StockFetcher Forums · General Discussion · Personal web page<< 1 ... 3 4 5 6 7 ... 9 >>Post Follow-up
671 posts
msg #105075
Ignore miketranz
2/20/2012 5:46:25 PM

Kevin,thank you for sharing your methods.An outstanding credit to SF.Question,the TAA method,when and where are you entering?Also,exiting? Miketranz...

3,854 posts
msg #105076
Ignore Kevin_in_GA
2/20/2012 6:26:14 PM

All trades (I tend to think of them more as reallocations) are made on the last trading day of the month. You reallocate 100% into the asset class with the highest 3 month ROC, and stay there until at the end of some subsequent month a new asset class takes the lead. That's when you sell and buy the new asset class.

You average 3-4 trades per year. Perfect for most 401k accounts, which tend to give you limited fund choices and curtail the frequency of reallocation.

16 posts
msg #105761
Ignore jldelta
4/1/2012 4:47:07 AM

'Morning, Kevin,

Your new blog link doesn't seem to be working this Sunday morning - it links to your 7 Feb post. I'm very interested in finding a practical trading application of Kelly's work, so am looking forward tor reading what you've come up with.


3,854 posts
msg #105762
Ignore Kevin_in_GA
4/1/2012 8:21:45 AM

That's because I haven't finished it yet! I should have it done later today.

What I want to do is to back test it similarly to what I do with the other systems, but that has to be done manually which will take some time. The Kelly allocation is based on the original Kelly f work for discrete bets, plus adaptations from Ernest Chan to fit it into a portfolio approach. Long story short, the ratios are based on ROC / (std dev)^2. I will use a 3 month look back (63 days to be exact) and only reallocate on the last day of each month.

Truth is, I don't know if this works better than the TAA system. I doubt it does, but for many investors the idea of being 100% invested in only 1 asset class is too unsettling, even if it works better for achieving their goals.


3,854 posts
msg #105770
Ignore Kevin_in_GA
4/1/2012 11:11:05 PM

Blog updated. While the Kelly f system did not perform as well as the current TAA system, I think this is because of the strong dependence on low volatility for good scores. I looked at a Sharpe ratio based system as well that does MUCH better, almost matching the TAA system in total return but with a lowered volatility and small max drawdown.

Stats are in the blog. I'll post some simple code for this approach using SF tomorrow.

16 posts
msg #105774
Ignore jldelta
4/2/2012 4:38:25 AM

Very interestng stuff, thank-you, Kevin. It's got my head spinning somewhat, reawkennig studies I looked into quite some years back into Sortino & Sterling as a possible alternatives to Sharpe. I concluded at the time that Sortino was best because it does not penalise upside volatility like Sharpe...

3,854 posts
msg #105776
Ignore Kevin_in_GA
4/2/2012 8:53:47 AM

Agreed, but it is much harder to calculate. My goal here is to share tools with people that can be readily used (and adapted) to meet their specific goals. I looked at Sortino the same way you did, but in the end went with Sharpe because it is easily determined and is more universally accepted as a measure of risk-adjusted returns.

683 posts
msg #105850
Ignore duke56468
4/8/2012 1:31:10 PM

Kevin_in_GA Stats are in the blog. I'll post some simple code for this approach using SF tomorrow.
2,301 posts
Hi Kevin...has this code been posted to SF somewhere yet? Thanks for all your help.

3,854 posts
msg #105851
Ignore Kevin_in_GA
4/8/2012 5:56:03 PM

Here is the code I am currently using. Note that I am using the 60 day ROC of the MA(3), rather than the expected ROC(63,1). I did this to help "smooth out" the impact of exceptionally good/bad days as the Start date for the lookback period. Not really sure if it makes a difference, so feel free to modify as you see fit.


SET{perf1A,ma(3) 60 days ago}
SET{perf1b, MA(3) / perf1a}
SET{perf, perf1b - 1}
SET{var1, perf*100}
SET{rfr, ind(^irx,close)}
SET{perf63, perf - RISKFREERETURN}
SET{VOL63a, STD63 / CMA(MA(3),60)}
SET{vol63, vol63a * 100}

SET{SHARPE1a, perf63 / VOL63a}
set{sharpe, sharpe1a * 0.5}

set{var1a, IND(SPY, var1)}
set{var1b, IND(IWM, var1)}
set{var1c, IND(EFA, var1)}
set{var1d, IND(AGG, var1)}

SET{RANK1A, COUNT(var1 is above var1a,1)}
SET{RANK1B, COUNT(var1 is above var1b,1)}
SET{RANK1C, COUNT(var1 is above var1c,1)}
SET{RANK1D, COUNT(var1 is above var1d,1)}

set{var2a, IND(SPY, sharpe)}
set{var2b, IND(IWM, sharpe)}
set{var2c, IND(EFA, sharpe)}
set{var2d, IND(AGG, sharpe)}

SET{var3a, COUNT(var2a is above 0,1)}
SET{var3b, COUNT(var2b is above 0,1)}
SET{var3c, COUNT(var2c is above 0,1)}
SET{var3d, COUNT(var2d is above 0,1)}

SET{var4a, var2A * var3a}
SET{var4b, var2b * var3b}
SET{var4c, var2c * var3c}
SET{var4d, var2d * var3d}

set{var5a, var4a + var4b}
set{var5b, var4c + var4d}
set{var5c, var5a + var5b}

SET{allocation1, sharpe / var5c}
set{allocation2, allocation1 * count(sharpe above 0,1)}
set{allocation, allocation2 *100}

ADD COLUMN RANK {TAA current rank}
ADD COLUMN var1 {3 month return (%)}
ADD COLUMN vol63 {volatility (%)}
add column sharpe {sharpe ratio}
add column allocation {Sharpe-based allocation (%)}
add column corr(spy,63, close) {correlation to S&P 500}

SORT ON COLUMN 6 ascending

640 posts
msg #105858
Ignore mahkoh
4/9/2012 8:29:44 AM

Nice work! If I understand correct I am to allocate my capital as follows: 36% SPY, 35% IWM and 29% EFA. You state these allocations are to be adjusted on a monthly basis.
Assuming these values change daily, if after 1 week I notice the actual allocation getting out of sync with the suggested allocation from this filter would it be advisable to reallocate?
And is it likely that the filter will allocate 100 % to one security (probably AGG) in case of a market meltdown?

StockFetcher Forums · General Discussion · Personal web page<< 1 ... 3 4 5 6 7 ... 9 >>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