% average trade gain over an average 6.6 day period. Anyone can chart this using the compound interest formula. Real money was used as another aspect of the forward test. The tester was on leave of absence and resigned at the end of the test. Proceds were used, in part, to design and build a custom Freightliner and a family tavel trailer. A hydraulically control DTM antenna was subsequently installed to make bird watching and trader concurrntly possible.

Bad and good curve fitting may occur. Elsewhere, recently, xburbx was informed by Spyder how to maintain a Universe with the press of a button. In the above description it is a feed for analysis oin the IAS. Most use the EXCEL version for speed and convenience. Making up the batting list on Sunday's leads to a week's trading using 30 minute charts manually. Following a routine is a result of having the tooling. The tooling is where curve fitting may occur (both design and usage). Sorting a Universe yields instruments that all look alike and it is based on percentile based quality. Qualty is remote from curve fitting, in my view. By having the instrument's data define the rank, duration, and frequency from a 6 month sample whch required five sample elements seems reliable in any market. Repeatable and reliable is what is on the table.

Actual trading demands rule following. Are the rules curve fit? By formulating the rules from a 30 year trading experience, they are inductive and not deduced. By using a net range of profits spanning hundreds to thousands and, at a 100,000 share upperlimit, low seven digits, the issue of market capacity incorporated in the sorting, has been handled.

Whether the method is widespread or not will never be an issue. We are fortunate in being able to filter out the less equipped potential traders by having a band of OCD's keep those people away from us. 4 out of 5 person reject this approach in simplisitc polling. I have never been informed by research r just common knowledge of any system not having bad curve fitting failing because of widespread public adoption.

Lastly, the use of indicators is addressed. They all work and more and more precision is attained by two basic techiques: tuning the defaults and signal selection.

The original designer defaults no longer work because of bridging. Good examples of bridging were the SPM and now the Talon idea testing. A lot of bridging is encapsulated in the "ideas". Pring ws a good example of a vendor who came late to the table regarding resetting defaults. The curve fitting aspect of indicators is not often surfaced but the avent of PC's revealed how electronic modifications in the delivery system does change market psychology. Good curve fitting was required to return the original "ideas" back to quality performance.

changing the intention of an indicator is bad curve fitting. MACD is a go/nogo indicator. Care has to be taken when adding new servicability.

My test of indicator performance is "tuning" that works on all fractals. This is a time bound measure of good curve fitting.

As new markets appear after the indicator initial dsing, the curve fitting issue mostly relates to the new market construct. The simplest example would be the change in a NYSE from before to after in the granularity. The shift from large to e mini in commodities futures is another example.

If mark to market becomes a regulatory requirement, new markets will be used to facilitate this. Indicator applications will need to be tuned using defaults and signal selection.

The Cash Cow is an indicator based ATS. The key ingredient is the relationship of market pace and volatility. I felt that this was key in Talon's verifying ideas via testing. Curve fitting is the underlying factor. I was not allowed to participate in that thread by the request of another non participant.

The indicators heyday was before the PC. Now the PC is part and parcel of using indicators. There is an automatic function involved in the use of tuned indicator signals. It is the curve fitting eliminator.

Indicators deal with cycling in markets. Making money exploiting cycles is what indicators do. Volatility is another aspect of market price activity. The constituent that allows a person to not have false signals defeat an ATS or plain indicator is market pace. Talon required me to operate on a 12yo level to deal with this and only use Arithmetic. My funny bone got the best of me, fortunately.

It has been explained how there are several concurrent influences when constructing testing. For indicators it gets tretcherous since they are mathematical designs primarily.

The solution is to deal in various degrees of precision. It becomes a combo of forcing good curve fitting and annihilating bad curve fitting. Joe, jerry and bill have the combo in reasoning for this. At this point in this thread the bad aspects of indicators still hold sway over the good aspects.
 
Wow, this is pretty consistant. Can you reverse action and rerun? (Meaning go short when the signal says go long)

Very interested to see if we can mirror this into profit. I always need more short strategies to look into.


Quote from Trader666:

No, Jack's concept of the price, volume relationship as described in the attached paper, "Catch Up with Tomorrow's Paper Today" is BROKEN. In it, Jack scores stocks using price, volume and A/D and the scores can range from 0 to 7.

I used Spydertrader's code for the scoring and tested buying the "0 to 7 turn" that's shown in the diagram on the last page. I sold each stock 5 days later and got the equity curve below. (5 days worked better than 1,2,3, or 4 days).

I had to exit after 5 days instead of at the "4 to 3 turn" because the model is so broken that the transitions from 4 to 3 are VASTLY outnumbered by the 0 to 7 transitions... enough to make the average trade last for YEARS. Which means that this concept is at odds with reality.

Of course Jack and his followers attacked my methodology and said it should have been applied to a prescreened "universe" of stocks but there's NO MENTION of that in the paper.... see for yourself. Jack also said I should have sold at the "4 to 3 turn" but I already explained why that won't work.

<img src=http://www.elitetrader.com/vb/attachment.php?s=&postid=2014607>
More...
 
You must log in or register to reply here.
Back
Top