Hello board,

I have been recently looking into way of quantifying breadth and got some inspiration from a guy who runs a site called emini watch. This indicator tries to mimic what he has done. It's not as good, but it does have some nice patterns.

For those who don't know, TRIN analyzes advancing issues vs. declining issues. It is a breadth measurement popularized (created?) by Dick Arms. It is an excellent way to gauge market strength without using any price data.

I figured I'd put my work up here and see if anyone gets an inspiration to make it better.

It uses the Nasdaq and NYSE trin figures - -- takes a logarithmic reading of each and inverts it (since TRIN is normally upside down).

These two numbers are averaged and smoothed with a 3 period exponential average.

Then this average is cumulated (added to the previous reading).

This number is then detrended (distance from a moving average taken).

Lastly the linear regression slope of this detrended reading is taken.

***IMPORTANT**** to set up this chart you need 3 data streams.

First is your symbol.

Second is the NYSE TRIN : $trin

Third is the Nasdaq TRIN: $trinq

I am not an expert at reverse engineering, but I like this concept of analyzing breadth. Maybe I have over manipulated the numbers. This is a work in progress, and contributions from the board are appreciated.

You may learn something from this guys site. He seems legit, sells his indicator for a reasonable price. Has good analysis of the market leading me to believe he is a trader first, programmer second. I think his site is eminiwatch.com, but you can google it if you're interested.



WS

[thrunner]

Thanks for sharing your solution. This indicator as written should perhaps be named WS Better Trin Oscillator, more in keeping with it's oscillation and the the original version it is trying to copy (The original has both a summed Trin indicator and an oscillator based on the summed Trin).
Depending on your data feed, you may also get a Trin or TrinQ of zero during an outage or holiday trading. Due to the fact that log(0) is undefined, you will get an error if not checked. The revised code:

inputs:len(7);

//value4=(log(c data2)*-1)*100;    log of zero is undefined, so check for data <=0
//value8=(log(c data3)*-1)*100;

value4=(iff(c data2 >0, log(c data2), 0) *-1)*100;
value8=(iff(c data3 >0, log(c data3), 0) *-1)*100;


value9=xaverage((value4+value8)/2,3);

value10=value9+value10[1];

value11=Detrend(value10, Len);


plot1(linearregslope(value11,len),"btrin");

plot2(0,"");

IMO, this oscillator as written is also one or two periods lagging behind the original oscillator.

[waveslider]

Thanks thrunner.

I agree the oscillator is not the same as the original. Using the linreg idea was mine, I suspected that was what he used.

This is a valid idea to pursue, I think. Though I might just pay the man since he had the idea. Figured it was worth a shot rev. engineering it.

Do you have any other suggestions to get it closer to the original?

[justlurkin]

ws,
Were you able to get this working accurately since the last post?

thanks

[waveslider]

hi justlurkin,
there was a bug. If you are interested, I have a better version. PM me. I'm in hawaii so there might be a tropical lag.

[realgonzo]

Could you post latest version here in ELD ??