Kevin_in_GA
4,599 posts
msg #92134
- Ignore Kevin_in_GA |
5/5/2010 2:52:52 PM
And for 20 out of 21 trading days you could watch TV or golf all day.
|
hmsb4494
81 posts
msg #92135
- Ignore hmsb4494 |
5/5/2010 3:00:58 PM
Now thats a living---500% gains and millions in commissions!!!
|
davesaint86
726 posts
msg #92136
- Ignore davesaint86 |
5/5/2010 3:12:23 PM
Kevin,
That's impressive. What does the 50/50 settings mean?
Thanks,
Dave
|
guymar
113 posts
msg #92137
- Ignore guymar |
5/5/2010 3:24:34 PM
My interpretation of Kevin's model returns 167% since June 2006. This means 10 000 USD become 26 796 USD at end of period (now). I am still optimizing it, and as I do manual backtesting, it takes time. If anyone wants my spreadsheet and knows
|
Kevin_in_GA
4,599 posts
msg #92138
- Ignore Kevin_in_GA |
5/5/2010 3:31:20 PM
50% weighting to the 3 month relative strength, 50% weighting to the 20 day relative strength.
PLEASE NOTE - ETFReplay.com uses a slightly different way to do this than the filter I wrote.
1. They look at the 20 day RS and rank them strongest to weakest. They then give a "1" to the top, a "2" to the next best, and a "3" to the lowest.
2. They do the same with the three month RS.
3. Now they multiply each ranking score by 50% and add them together. This is the same as just adding them and selecting the lowest aggregate score as the top pick. The way I do it is to sum the absolute score, not the ranking. Therefore there are occasional differences in ETF selection.
Their method yields a better overall return than mine, but I can't figure out how to sum the positional ranks for each ETF in a single filter. You can do this by hand, and it only takes about a minute, but it would be nice if I could build it into the filter - so far no luck.
|
guymar
113 posts
msg #92139
- Ignore guymar |
5/5/2010 3:41:09 PM
oops, part of my message got truncated. Basically: I am doing manual backtesting to have precise results, my spreadsheet is available if we can find a way to share it without being spammed.
I would prefer to use weekly rebalancing with 5 and 35 period together with the P¨*VMO oscillatior. This should work, but not for 401k. Give me a couple of days to come back with results. On the other hand: don't forget that we would really need to backtest rebalancing at every day of the month to eliminate "luck" from the equation. Meaning: result on a portfolio rebalancing eacht 2nd, each 3rd and so on to see if results are consitent. Don't forget, even over 4 years, you have only 48 occurences, hardly statistical proof.....
|
Kevin_in_GA
4,599 posts
msg #92142
- Ignore Kevin_in_GA
modified |
5/5/2010 4:14:07 PM
Guymar:
Here I might disagree with you - from my point of view, the objective is to determine if the relative strength method is one that can consistently deliver higher returns than the SPY over a reasonable time frame (say 1 year or longer).
Statistical sampling here can be in weeks, months, or quarters. Since Jan 1 2003 there have been 384 weeks, 88 months, and roughly 22 quarters.
Regardless of the rebalancing period used, this method delivers higher returns (SPY return at 53.2% as of close today):
weekly - 168.9%
monthly - 383.5%
quarterly - 338.2%
|
ciscoslaves
5 posts
msg #92145
- Ignore ciscoslaves
modified |
5/5/2010 4:52:01 PM
Hi, in case anyone is interested in the results, I backtested using Vanguard Sector ETFs and SHY as a proxy for cash to Jan 2007.
I used:
VCR , VDC , VDE , VFH , VHT , VIS , VGT , VAW , VNQ , VOX , VPU
I picked the top three results each month and bought equal positions. If SHY was indicated by the original filter, I went to cash for the entire month.
These are the numbers I came up with, with compounded :
2007 - 12.4% return
2008 - 14.15% loss
2009 - 36% return
2010 - 11.9% return
Overall, from Jan 2007 - now, a 46.2% return. Not bad, but not as good as Kevin's simpler strategy of just using EEM, IWM, and SHY.
|
davesaint86
726 posts
msg #92146
- Ignore davesaint86 |
5/5/2010 4:58:40 PM
I went to the ETFReplay site and ran the relative strength backtest using IWM, EFA, and SHY. SPY was the benchmark. The gains since 2003 was 147%. The weighting I used was 40, 30, 30.
Question - How come the bar chart % gains do not match what is in the big chart? Which one is correct?
Thanks,
Dave
|
Kevin_in_GA
4,599 posts
msg #92147
- Ignore Kevin_in_GA
modified |
5/5/2010 5:04:50 PM
Dave: Look more closely - you need to subrtract the initial 100,000 investment from the chart to get the return.
Also, try 50% on 3 month, 50% 20-day, 0% volatility. Higher returns but also higher volatility.
EDIT: I ran these ETFs at the settings I listed above - return since 2003 is up to 215.5% versus the SPY at 52.3%
Volatility was at 16.1% versus 21.4% for the SPY
If you run it with VWO, the return is now up to 324.2%
If you run it with EEM, the return is a whopping 477.1%, more than double that EFA generated.
This is why EEM was my emerging market ETF of choice ...
|