arw functions

Questions about MultiCharts and user contributed studies.
cael
Posts: 89
Joined: 06 Sep 2006
Has thanked: 2 times

arw functions

Postby cael » 14 Feb 2008

Is it possible to control which sub-chart the arrow is plotted?

User avatar
TJ
Posts: 7740
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2221 times

Postby TJ » 14 Feb 2008

It would be great if MC can extend the plot function into plotting in different subcharts.

I can have one indicator script, plotting arrows, lines, texts, etc., on the main chart as well as other subcharts.

This would really put MC above TS.

cael
Posts: 89
Joined: 06 Sep 2006
Has thanked: 2 times

Postby cael » 14 Feb 2008

i take that as no.

but why limits to arrows only, you can also plot other shapes!

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Re: arw functions

Postby Andrew Kirillov » 15 Feb 2008

Is it possible to control which sub-chart the arrow is plotted?
It is possible

Arw_New_self
Displays an object, consisting of an up or a down arrow located at the specified bar and specified price value, on the SubChart containing the study; returns an object-specific ID number, required to modify the object.

Usage
Arw_New_self (BarDate, BarTime, PriceValue, Direction)

Parameters
BarDate - a numerical expression specifying the date of the bar at which the object is to be placed; the date is indicated in the YYYMMdd format, where YYY is the number of years since 1900, MM is the month, and dd is the day of the month

BarTime - a numerical expression specifying the time of the bar at which the object is to be placed; the time is indicated in the 24-hour HHmm format, where 1300 = 1:00 PM

PriceValue - a numerical expression specifying the price value (vertical position, corresponding to a value on the price scale of a chart), where the object is to be placed

Direction - a logical expression specifying the direction of the arrow; True = Down and False = Up

Example
Place, on the SubChart containing the study, an up arrow at the top of a bar if the Open price has increased incrementally over the last three bars:

If Open>Open[1] And Open[1]>Open[2] Then
Value1=Arw_New_self(Date,Time,High,False);

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 15 Feb 2008

It would be great if MC can extend the plot function into plotting in different subcharts.

I can have one indicator script, plotting arrows, lines, texts, etc., on the main chart as well as other subcharts.

This would really put MC above TS.
We allow users to plot trendlines, arrows and text in subcharts.

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 15 Feb 2008

This would really put MC above TS.
The real feature that puts MultiCharts above all existing backtesitng products is True Dynamic Portfolio Backtester which supports EL.
I'm surprised that nobody on the forum is discussing it...

User avatar
TJ
Posts: 7740
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2221 times

Postby TJ » 15 Feb 2008

This would really put MC above TS.
The real feature that puts MultiCharts above all existing backtesitng products is True Dynamic Portfolio Backtester which supports EL.
I'm surprised that nobody on the forum is discussing it...
that shows you how much we need it.

Backtesting was hot in the early 90s, exactly when TS came on the scene. TS capitalized on that wave. But time has changed. people are more sophisticated today than 20yrs ago. People realized that backtesting/curve fitting has limited value. While we still want backtesting, it is not as important a feature as many other things that can help make us money now. (eg integrated playback with simulated order management).

User avatar
TJ
Posts: 7740
Joined: 29 Aug 2006
Location: Global Citizen
Has thanked: 1033 times
Been thanked: 2221 times

Postby TJ » 15 Feb 2008

It would be great if MC can extend the plot function into plotting in different subcharts.

I can have one indicator script, plotting arrows, lines, texts, etc., on the main chart as well as other subcharts.

This would really put MC above TS.
We allow users to plot trendlines, arrows and text in subcharts.
I am suggesting ONE indicator that can simultaneously plot into Subchart#1 and Subchart#2... etc.

I think this extension is easily achievable. It will really distinguish MC from TS.


example of an application:
I have a complex indicator that consists of hundreds of lines of code. It plots two lines, when they cross it generates a signal (appears like a MACD).

if this indicator can plot in multiple-subcharts, then I can simply program it to plot an arrow (or text) on the main chart when a signal is triggered.

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 15 Feb 2008

TJ,
I respectfully disagree. The portfolio back testing solves curve fitting problem, because you can test the same idea on thousands of instruments. It is a way to understand what really works and what doesn't.
I can’t understand how playback will allow you to make money, Even if you will playback charts thousands of times you will be driven by emotions and guess. You will never know when you were wrong and when you were lucky.
You are wrong about 90s. Algo-trading is much more popular now days.
You can see changes in FIX protocol you can monitor growth of algo-trading reported by respected analytical agencies.

Backtesting was hot in the early 90s, exactly when TS came on the scene. TS capitalized on that wave. But time has changed. people are more sophisticated today than 20yrs ago. People realized that backtesting/curve fitting has limited value. While we still want backtesting, it is not as important a feature as many other things that can help make us money now. (eg integrated playback with simulated order management).

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 15 Feb 2008

[

example of an application:
I have a complex indicator that consists of hundreds of lines of code. It plots two lines, when they cross it generates a signal (appears like a MACD).

if this indicator can plot in multiple-subcharts, then I can simply program it to plot an arrow (or text) on the main chart when a signal is triggered.
You can draw an arrow in the subchart containing the indicator and in the subchart containing the data series that the indicator is applied to.

drwar
Posts: 218
Joined: 31 Jul 2005

Postby drwar » 16 Feb 2008

[quote][quote] This would really put MC above TS.[/quote]

The real feature that puts MultiCharts above all existing backtesitng products is True Dynamic Portfolio Backtester which supports EL.
I'm surprised that nobody on the forum is discussing it...[/quote]

Andrew
Frankly, there are several reasons that I wouldn't be doing this Yet, and I say yet because I hope that at some point you have the pieces in place to facilitate this.

1) MC does not yet support setting up and maintaining portfolios. Manual visual conformation and selection is always going to preceed backtestiing.
2) Portfolio backtesting and backtesting in general for futures symbols requires splicing together and creating custom symbols. TS has released a version of this albeit insufficent but MC needs to have this to be able to create enough symbol history.
3) There are too many issues still with signals in general to worry about jumping to portfolios.

I know you proceeding on some or all of these fronts but from my viewpoint until they are addressed I am not going to worry about portfolios.

Just my view

J~

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 18 Feb 2008

Jerry,
Although I do see some logic behind your argumentation, I want to disagree in some aspects:

1) MC does not yet support setting up and maintaining portfolios. Manual visual conformation and selection is always going to preceed backtestiing.
1. MC portfolio Backtester allows you to have several WSP which is equal to watch list.
2. Our back testing concept implies that all symbols are equal for you strategy and it selects what it needs using its logic.


2) Portfolio backtesting and backtesting in general for futures symbols requires splicing together and creating custom symbols. TS has released a version of this albeit insufficent but MC needs to have this to be able to create enough symbol history.
You are talking about subset of the users requirements, but for some reason it looks like generalization. eSignal, IQfeed and other provide continues futures. So it is not a necessary for all to make custom symbols. I agree that some people need custom futures, but it is a minor

3) There are too many issues still with signals in general to worry about jumping to portfolios.

Please list.

Nick
Posts: 496
Joined: 04 Aug 2006
Has thanked: 4 times
Been thanked: 24 times

Postby Nick » 19 Feb 2008

Interesting discussion. I can understand how you are differentiating MC with the portfolio testing stuff. OK maybe I am bias. My leaning is towards more visual tools and core charting features being a discretionary/visual trader. My strong hunch is that this is potentially a larger market. Maybe TS support are being clever going after a different segment (albeit perhaps more of a niche). I still wonder how much interest there will be in portfolio optimisation with the 'average retail trader' (whoever that might be!)

I have to say that traditional 'back testing' type tools aren't of great interest to me (right now at least). However when the tools are top notch I might be inclined to do some research with them.

I am interested in data mining and statistical analysis. This is the same problem approached from the other end. Let me find patterns in the data set so I can design a system to exploit them, rather than let me run this dataset through this system and optimise the parameters. Trouble is with large data sets is computers (and people for that matter) see patterns that are there simply because the data set is large. I am not aware of any trading specific tools for this purpose, Excel and MatLab are probably as good as it gets.

Anyway I look forward to the webinar on this topic, hopefully it will capture my imagination.

Cheers.

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 20 Feb 2008

Trouble is with large data sets is computers (and people for that matter) see patterns that are there simply because the data set is large. I am not aware of any trading specific tools for this purpose, Excel and MatLab are probably as good as it gets.
Nick,
Thanks for the input. I'm not sure I understand the issue you mentioned. Could you specify?

Nick
Posts: 496
Joined: 04 Aug 2006
Has thanked: 4 times
Been thanked: 24 times

Postby Nick » 20 Feb 2008

Its certainly not issues with MC. Large data sets can appear to have non random repeating patterns just by virtue of the fact they are large data sets, i.e. its an illusion. Anyway thats not really the point :D

I think data mining for things that appear to be statistically significant may be quite rewarding. There do not seem to be many (any?) tools to do this. For example conventional wisdom has it that the high or low of the day is 'often' 'close' to the 'early' part of the session. To examine this and try and establish the effect of varying 'often' 'close' and 'early' the easiest way is to pull the data into excel.

Now using traditional system tools you could model an opening range breakout system and optimise various parameters but really that's approaching things quite differently.

Another example. I quite like simple floor pivots for a variety of tasks (yesterdays H+L+C/3). When someone told me that the S&P traded at the Pivot 70% of days - that sounded interesting. To establish if it was true I sucked the data into excel sorted it entered a couple of formulas and did in fact verify that about 70% of the time the ES would trade within 2 ticks of the pivot. It was a PitA working out how to do it and no doubt would be easier next time. Is it valuable to know where price will be tomorrow at some time with 70% reliability? perhaps not on its own but a good place to start puling other statistics.

What about uncovering patterns that you do not know of yet? Or more complex patterns. Data mining coupled with statistical analysis to look for 'edges'. Once you find an 'edge' it should be pretty easy to put together a system to exploit it. People seem to want to 'throw together' a few indicators and have an optimiser find those that are fittest for making money. I guess one way is like doing geological research to find gold the other is digging lots of holes till you have gold in your bucket. I guess both ways you end up with some gold :D

Anyway I should repeat I am not a 'system' trader, maybe because the tools don't exist for the way I am interested in approaching things. In the dim distant past i have pursued the "lets try a bunch of stuff with the help of GA's" method with not that much success. What always surprised me was how stuff that back tested well was lousy forward testing. Evidence that the patterns discovered in backtesting despite being patterns here actually random. Having said that being more experienced now, there are a whole bunch of 'inefficiencies' and semi predictable behaviours in the market that I recognise are exploitable. Maybe that would give better starting spots. Perhaps the new tools in MC will encourage me to have a go at doing that.

Cheers,
Nick.

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 21 Feb 2008

Its certainly not issues with MC. Large data sets can appear to have non random repeating patterns just by virtue of the fact they are large data sets, i.e. its an illusion. Anyway thats not really the point :D
I see some logic in your thesis, but it is not true in all cases. If you have large data sets and great number of trades I think it is not an illusion, but rather a pattern. Let say if you have 1000 symbols with 100 days of data. If you have 4-5 good trades on each symbol with the using the same strategy I think it is a market pattern and the strategy simply exploit it, because 4000 trades is significant number to be statistically valid.



I think data mining for things that appear to be statistically significant may be quite rewarding. There do not seem to be many (any?) tools to do this. For example conventional wisdom has it that the high or low of the day is 'often' 'close' to the 'early' part of the session. To examine this and try and establish the effect of varying 'often' 'close' and 'early' the easiest way is to pull the data into excel.

Now using traditional system tools you could model an opening range breakout system and optimise various parameters but really that's approaching things quite differently.

Another example. I quite like simple floor pivots for a variety of tasks (yesterdays H+L+C/3). When someone told me that the S&P traded at the Pivot 70% of days - that sounded interesting. To establish if it was true I sucked the data into excel sorted it entered a couple of formulas and did in fact verify that about 70% of the time the ES would trade within 2 ticks of the pivot. It was a PitA working out how to do it and no doubt would be easier next time. Is it valuable to know where price will be tomorrow at some time with 70% reliability? perhaps not on its own but a good place to start puling other statistics.

What about uncovering patterns that you do not know of yet? Or more complex patterns. Data mining coupled with statistical analysis to look for 'edges'. Once you find an 'edge' it should be pretty easy to put together a system to exploit it. People seem to want to 'throw together' a few indicators and have an optimiser find those that are fittest for making money. I guess one way is like doing geological research to find gold the other is digging lots of holes till you have gold in your bucket. I guess both ways you end up with some gold :D

Anyway I should repeat I am not a 'system' trader, maybe because the tools don't exist for the way I am interested in approaching things. In the dim distant past i have pursued the "lets try a bunch of stuff with the help of GA's" method with not that much success. What always surprised me was how stuff that back tested well was lousy forward testing. Evidence that the patterns discovered in backtesting despite being patterns here actually random. Having said that being more experienced now, there are a whole bunch of 'inefficiencies' and semi predictable behaviours in the market that I recognise are exploitable. Maybe that would give better starting spots. Perhaps the new tools in MC will encourage me to have a go at doing that.
Nick,
MC portfolio Backtester allows you to find patterns avoiding over-optimization. If you apply some algorithm to 1000 shares and use genetic optimizer to find the values (rules) that makes sense.
For instance if you optimize simple moving average (it is just an example) and see that 19 is a good value, but 20 doesn’t work. It is cleat that it is curve-fitting. But if you see that MA works in the range of 20-40 and doesn’t in the range of 10-20, you’ve got a pattern.
It is not likely to curve-fit if you use 1000 symbols and have 4000 trades. Besides you can test on out-of-sample data to find if it works.

I don’t understand why you tried to use Excel if MC gives all tools to do quality backtesting?

Nick
Posts: 496
Joined: 04 Aug 2006
Has thanked: 4 times
Been thanked: 24 times

Postby Nick » 24 Feb 2008

Hi Andrew I think yu miss my key point. I don't want to backtest (well not a t first). I want to do data mining and statistical analysis. It approaches things from the other end. With backtesting you need an algorithm study or system to start. I want to find interesting patterns in the data that might lend themselves to designing a system around.

Have a quick re-read of my last post and the examples.

I realise that this is likely a version 5 6 or even 7 feature :)

User avatar
Andrew Kirillov
Posts: 1589
Joined: 28 Jul 2005
Has thanked: 2 times
Been thanked: 31 times
Contact:

Postby Andrew Kirillov » 25 Feb 2008

:)


Return to “MultiCharts”