If the Frequently Answered Questions (FAQ) is not available for your board (the link to the FAQ is shown always below each post whilst reading it), send one to Shrewd'm using the contact link
- Manlobbi
Halls of Shrewd'm / US Policy❤
No. of Recommendations: 22
We now have 15 years of data since the Bear Catchers (BCCs) were discovered. According to Zeelotes, 2009 was the first full year that an investor could use all 3 BCCs to time the market. This post looks into two market timing strategies using the BCCs, and compares them to simply buying and holding an S&P 500 index fund or ETF.
If you don't have time to read the below, my conclusion is that one of the two BCC strategies I looked at has been a failure. The other one can be considered either a success or a failure, depending on how you define your success criteria.
Ok, let's dive in.
First, I'm defining a bear market as a peak-to-trough drawdown of 20% or more in the S&P 500. I'm also defining a
severe correction as a drawdown of 15% or more. Since 1/1/2009, we've experienced 3 bear markets and 3 severe corrections, as shown in the table below.
Let's consider 2 market timing strategies: MT1 and MT2. The buy and sell rules are as follows.
MT1 Market Timing StrategySell SPY only if all 3 BCCs turn bearish
Buy SPY if at least 1 BCC turns bullish
MT2 Market Timing StrategySell SPY if at least 1 BCC turns bearish
Buy SPY only all 3 BCCs turn bullish
Obviously, MT2 is for more skittish investors.
Now, let's define some success criteria. Let's assume we're retired and it's much more important to limit drawdowns than to beat the market. But what exactly does it mean to "catch" a bear market? Does it mean we want our strategy to lose 0% while everyone else is losing 20%? Nah, that's not realistic. Let's say that we consider the strategy to be a success if it limits our drawdown to 10% or less, while buy-and-hold investors are losing 15% or more.
Here's how the MT1 and MT2 strategies have done in the past 15 years.
BCCs Post Discovery (1/1/2009 through present)
MT1 MT2
Buy & hold Sell if all 3 Sell if at least 1
S&P 500 BCCs bearish BCC bearish
Max drawdowns
2009 bear market -27.0% -27.0% -5.6%
2010 correction -15.6% -15.6% -16.0%
2011 correction -18.6% -17.4% -7.6%
2018 correction -19.4% -19.4% -12.5%
2020 bear market -33.5% -33.5% -11.9%
2022 bear market -24.3% -18.3% -2.2%
Performance
CAGR 14.4% 13.9% 7.90%
SAWR 11.9% 11.3% 6.80%
LDDD3 6.0% 6.0% 5.40%
MDD -33.5% -33.5% -16.2%
$100K becomes $769,000 $717,000 $317,000
As you can see, the MT1 strategy failed to limit drawdowns in all 6 cases! During the bear markets and corrections, MT1 dragged investors into the same massive declines as the overall market. So MT1 is objectively a failure when it comes to successfully catching bear markets. I have to admit, I was surprised by this result.
MT2 does much better, but it's not perfect. MT2 did limit drawdowns to below 10% during 3 of the 6 bad periods. But it still exposed investors to a 16% drawdown in 2010, a 13% decline in 2018, and a 12% decline in 2020.
The reduction in volatility came at a cost. Investors using MT2 only had a Safe Withdrawal rate of 6.8% and a CAGR of 7.9%. Whereas their buy-and-hold friends had a SAWR of 12%, a CAGR of 14%, and ended up with more than double the amount of money at the end.
I hope this post helps you understand the tradeoffs when using these BCC strategies.
You can backtest these yourself using the gtr1 link below:
https://gtr1.net/2013/?!!QlpoMTFBWSZTWa!2BNzl4AAnJ...Set BCC > 0 to test MT1 and BCC = 7 to test MT2.
Mechinv
No. of Recommendations: 2
one of the two BCC strategies I looked at has been a failure. The other one can be considered either a success or a failure
That's a great idea but isn't testing those 2 strategies only the first step before coming to hard conclusions? With the next testing them separately to find out which ones worked better to get out and which ones better to get in again? And to do the same for all combinations of them?
(I am not familiar with gtr1 so don't know how to do it.)
No. of Recommendations: 7
isn't testing those 2 strategies only the first step before coming to hard conclusions? The two strategies I tested are the same two that Robbie Geary himself described at this post:
http://www.datahelper.com/mi/search.phtml?nofool=y...Note that Robbie tested from 1926 through Aug. 2019, whereas I tested the post-discovery period from 1/1/2009 to see if the results still hold up.
BCC > 0 means hold SPY as long as
any one of the BCC indicators is bullish. This is the MT1 strategy I tested. Post-discovery, this strategy has
not worked. It has exposed investors to the same massive declines during bear markets as the S&P 500 itself.
BCC = 7 means hold SPY only if
all 3 BCCs are bullish. This is the MT2 strategy I tested. During the 6 severe corrections or bear markets in the post-discovery period, this strategy did limit declines to 10% or less during 3 of those periods. But it exposed investors to declines ranging from 12% to 16% during the other 3 periods. And, after 15 years, you wound up with less than half the amount of money compared to just buying and holding the index (CAGR = 7.9% compared to 14.4% for buy and hold). And you got whipsawed a lot (look at how many times NH-NL flipped just this year). And your Safe Withdrawal Rate declined to 6.8% compared to 12% for the index.
You're certainly welcome to test other values and conditions using the gtr1 link I gave at the top of this thread. Each BCC is represented by a binary digit where 0 is bearish and 1 is bullish. So the BCC state is 3 bits, and the range of values is 0 (all bearish) through 7 (all bullish). With gtr1, you can test BCC using the operators <, >, = and | (or). For example BCC = 1|3|5 means 1 or 3 or 5.
It's important to do periodic post-discovery testing to make sure the conclusions you arrived at 15 years ago are still valid today. Good luck to all.
No. of Recommendations: 10
So MT1 is objectively a failure when it comes to successfully catching bear markets.
A valuable analysis! As well, other backtests (5-10 years ago) have come up with a "NH/NL and any 1 other bearish - out, any 2 bullish back in" as, to my memory, being the most effective permutation (BCC <4 or something out, BCC >=4 or 5) - others could dig through datahelper to confirm.
Yep, the S&P has crushed all. One caveat to the conclusion above: Mechinv's starting date of 1/1/09 is "cherry-picked" to align with a definition of post-discovery. When the BCCs were developed, the backtests included the top of 07 and the bear market of 08 as well as several earlier "long-developing" bear markets. IF IF IF the BCCs had been available, known and used by starting lets say mid-07, an investor would have saved a massive drawdown and that significantly affects this conclusion.
Similar to many of the screens being affected to the wrong side, by many market and technological changes since the late 90s / early Aughts, the "99D" and the SMA Slope (and the 26wk/52wk ratio) were tailored around slower-developing and longer lasting bear markets that had been seen up to that point. There hasn't been one of those since 08-09. The shorter / shallower corrections don't fit, and the '22 bear is really the only one that comes close - but because it had a sharp peak with a fast drawdown (interest rate policy change-driven), a lot of the damage was done before the slower acting 2 fired bearish.
Which is a part of the reason that Zeelotes developed and backtested a 150-day version of the "99D" BC which had somewhat better results by keeping people in the market longer.
FC
No. of Recommendations: 0
Each BCC is represented by a binary digit where 0 is bearish and 1 is bullish. So the BCC state is 3 bits, and the range of values is 0 (all bearish) through 7 (all bullish). With gtr1, you can test BCC using the operators <, >, = and | (or). For example BCC = 1|3|5 means 1 or 3 or 5.
Great! Exactly what I need to know as I love your backtesting BCC idea and am intrigued to test what permutation works best (and how the better ones compare to the index). Thank you.
No. of Recommendations: 10
https://www.shrewdm.com/MB?pid=785305331Results for Growth Trend Timing (GTT), 1/4/2009 thru 3/4/2024, weekly
(MaxDD is the maximum 52-week drawdown)
B&H
CAGR 14.0%
Stdev 17%
MaxDD -51%
Sortino 2.69 (MAR: 3%)
# trades: 1
1st Buy: 1/4/2009
GTT, 43 wk SMA
CAGR 10.9%
Stdev 13%
MaxDD -18%
Sortino 1.93
# trades: 27
Successful OUT: 6 (43%)
Failed OUT: 8
1st Buy: 6/8/09
GTT, 43 wk SMA (sell at 2% below SMA)
CAGR 12.2%
Stdev 13%
MaxDD -18%
Sortino 2.77
# trades: 17
Successful OUT: 6 (67%)
Failed OUT: 3
1st Buy: 6/8/09
I'd need the dates in order to compare with the 6 MT drawdown periods.
Eyeball estimate dates, 1/19/20 - 5/31/20
B&H
CAGR -7.4%
MaxDD -32%
GTT
CAGR -13.6%
MaxDD -12%
GTT (-2% SMA)
CAGR -13.6%
MaxDD -12%
GTT appears to greatly improve the MaxDD.
================================================================
The 30 year period 1/1/95 - 3/4/24:
B&H
CAGR 10.6%
MaxDD -51%
Sortino 0.70
GTT
CAGR 11.3%
MaxDD -18%
Sortino 1.80
GTT (-2% SMA)
CAGR 12.0%
MaxDD -18%
Sortino 2.04
No. of Recommendations: 10
No. of Recommendations: 8
Okay, I am curious about the focus on BCC versus GTT. Is it just that BCC was published earlier?
The GTT paper was published in January 2016, but the "discovery" date is certainly before then, perhaps by one or two years. Nonetheless, it is instructive to look at how it worked out after publication.
GTT certainly seems to hold up post publication. Maximum drawdown is half that of buy&hold, with only 25% lower CAGR.
A longer backtest of the last 30 years shows GTT with much lower MaxDD and actually higher CAGR.
The raison d'être of a timing scheme is to reduce the drawdowns; avoiding the full brunt of a large decline. "John Serrapere's strategy of capture 75% of the S&P’s upside and [only] 50% of its downside."
GTT looks to do this better than BCC, -18% vs. -34%.
--------------------------------
Post publication of the Growth Trend Timing (GTT) paper
Results for Growth Trend Timing (GTT), 1/3/2016 thru 3/4/2024, weekly
(MaxDD is the maximum 52-week drawdown)
B&H
CAGR 13.8%
Stdev 17%
MaxDD -32%
Sortino 1.87 (MAR: 3%)
# trades: 1
GTT, 43 wk SMA
CAGR 10.5%
Stdev 13%
MaxDD -18%
Sortino 1.62
# trades: 19
Successful OUT: 5 (50%)
Failed OUT: 5
GTT, 43 wk SMA (sell at 2% below SMA)
CAGR 11.7%
Stdev 13%
MaxDD -18%
Sortino 2.29
# trades: 13
Successful OUT: 5 (71%)
Failed OUT: 2
================================================================
The 30 year period 1/1/95 - 3/4/24:
B&H
CAGR 10.6%
MaxDD -51%
Sortino 0.70
GTT
CAGR 11.3%
MaxDD -18%
Sortino 1.80
GTT (-2% SMA)
CAGR 12.0%
MaxDD -18%
Sortino 2.04
If I read the GTR1 results correctly, for 19951229 to 20240304
CAGR: 11.789%
MDD: -33.5%
https://gtr1.net/2013/?s19951229e20240304lf-1lp-1h...Does GTR1 have the capability of timing with GTT? I suspect not, since GTR1 predates GTT discovery.
No. of Recommendations: 0
rayvt,
Is the only component of GTT a 43 week SMA?
No. of Recommendations: 9
Is the only component of GTT a 43 week SMA?
43 week SMA for sell/buy timing signals.
The GTT paper uses 12 month SMA, I prefer 10 month (43 weeks).
Sell signal ignored if BOTH these FRED indexes are positive YOY.
INDPRO Industrial Production Index, Index 2012=100, Monthly, Seasonally Adjusted
RRSFS Real Retail and Food Services Sales, Monthly, Seasonally Adjusted
If *either* is down YOY, then you are to obey any SMA sell signal.
In a nutshell, you only consider selling if the SMA is negative when the economy _might_ be in a recession. When the FRED indexes both indicate the the economy is _not_ in a recession, then stay invested (do not sell) regardless of the SMA.
The GTT paper is long and confusing. There are sites that present the meat without the fluff and exhaustive data.
No. of Recommendations: 1
Okay, I am curious about the focus on BCC versus GTT. Is it just that BCC was published earlier?
Its complicated. Another question is why is the focus only on LargeCap instead of the the broad market like Sp1500?
GD_
No. of Recommendations: 6
Another question is why is the focus only on LargeCap instead of the the broad market like Sp1500?
Good question. Perhaps because there is more data.
SP1500 began in May 1995. I can't find historical data for SP1500 on Yahoo. Yahoo does have the ETF (SPTM) from 10/2000)
Yahoo does have S&P500 (^GSPC) from 1950. Dividend yield of the S&P500 is available from other sources, so I can figure out the dividend-reinvested total returns. (Approximately.)
Yahoo has the Russell 1000 (^RUI). I did not bother to look for its dividend yield, so can only figure price-only returns.
Here is comparison of like-to-like for the period 1/7/2002 to 3/15/2024. Timing signal is the index, same as the investment.
B&H, then GTT, then GTT@-2%
S&P500 w/div
8.9%
9.4%
10.3%
S&P500 w/o div
6.8%
7.8%
8.6%
S&P1500 w/o div
7.0%
7.4%
8.2%
Russell 1000 w/o div
7.0%
7.6%
8.1%
It appears that the strategy works similarly for all 3 indexes.
No. of Recommendations: 4
How did it perform with a scaled approach versus All-in/All-out?
BCC Risk-On Risk-Off
7 100% 0%
6 86% 14%
5 71% 29%
4 57% 43%
3 43% 57%
2 29% 71%
1 14% 86%
0 0% 100%
JRB
No. of Recommendations: 10
Dear all.... I think one ought to understand the motivations and underpinning assumptions of each of these rulesets/indicators and then make an informed call.
FULL DISCLOSURE - I follow all 4. And here's the reason why.
But before I forget : "GTT looks to do this better than BCC, -18% vs. -34%." - Ray : This statement might be false.
Zee.... Check the tables #s please - NAHL by construct is the fastest of the lot - so to see an MDD of -34% on it is surprising. IIRC - it was basically negative all the way in GFC 2008 period - so the only option is COVID ..... and by my own tweaked version NAHL exit is 2/24/2020.
Is there a chance you juxtaposed the DBE and NAHL#s - because that would be my next counter - DBE took it in the brunt in COVID - again by construct .... Index was at high 2/19 - which means no DBE possible before mid-July appx - thus that -16% for DBE looks suspect also.
Here's my commentary /understanding of each of these and I'll go slower to faster ie in terms of trigger frequency
(1) DBE-BCC: Dying Bullish - as the name itself suggests motivation is to "Wait and See" whether Bullish euphoria has disappeared. For all practical purposes MDD may not exactly be the best metric. As I understood from Jim's comments - his observation was - typically Start of a Bear provides one last opportunity to exit closer to the last high - and it typically falls in the 5-6 months out period.
Of course, if the index has marched onwards - this stops you from selling out in corrections.
This is basically a TIME ONLY indicator - price check is simply No New High
(2) Slope-BCC : This is a geometrical indicator - its practically 2nd order. Its the 2 week change of Slope ( First derivative) of the 200d MA. (ie appx 50wk SMA equivalent). The change in slope being 2week also helps smoothen it out a bit
This is a PRICE ONLY indicator - but with a 200d SMA and a 2 week change in it ( Slope change <0) its a double smoother.
The exit closeness to the peak is purely defined by how close the 200d MA was to the high at the beginning and the sharpness of the fall. You will have to endure the first leg in the brunt of the face. Its hypothesis is - it will continue.
That of course has been its bane a bit in the post 2009 cycle - ie in some of the deeper corrections - it will signal right close to the bottom
(3) GTT - Macro : This one is a bit different - it actually has an over-arching macro Overlay. And hence is fairly complicated.
Why ? It depends on how FAMILIAR you all are with US Macro series publications. Post-facto data ( ie for eg if you download today from FRED) is literally the FINAL ie Last revision - that typically can happen almost 2-3 quarters ie 6-9 months AFTER ie Look-Ahead bias.
The initial OVERLAY check is simple : Before even looking into the price action - you look at the 2 aforementioned indicators ( or actually your favorite RECESSION indicator) and follow the monthly FRED release update - which is typically towards the 3rd week of the month - and see if they have gone Negative YOY. Else - whatever price action is you hold.
Hypothesis : Unless there are recessionary clouds in the horizon - there cant be deep bears. Your expectation should be ~ -20% drawdowns on this - because there are technical/structural bears ( eg LTCM crisis etc)
But there's an underpinning to it - lets take the last 3 triggers on this
(a) COVID Sell : 1st week of March Open. ie immediate sell - Why? - INDPROD was YoY negative since late 2019 ie Economy was slowing. This was also caught by the Yield Curve inversion - another EXCELLENT macro indicator
(b) 2022 Bear : Late sell - April. Due to RRSFS - but its extremely tight - it went briefly negative in Mar 2022 - and practically remained positive - at least in the periods when the SMA fell below, while when in the interim months when it printed negative the SMA was above.
(c) 2018 Just-Bear: It makes you sell right at the wrong place - towards end of Dec 2018 - as RRSFS goes negative. The SMA went negative with the -2% threshold in Oct '18 itself
(4) NAHL ( With your preferred version of MA smoother) : This is the ONLY Breadth dependent indicator of this lot.
Very simple hypothesis - for a Bull market there needs to be more highs that lows. Its also the reason why this is not necessarily a Bear Indicator - its a "Non-Bull" period indicator. Hence its fast to get out - and of course also relatively fast to get in.
You will have a LOT OF TRADES with this - if its your ONLY one. But as Zee's table shows - in the time since GFC - there's been a decent amount of stocks which have languished - ie breadth really has not been uniform across the market - hence it actually has had the benefit of some really good side-stepping trades
eg. The last exit on the 2021-22 Bear - it was IIRC 1 day within the high ( could well be 1 day BEFORE!)
It operates on the notion - you cant CHEAT MARKET BREADTH!
In summary - these are following posits and timers
(1) DBE : Last breath ( not breadth) of the bulls gasp - Sell 5 months out and run for cover!
(2) Slope: Pure geometric Change in Trend detection - hence double smoothed. You know what you are getting into and that's exactly what you get.
(3) GTT : Macro DOMINANT. No Recession - No fundamental problem in markets. Else Geometry ie 43wkMA Change in Trend
(4) NAHL: Market (ie Daq) Breadth ... Cant hide from breadth. Bull - please show me you have strength. Hence possibly different from the others.
Actually both DBE and NAHL are - DBE is the last gasp check. NAHL is a Non-Bull check.
Hope this made sense and helps!
Best
No. of Recommendations: 0
JRB : That is a bit fallacious - due to how GTT works IIRC
Those are discrete steps /stages - ie it doesn't step +1 necessarily.
Hope that made sense
No. of Recommendations: 2
anchak wrote: Check the tables #s please - NAHL by construct is the fastest of the lot - so to see an MDD of -34% on it is surprising.
I used the links from Robbie's post linked to earlier in this thread and simply changed the date range, nothing else. Whether it is right or not I did not go through the laborious effort to confirm one way or the other. But I'd be surprised if it was incorrect. Though since he posted those links long ago it is possible his method on GTR1 has changed regarding Bear Catchers.
No. of Recommendations: 4
In the 67 years since 1957, BCC=0 timing had 12 years with excess returns above 2%, and 11 years with excess returns less than -2%. BCC did well when there was a recession lasting more than 9 months. The top 10 excess return years line up with recessions:
excess recession recession length
Year 500ew BCC0 return .... start end months
1962 -12 -4 8 Apr-1960 Feb-1961 10
1969 -18 -10 8 Dec-1969 Nov-1970 11
1970 5 20 15
1974 -23 12 35 Nov-1973 Mar-1975 16
1981 5 9 5 Jul-1981 Nov-1982 16
1982 30 39 9
2001 0 6 6 Mar-2001 Nov-2001 8
2002 -18 3 20
2008 -39 -8 32 Dec-2007 Jun-2009 18
2009 46 52 5
Aug-1957 Apr-1958 8
Jan-1980 Jul-1980 6
Jul-1990 Mar-1991 8
Feb-2020 Apr-2020 2
BCC0 is {SP500EW_sell_if_All_Bearish}
https://gtr1.net/2013/?~SP500EqualWeight:h63::sp50...https://gtr1.net/2013/?~SP500EW_sell_if_All_Bearis...
No. of Recommendations: 4
trying a repost of the table using PRE tags:
excess recession recession length
Year 500ew BCC0 return .... start end months
1962 -12 -4 8 Apr-1960 Feb-1961 10
1969 -18 -10 8 Dec-1969 Nov-1970 11
1970 5 20 15
1974 -23 12 35 Nov-1973 Mar-1975 16
1981 5 9 5 Jul-1981 Nov-1982 16
1982 30 39 9
2001 0 6 6 Mar-2001 Nov-2001 8
2002 -18 3 20
2008 -39 -8 32 Dec-2007 Jun-2009 18
2009 46 52 5
Aug-1957 Apr-1958 8
Jan-1980 Jul-1980 6
Jul-1990 Mar-1991 8
Feb-2020 Apr-2020 2
No. of Recommendations: 3
I checked it ..... that draw ie -30% is because of continued under-performance of the strategy 2013-2016 ( ie almost continuous small losses due to whipsaws). Not because it didnt exit in the bears
No. of Recommendations: 12
How did it perform with a scaled approach versus All-in/All-out?
BCC Risk-On Risk-Off
7 100% 0%
6 86% 14%
5 71% 29%
4 57% 43%
3 43% 57%
2 29% 71%
1 14% 86%
0 0% 100%
I think BCC is being misconstrued here as representing a scale.
A value of 0 means all signals negative.
A value of 7 means all signals positive.
But in between there is no scale.
The values 1,2, and 4 each mean that one of the indicators is positive and two are negative.
The values 3,5, and 6 each mean that two of the indicators are positive and one is negative.
So values of 1,2, and 4 are effectively equal, unless you decide to rank the three indicators relative to one another.
The same goes for values of 3,5, and 6.
Elan
No. of Recommendations: 9
I believe this is how that works:
BCC BCI BCII BCIII
(Bear Catchers Combined) (SMA Slope Change) (NH/NL Ratio) (Dying Bullish Euphoria)
0 Bear Bear Bear
1 Bull Bear Bear
2 Bear Bull Bear
3 Bull Bull Bear
4 Bear Bear Bull
5 Bull Bear Bull
6 Bear Bull Bull
7 Bull Bull Bull
No. of Recommendations: 4
It appears that the strategy works similarly for all 3 indexes.As expected for those 3 cap-weighted indexes and understand the data problem.
If I recall GTR1 has a SP1500EW index available which would be closer to representing the MI screens.
And IMO be a better Benchmark.
Example of existing ETF choices available:
INVESCO ISHARES INVESCO SPDR SPDR
RUSSELL MSCI S&P S&P S&P
1000 USA` 500 1500 500
EQ WT EQ WT EQ WT COMPOSITE ETF TRUST
ETF ETF ETF ETF ETF
Symbol EQAL EUSA RSP SPTM SPY
Market price $44.60 $88.57 $164.19 $62.80 $512.86
Mar-18-2024 Mar-18-2024 Mar-18-2024 Mar-18-2024 Mar-18-2024
Market total return (1 year) 8.09% 16.57% 13.01% 28.69% 30.18%
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Market total return (3 years) 4.36% 6.76% 8.43% 11.35% 11.78%
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Market total return (5 years) 8.79% 10.67% 11.35% 14.24% 14.63%
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Market total return (10 years) -- 9.98% 10.23% 12.22% 12.58%
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Premium / discount (1 year avg.) -0.01% 0.02% 0.01% 0.01% 0.01%
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Beta (month-end 3 years) 0.98 1.02 0.98 1 1
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Standard deviation (month-end 3 years) 18.6 18.72 18.29 17.7 17.63
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Sharpe ratio (month-end 3 years) 0.17 0.3 0.38 0.54 0.57
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
Alpha (month-end 3 years) -6.62 -4.73 -2.89 -0.49 -0.07
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
R2 (month-end 3 years) 85.96 92.71 89.66 99.83 100
Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024 Feb-29-2024
GD_
No. of Recommendations: 9
Likely a tempest in a teapot.
Roger Nusbaum:
"I don't really think it matters which trigger is used as no single trigger can be the best for all times but they can be effective which is the priority as I see it. Here effective is simply defined as
avoiding the full brunt of a large decline.
No one rule is always correct. they all give false signals."
"The moving average timing strategy makes the majority of its money by
avoiding large, sustained market downturns. " --
http://www.philosophicaleconomics.com/2016/01/movi...
No. of Recommendations: 2
https://en.wikipedia.org/wiki/Tempest_in_a_teapotTempest in a teapot - Wikipedia
Tempest in a teapot (American English), or also phrased as storm in a teacup (British English), or tempest in a teacup, is an idiom meaning a small event that has been exaggerated out of proportion. There are also lesser known or earlier variants, such as storm in a cream bowl, tempest in a glass of water, storm in a … See more
No. of Recommendations: 1
I think BCC is being misconstrued here as representing a scale.
A value of 0 means all signals negative.
A value of 7 means all signals positive.
But in between there is no scale.
The values 1,2, and 4 each mean that one of the indicators is positive and two are negative.
The values 3,5, and 6 each mean that two of the indicators are positive and one is negative.
So values of 1,2, and 4 are effectively equal, unless you decide to rank the three indicators relative to one another.
The same goes for values of 3,5, and 6.
In that case, what was the original intent for assigning weighted values (i.e., 4 for DBE, 2 for NH/NL, 1 for SMA), giving a range of 0 to 7, instead of a value of one for each indicator, giving a range of 0 to 3? It seems to me that weighted values complicate the system if they don't have any meaning.
JRB
No. of Recommendations: 1
It appears that the strategy works similarly for all 3 indexes.Size matters
INVESCO ISHARES INVESCO SPDR SPDR
RUSSELL MSCI S&P S&P S&P
1000 USA 500 1500 500
EQ-WT EQ-WT EQ-WT COMPOSITE TRUST
Issue type ETF ETF ETF ETF ETF
Technicals EQAL EUSA RSP SPTM SPY
Market price $45.64 $90.40 $167.87 $64.18 $523.79
Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024
Price performance (today) 0.70% 1.00% 0.77% 0.71% 0.64%
Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024
% Price off 10 day SMA 1.49% 1.61% 1.49% 1.78% 1.81%
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
% Price off 20 day SMA 2.19% 2.53% 2.31% 2.39% 2.40%
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
% Price off 50 day SMA 3.96% 5.26% 4.82% 5.21% 5.21%
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
% Price off 200 day SMA 8.77% 12.88% 11.42% 14.23% 14.38%
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
% Change 52 week high 0.48% 0.51% 0.65% 0.63% 0.60%
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
% Change 52 week low 22.20% 28.17% 25.92% 33.88% 34.50%
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
Days since new 52 week high 13 1 8 1 1
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
Days since new 52 week low 146 146 146 363 363
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
Volume 8.0K 8.5K 3.3M 201.4K 25.4M
Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024
Today/10 day avg. volume 0.6 0.3 0.5 0.5 0.3
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
30 day avg. volume 28.1K 57.1K 6.3M 533.3K 70.4M
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
90 day avg. volume 60.4K 55.7K 6.8M 551.1K 75.5M
Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024 Mar-21-2024
10 day avg./90 day avg. volume 0.4 0.5 0.9 0.8 1
Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024 Mar-20-2024
GD_
No. of Recommendations: 4
JRB
In that case, what was the original intent for assigning weighted values (i.e., 4 for DBE, 2 for NH/NL, 1 for SMA), giving a range of 0 to 7, instead of a value of one for each indicator, giving a range of 0 to 3? It seems to me that weighted values complicate the system if they don't have any meaning.
There was no intent to assign weighted values. The reason is below.
From GTR1 for Dummies.
With 3 binary signals, BCC can have 8 possible values (2^3). BCC is equal to 1 times Bear Catcher I's state, plus 2 times Bear Catcher II's state, plus 4 times Bear Catcher III's state, which produces an integer from 0 to 7, each corresponding to one of the eight possible states that all three Bear Catchers can be in. The following table summarizes the eight states and the value of BCC:
BCC BCI BCII BCIII
Value SMA Slope NH/NL Ratio Dying Bullish Euphoria
0 Bear Bear Bear
1 Bull Bear Bear
2 Bear Bull Bear
3 Bull Bull Bear
4 Bear Bear Bull
5 Bull Bear Bull
6 Bear Bull Bull
7 Bull Bull Bull
The next table summarizes the values you can set BCC equal to in backtests to require certain Bear Catcher states:
BCC Bear Catcher State
1!3!5!7 BCI Bullish
2!3!6!7 BCII Bullish
4!5!6!7 BCIII Bullish
> 0 At least one Bullish
No. of Recommendations: 2
JRB .... there's an elegance to Robbie's thinking - although I dont use it in that order.
In my BCC I have these as the combination
(1) NHNL
(2) Slope
(4) DBE
The order is geometric and basically if they enter in order of expecatation - NHNL > Slope > DBE
those can be tested for values of 1, 2, 4 in the BCC=
Also you can test 6 and 4 as exits on the way out - ie NHNL went positive but other 2 remained and DBE remained etc.
You can always come up with your own - its a Linear combination function
Best
No. of Recommendations: 1
Sorry combinatorial values are 1, 3, and 7
and exit states as 6, 4
No. of Recommendations: 8
In that case, what was the original intent for assigning weighted values (i.e., 4 for DBE, 2 for NH/NL, 1 for SMA), giving a range of 0 to 7, instead of a value of one for each indicator, giving a range of 0 to 3? It seems to me that weighted values complicate the system if they don't have any meaning.
It's just a familiar notation for anyone with experience in encoding bits and bytes as 1s and 0s. For some of us it comes naturally as a concise notation of a system's state. For others it may be confusing. I don't believe any weighting was intended.
Elan
No. of Recommendations: 0
>> We now have 15 years of data since the Bear Catchers (BCCs) were discovered. <<
As at least half of the whole thread drifted into correcting misperceptions about how BCC is binary encoded, away from the original subject of the thread, whether (or how well) BCC worked post-discovery:
What are your conclusions?
No. of Recommendations: 3
I don’t think that anyone can definitively say how BCC will work in the future!
As far as I’m aware Zee posted the first complete bear catchers combined signal with all three indicators
on 3/6/2009 in a post Bear Catchers Compared.
http://www.datahelper.com/mi/search.phtml?nofool=y...Bear Catchers Combined is fundamentally unchanged except for minor tweaks to improve the noise level
and effectiveness of each signal from that point.
SPY with BCC compared with SPY alone and VWELX a 60/40 equivalent post discovery from 20090306.
BCC_SPY SPY VWELX
CAGR: 14.1 16.6 11.1
TR: 622.9 904.9
Log2TR: 2.9 3.3
SAWR(20; 0.95): 11.4 13.0
GSD(20): 15.3 17.2 10.2
DIGSD(20; 0%): 18.5 20.1
LDD(20; 0%): 10.0 10.7
LDDD3: 6.1 5.8
MDD: -33.5 -33.7 -20.2
UI(20): 6.0 5.7
Sharpe(20): 0.9 1.0 0.9
Beta(20): 0.8 1.0
TI(20): 16.8 15.7
AT: 0.9 0.0 0.0
So far hasn’t been an advantage but that is understandable in that the only significant bear like event
was the ultra fast covid event which in actuality was so short that it wouldn’t have significantly affected
most of us. The proplem with market timing statistics is that the events are so few and so different
throughout history that is almost impossible to actually apply any statistical significance to any method.
My not so confident method: I look at the BCC signal, market valuation, three different signals on
Allocate Smartly based on Unemployment, Risk Parity and some so called canary signals and usually go
with a riskier approach than an advisor would suggest. So far over the last 30+ investing years I’ve come
out with slightly less drawdowns and slightly better returns than the overall market but who knows
what’s coming.
RAM
No. of Recommendations: 20
As at least half of the whole thread drifted into correcting misperceptions about how BCC is binary encoded, away from the original subject of the thread, whether (or how well) BCC worked post-discovery:
What are your conclusions?
For the 99 day signal (no new high lately), since the original post 2008-09-26
Original version as described has been bullish 84.1% of the time.
S&P 500 real total return (not annualized) during bullish periods has been 226.6%
S&P 500 real total return (not annualized) during the rest of the time has been 4.9%.
So, if you had been long in the bullish periods and had no return the rest of the time, your portfolio size would be worse off by 4.9% having used this as a timing signal.
This result is mainly because of whipsaws, getting out too hastily. (and secondarily because it is a trend following system, and the 2020 plunge was right after market highs)
For variants with the day count of 115 days or longer (up to maybe 175), a long/no-return strategy would have beat the S&P. Lately, it looks like a 135 day timeout might be about optimal.
So...the periods tagged as bullish were indeed mostly pretty bullish. Daily CAGR 1.28%/year higher than the overall time period since the 2008 post.
But the periods NOT tagged as bullish were not reliably bad.
This is a failure if seen as a long/short timing signal, but somewhat in keeping with the original intent of the signal: identifying times it is safe to go in the water, without an attempt to say much about the market at other times. The two signal states are in effect "bullish" and "beats me", not bullish and bearish.
Jim
No. of Recommendations: 6
What are your conclusions?
I had explained my conclusions at the start of this thread, but let me restate them in baseball terms.
The strategy where you got out only if ALL 3 bear catchers turned bearish had a batting average of zero percent at keeping you out of bear markets since 2009.
Over the past 15 years, there were 6 times that the market declined 15% or more. Using the above strategy dragged you through through the full brunt of the market decline all 6 times. Your drawdown was just as severe as that of a buy and hold investor all 6 times.
So that strategy has been a clear failure as a bear detector.
How about the strategy where you got out if ANY ONE of the bear catchers turned bearish? That one was able to limit your drawdown to 10% or less 3 out of the 6 times. So you can say it has a batting average of 50% when it comes to keeping you out of bear markets. Not that great.
And there's a heavy price to pay for this 50% batting average. As a result of all the whipsaws, you wound up with less than half the money and half the Safe Withdrawal Rate as your buy and hold friends.
Take a look at this year so far. S&P 500 is returning 10.2% year to date, whereas someone using the second strategy above is only up 4.5% YTD after being whipsawed 3 times by NH/NL flipping from bullish to bearish and vice versa.
For all the detailed numbers and the periods of the market declines, see my original post at the top of this thread.
No. of Recommendations: 4
there's a heavy price to pay for this 50% batting average. As a result of all the whipsaws....
Doesn't that essentially mean that the only worthwhile mechanical strategy with the goal to keep you out of bear markets would be one that does not even tries to identify "normal" bears but only catastrophic ones (Dotcom Bubble, GFC), keeping you fully invested at all other times? By trying to identify the characteristics only those extremes might have in common?
