StockFetcher Forums · Filter Exchange · TSI question for Kevin_in_GA<< >>Post Follow-up
534 posts
msg #97616
Ignore glgene
11/25/2010 11:37:29 PM


I reread some of your previous posts, and I have a TSI-related question for you (if you don't mind):

Which TSI setting do you currently use on a weekly basis:




I saw that you referenced both in a couple of your previous posts. Not sure which one (if either) you're still using.

There isn't much said about TSI (True Strength Index) on page 107 in the manual.

Thanks in advance for any help you can offer. I like your thoroughness.


4,536 posts
msg #97619
Ignore Kevin_in_GA
11/26/2010 6:30:46 AM

I started using the 3,9 settings, but after a more thorough analysis, I have switched to the 5,5 settings for the weekly TSI

From the other thread, posted on 10/12/2010. This should answer your question.

I finally got around to optimizing The TSI settings.

Using StrataSearch, I was able to have it calculate the weekly TSI data for SPY, IWM, AGG, and EEM going back to 2004. I them imported these into Excel and calculated the equity curves for each of 16 settings - weekly TSI(X,Y,1) where X = 3,5,7,9 and Y = 3,5,7,9). Luckily, the symmetric nature of double EMAs meant that I only had to calculate 10 different combinations (the triangular matrix for you math nerds). It did it for 2004 -2010 as well as 2007-2010. It seemed that most of the gains were actually made during the more volatile past 4 years.

Starting from 1/1/2007 until 8/27/2010, you got the following data:

weekly TSI(3,3,1) = 52.6% return since 2/2/2004, 45.6% return since 1/3/2007. 95 trades made since 2007.
weekly TSI(3,5,1) = 48.6% return since 2/2/2004, 50.8% return since 1/3/2007. 90 trades made since 2007.
weekly TSI(3,7,1) = 53.9% return since 2/2/2004, 50.9% return since 1/3/2007. 79 trades made since 2007.
weekly TSI(3,9,1) = 62.7% return since 2/2/2004, 45.7% return since 1/3/2007. 60 trades made since 2007.

weekly TSI(5,5,1) = 67.0% return since 2/2/2004, 59.4% return since 1/3/2007. 70 trades made since 2007.
weekly TSI(5,7,1) = 57.0% return since 2/2/2004, 35.2% return since 1/3/2007. 52 trades made since 2007.
weekly TSI(5,9,1) = 53.8% return since 2/2/2004, 42.3% return since 1/3/2007. 47 trades made since 2007.

weekly TSI(7,7,1) = 50.9% return since 2/2/2004, 42.3% return since 1/3/2007. 47 trades made since 2007.
weekly TSI(7,9,1) = 40.8% return since 2/2/2004, 30.5% return since 1/3/2007. 44 trades made since 2007.

weekly TSI(9,9,1) = 28.3% return since 2/2/2004, 25.8% return since 1/3/2007. 40 trades made since 2007.

Clearly the Weekly TSI(5,5,1) settings are the best within this set. Looking at the data from 2007, about 10% of the trades suggested were within 2 weeks of the previous trade, so that might be an issue of you are limited to only two reallocations per month.

For comparison:

SPY Buy and Hold: -19% return since 1/3/2007.

Diversified Portfolio (equal weights in all four ETFs): -0.1% return since 1/3/2007. Obviously there is value in having a diversified portfolio, but using strategic asset allocation based on buying into strength (either using RS or the weekly TSI) yields a much higher return at lower overall volatility.

Just looking back to September, you still are making a small profit at a much lower volatility. There will be times where you do not beat the SPY, and times (although limited) where you might lose money - that's all part of investing.

This system is designed to be easy for many casual investors or those managing their 401k accounts to use. Happy to answer your questions here, but it is very self-explanatory.

534 posts
msg #97634
Ignore glgene
11/27/2010 9:38:20 AM

Thanks, Kevin, for your prompt and illustrative response!

So with your weekly TSI(5,5,1) settings and using your example ETFs (SPY,IWM,EEM,AGG), I have 2 questions:

1) At what point would you buy NONE of the above if all of your 4 ETF TSI readings for the week dropped below, say, 50 (or whatever cutoff you imposed)?

2) Using the script below (taken from an earlier posting by you and adjusted to 5,5,1), I don't understand why the TSI "current" reading when run 1 week ago (11/18/2010) doesn't match the TSI reading of 1 week ago when run on 11/26/2010). In other words, why does IWM below show 72.59 and 51.08, when I would think they should be identical, matching numbers?

................................................................... rstoday reading
..................rstoday ....... tsi-1w .............when run 11/18/2010
IWM ......... 51.49 .......... 72.59 ............................ 51.08 .....................

SPY ......... 25.73 .......... 71.66 ............................ 43.75 .....................

EEM ........ -2.65 ........... 66.28 ............................ 23.22 .....................

AGG ........ -37.40 ......... 17.37 ............................ -33.64 ...................

set{TSIRank, weekly TSI(5,5,1)}
add column separator
add column TSIRank {rsToday}
add column TSIRank 1 week ago{TSI -1w}
add column TSIRank 2 weeks ago{TSI -2w}
add column TSIRank 3 weeks ago{TSI -3w}
add column TSIRank 4 weeks ago{TSI -4w}
add column TSIRank 5 weeks ago{TSI -5w}
add column separator

sort column 6 descending
chart-display is weekly
chart-time is 6 Months

23 posts
msg #97636
Ignore sierraseller
11/27/2010 11:35:40 AM

Kevin: Clever approach. Have you considered adding GLD, TLT and maybe USO, and a US$ or a Euro etf to the symbol list to provide a counterbalance when the other ETF's are in a decline?

4,536 posts
msg #97642
Ignore Kevin_in_GA
11/27/2010 7:41:07 PM

Of course - the key here is to be sure that you have a sufficiently uncorrelated set of assets. I posted this a few weeks back in the other thread. Please note that in this filter they are NOT sorted by TSI value, but are positioned such that the added columns create a correlation matrix (this way you can see exactly to what degree each ETF correlates with the others).



and add column corr(agg,100,Close)
and add column corr(gld,100,Close)
and add column corr(iwm,100,Close)
and add column corr(moo,100,Close)
and add column corr(tlt,100,Close)
and add column corr(uup,100,Close)
and add column corr(vnq,100,Close)
and add column corr(vwo,100,Close)

SORT ON COLUMN 1 ascending



I am not including SPY this time, since historically whenever SPY is a buy one can make better returns investing in IWM.

I added TLT and UUP as these are either inversely correlated or uncorrelated to most equities. Added in MOO to make sure that commodities other than metals are included.

534 posts
msg #97644
Ignore glgene
11/27/2010 10:35:29 PM


What is the subject heading for your post a few weeks back ("....I posted this a few weeks back in the other thread.) I missed it.

How do you use the correlation info to make your weekly pick?


4,536 posts
msg #97645
Ignore Kevin_in_GA
11/27/2010 11:31:59 PM

Gene: Here it is ...

And I do not use the correlation data to pick trades, but rather to make sure that the list of ETFs being used is sufficiently diverse.

StockFetcher Forums · Filter Exchange · TSI question for Kevin_in_GA<< >>Post Follow-up

*** Disclaimer *** does not endorse or suggest any of the securities which are returned in any of the searches or filters. They are provided purely for informational and research purposes. does not recommend particular securities., Vestyl Software, L.L.C. and involved content providers shall not be liable for any errors or delays in the content, or for any actions taken based on the content.

Copyright 2016 - Vestyl Software L.L.C.Terms of Service | License | Questions or comments? Contact Us
EOD Data sources: DDFPlus & CSI Data Quotes delayed during active market hours. Delay times are at least 15 mins for NASDAQ, 20 mins for NYSE and Amex. Delayed intraday data provided by DDFPlus