pete
05-16-2012, 08:37 AM
A Big Thank you to mike, pascal, billy, bob, anindo, you all had influence on this and made it possible. Reading your posts made my day for a lot of days in a row.
Background:
Billyīs and Bobīs postings from 2010 got me on the Weinstein stage model (again), Bobīs backtest on single stocks made me think of expanding this to industry groups. Anindo delivered the input for the stage definitions. Just now I have seen that he also applied it to industry groups in 2009 already.
As I donīt have historical ibd industry group data around I tried the MG and DJUS subsets for short history. Number of Stocks in Stage4 as Indicator for downtrend risk looked interesting from eyeballing:
14246
Fast Historic Crosscheck of previous findings:
8-16-2009 08:26 PM
Anindo Majumdar-VIT
Running my TC 2000 scan shows the market still in a healthy condition.
Stage 1 *(Base building Accumulation) * - 13
Stage 2 *(Uptrending) * * * * * * * * * * * * * - 219
Stage 3 *(Distribution) * * * * * * * * * * * * *- 5
Stage 4 (Downtrending) * * * * * * * * * * * *- 2
As you can see most groups are uptrending which is a good sign for the market. I expect any pullbacks to be shallow and contained. If the market starts topping here I would expect a lot of groups in stage 2 to go into stage 3. Anindo
yyyymmdd st1 st2 st3 st4
20090814 3 229 5 2
20090817 3 207 26 3
Not perfect, but close enough for me, as its no point-in-time data.
So how about a larger timeframe? Expanding the data to start 2000: Warning signs starting July 1999 til 2003, not bad. No sell trigger from Arpil 2003 to April 2004, still interesting.
What I expected, that not worked out that way:
Not much difference to Bobs single Stock backtest, I had expected a different, more precise picture using the industry sets.
Using structures with less or more subindices doesnīt seem to make a difference, I had expected that the datasets with 239 subindices would have made a more precise picture than the ones with 40 or 100.
Using value weighted or equal weighted indices doesnīt seem to make a difference, however this should be studied in more detail.
Even more history:
14248
Again some changes in data subset:
14247
A first test on monthly data from the 1930īs looks pretty similar, but the monthly data needs some more checking.
From the ongoing work I conclude so far:
Number of Industryindices in stage4 above a trigger level could indicate market risk.
Actual we had one sell trigger on 20120509, so from this perspective the correction since April high is not in high risk territory.
For further study:
Number of Industryindices in stage1 above a trigger level could indicate prolonged uptrend to follow.
Pete
Background:
Billyīs and Bobīs postings from 2010 got me on the Weinstein stage model (again), Bobīs backtest on single stocks made me think of expanding this to industry groups. Anindo delivered the input for the stage definitions. Just now I have seen that he also applied it to industry groups in 2009 already.
As I donīt have historical ibd industry group data around I tried the MG and DJUS subsets for short history. Number of Stocks in Stage4 as Indicator for downtrend risk looked interesting from eyeballing:
14246
Fast Historic Crosscheck of previous findings:
8-16-2009 08:26 PM
Anindo Majumdar-VIT
Running my TC 2000 scan shows the market still in a healthy condition.
Stage 1 *(Base building Accumulation) * - 13
Stage 2 *(Uptrending) * * * * * * * * * * * * * - 219
Stage 3 *(Distribution) * * * * * * * * * * * * *- 5
Stage 4 (Downtrending) * * * * * * * * * * * *- 2
As you can see most groups are uptrending which is a good sign for the market. I expect any pullbacks to be shallow and contained. If the market starts topping here I would expect a lot of groups in stage 2 to go into stage 3. Anindo
yyyymmdd st1 st2 st3 st4
20090814 3 229 5 2
20090817 3 207 26 3
Not perfect, but close enough for me, as its no point-in-time data.
So how about a larger timeframe? Expanding the data to start 2000: Warning signs starting July 1999 til 2003, not bad. No sell trigger from Arpil 2003 to April 2004, still interesting.
What I expected, that not worked out that way:
Not much difference to Bobs single Stock backtest, I had expected a different, more precise picture using the industry sets.
Using structures with less or more subindices doesnīt seem to make a difference, I had expected that the datasets with 239 subindices would have made a more precise picture than the ones with 40 or 100.
Using value weighted or equal weighted indices doesnīt seem to make a difference, however this should be studied in more detail.
Even more history:
14248
Again some changes in data subset:
14247
A first test on monthly data from the 1930īs looks pretty similar, but the monthly data needs some more checking.
From the ongoing work I conclude so far:
Number of Industryindices in stage4 above a trigger level could indicate market risk.
Actual we had one sell trigger on 20120509, so from this perspective the correction since April high is not in high risk territory.
For further study:
Number of Industryindices in stage1 above a trigger level could indicate prolonged uptrend to follow.
Pete