Welcome Guest, please sign in to participate in a discussion. Search | Active Topics |

Testing entry reliability in Telechart Rate this Topic:
Previous Topic · Next Topic Watch this topic · Print this topic ·
brexxx
Posted : Monday, May 28, 2007 2:37:46 PM
Registered User
Joined: 7/24/2006
Posts: 131
I would like to test variable entry setup conditions against a random entry system with various hold times (five days in this example). I'm curious to know how some other people might test for this kind of information in Telechart or with the Blocks player. (I do not plan on upgrading to Blocks2.0 until I have attended the seminar, so for now, this is how I test entry reliability). I'm posting this because I would like to know if I have any formula errors, or if there is a more efficient way to do this in Telechart; some of the calc times can be pretty long when plotted as a custom indicator.

Stochastics (35, 10, 11) Test #2 - (Signal 5 Days Ago) [LONG ENTRY ONLY]

PCF Formulas Involved:

1. (STOC35.10.5 > 20 AND STOC35.10.6 > 20) :
Stochastics 35.10 is above 20 five days ago AND six days ago.

2. (STOC35.10.5 > AVG(STOC35.10.5,11)) :
Stochastics 35.10 is above above 11sma trigger five days ago.

3. (STOC35.10.6 <= AVG(STOC35.10.6,11) :
Stochastics 35.10 is below 11 sma trigger six days ago.

4. (C5 > AVGC50.5) :
Close is above 50-bar sma five days ago.

5. (XAVGV38.5 > 1500)
Exp. average volume (38-bar period) is greater than 150,0(oo) five days ago.

Formulas 1-5 are the SETUP conditions. Note that the formula will return a true value (-1) when ALL five of the above conditions were met FIVE DAYS AGO. The goal is to then see how price performed following an entry signal after five days. For my purposes, Entry will be at the open of the bar following the signal bar, and the position will be exited at the close on the fifth holding day. If the close (on the day I receive the signal, which would be five days after the original signal was received) is above the entry price (O4), then call that instance a "True Positive," if price has decreased or has not changed, then call that instance a "False Positive."

6. (C > O4) :
True positive; price has increased in the five days following the signal.

7. (C <= O4) :
False positive; price has decreased or has not changed in the five days following the signal.

In addidtion to plotting true and false positives, I would also like to know the percent change in price from the entry price (O4) to the exit price.

8. ((C-O4)/O4) * 100

TRUE POSITIVES:
(STOC35.10.5 > 20 AND STOC35.10.6 > 20) * (STOC35.10.5 > AVG(STOC35.10.5,11)) * (STOC35.10.6 <= AVG(STOC35.10.6,11)) * (C5 > AVGC50.5) *(XAVGV38.5 > 1500) * (C>O4) * ((C-O4)/O4) * 100

FALSE POSITIVES:
(STOC35.10.5 > 20 AND STOC35.10.6 > 20) * (STOC35.10.5 > AVG(STOC35.10.5,11)) * (STOC35.10.6 <= AVG(STOC35.10.6,11)) * (C5 > AVGC50.5) * (XAVGV38.5 > 1500) * (C<=O4) * ((C-O4)/O4) * 100

Thank you for any input.
Kevin
scottnlena
Posted : Monday, May 28, 2007 3:51:05 PM

Registered User
Joined: 4/18/2005
Posts: 4,090
you might could use "Backscanner" for that.

Other wise create a spread sheet in excell and track them by hand. YOu'll need several hundred at least... or until the numbers stop changing signifficantly. This is what I do.

check my post on "the simple aproach".

Basically crate % true indicators for your conditions and the various excell collumns and rows you need to tracke the various bits of data and and start going through charts. make shure you don't cheat the numbers because you aren't serving your self the best that way. You need a true representation.
brexxx
Posted : Monday, May 28, 2007 4:03:43 PM
Registered User
Joined: 7/24/2006
Posts: 131
Thank you for your input, Scott. Much appreciated. I will check your post.
You and I are on the same page. I have an excel spreadsheet w/ the SP-500 components. The first hundred symbols (alphabetically) are tested (I use four-year periods: 1-1-03 - 12-29-06, for example) against a 5-day hold time, the next hundred, 10-days, etc...
As you probably know, using telechart I have to manually count all of the trues and then all of the falses, which can be time consuming. I plan on getting backscanner (now free with a data subscription to the new Blocks) mid-June, but find myself in limbo until then. Do you know of anyway to automate the counting process in Blocks 1.3?
Thanks again,
Kevin
P.S. The formula should be plotted as a custom indicator as opposed to a % true indicator, just to clarify)
diceman
Posted : Monday, May 28, 2007 6:51:59 PM
Registered User
Joined: 1/28/2005
Posts: 6,049
brexxx

One of the way that you can do counts.

A true false condition will equal 1 when put inside
an ABS() function.

For example:

ABS(C>C1) will equal 1 when C is > than C1.
(zero when not)

So if I want to find out how many days C>C1 in the
last 250 days.

I would create this:

ABS(C>C1)*250

Then smooth it with a simple 250 day moving average.
( with the smoothing average that's in the custom indicator
box , you must use simple)

(smoothing it and multiplying it by 250 recreates the sum)
(we are recreating the sum lost in the averaging process)

This is an easy way to create "counts"
(you must make sure your smooth length
and your multiplication are the same)

(for the last 100 days you would create:
ABS(C>C1)*100 and smooth by 100)

You would click on the indicator and sort by actual value.

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

You can also create "equity graphs" for true false
trading systems by using the cumulative indicator.

This works well for things like 5mav>20 mav= buy.
(5MAV<20MAV=SELL)

You would create the buy trigger the day before.
(as the up condition)
You would make the down condition impossible.
(C>10000)
Then use the 1 day percent change as the value to
add or subtract.

(basically you are adding the percent change when on
a buy and flat-lined when on a sell (cash))

The one problem is this will trade for the "life" of the
stock.

To get what happened in the last year. You would click
on the indictor and sort by net difference over 250 days.
(basically a percent equity line)

(hope that's not all too confusing)


Thanks
diceman














brexxx
Posted : Monday, May 28, 2007 7:14:26 PM
Registered User
Joined: 7/24/2006
Posts: 131
Diceman:
Thank you for your detailed reply. I will look over what you said carefully and post any questions I might have. (The equity graph interests me very much). Thanks again for your time today, and in the past.
Any ideas on adding commission, slippage, and stops in order to find expectancy? I would imagine Backscanner is the best way to do it, but I tend to think it can be done in Telechart. Thoughts?

Kevin
diceman
Posted : Monday, May 28, 2007 7:39:52 PM
Registered User
Joined: 1/28/2005
Posts: 6,049
Once you have a trading return.
(lets say 33% over 1 year as a number)

You can "count" how many times you had a
buy trigger.

Make an assumption on how much equity
you would have devoted to the system.
(Then adjust for commission and slippage)

To guestimate a return.

You can also "count" how long you were on a buy
to determine "exposure".

Realize that this is an approximation of your return
because it is true/false.
(real trading would probably buy at the open, In this
case the 1st days percent change will go into your
account even if you couldn't get in at that level)

Basically you are working with what you have.

(My main goal in looking at this was to find stocks
that were "friendly" to indicators)



Thanks
diceman


Users browsing this topic
Guest-1

Forum Jump
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.