Timing - What worked and did for October 2014 down turn.

Research
1

Hi All:

We don’t yet know if this is a major down turn or a minor correction, but it has generated some discussion already.

The discussion got started in another forum section. See it here.
https://www.portfolio123.com/mvnforum/viewthread_thread,8100_lastpage,yes#41660

Here’s my take (as of Oct. 15, 2014).

My three systems use two forms of timing.

My “Macro” timer has not yet given an exit signal. But my “Micro” timer has me approximately 50% in cash as of last week. One method is 33% in cash, another is 50% in cash and the third is 66% in cash.

My “Macro” timer calls an exit if the SP estimates are down AND the moving average of the QQQ is also going down. When it says to exit, I exit ALL positions regardless if any individual stocks are still going up. This timer has not yet said to get out. All three of my systems currently use the same macro timer, but I may decided to use different variations of the timer for each of the systems at some point in the future.

In contrast my “Micro” timer looks at individual stocks and requires them to have an upward path in the medium term. If the medium term path of an individual stocks turns down, that stock is sold. Any potential replacement stock must be going in an upward path AND also be above a minimum rank (eg, rank > “X”). I set “X” to the highest possible level which still allows the method to be fully invested for about 75% of the time in sim tests. If the market is “healthy” for the stocks my systems look for, then the very top stocks on the rank list will be in an uptrend in the medium term. If the highest ranked stocks are not currently going up, I don’t want lower ranked stocks even if they are going up.

Oh, I turn the Macro timer off when running sims to determine the optimum level for “X” for the Micro timer, but turn it back on for my live portfolios.

In back tests the Micro timer does not add much value. the Macro one does all the heavy lifting in back tests. I use the Micro timer because I don’t fully trust my Macro timer. I’ve done a lot of work on market timers over the years and have discovered that I can make Macro market timers that are virtually perfect on past data but more hit and miss when running live. Why not just use the Micro timer without the Macro? Because the back tests give higher profits and lower drawdowns with the Macro timer than without it. My hope is the Macro timer will work at least some of the time in the future. If it works half the time for calling an appropriate exit, I’ll be happy. The Macro timer is very aggressive for getting back in (it will give a BUY signal if EITHER the QQQ is helding up OR the SP earnings estimates are heading up, so I’m not worried it will miss any significant market rebound off a low. If the Macro sometimes fails to call an exit (from experience I know it will fail to do so at least some of the time, or at least be rather late giving its exit signal), I’m counting on the Micro timer to give partial protection.

Regards,
Brian

2

I was curious as well about the recent October drop and checked the R2Gs as a proxy for what the community actually trades and how everybody did.

I viewed all models and restricted the selection to a launch period of at least three months (to be far enough away from the recent turmoil).
Afterwards, I ranked all models by their one (respectively three) month(s) return. Only 16 out of 193 models have a positive one month return. For the three months return it’s 29 models with positive performance. Not surprisingly, the trader models are more volatile. Checking the R2G front page’s current “Average Out-of-Sample Statistics” as of today is actually a little scary.

My macro bear market rules were not triggered yet, either. Looking at a long term S&P 500 chart or Shiller’s 10 year PE (or many other market indicators), the market might be overvalued. Maybe there will be medium setbacks - but then again, what are the alternatives to owning equity right now?

Market timing is a lot of speculation, and (macro) events are hard to predict. I am currently more working on systems that invest only after sharp drops and then gradually phase out as the market recovers.

3

Hi fips,
I think this is a general dilemma we wan’t solve (at least not for the future).
Long and drawn out declines are easy to escape with the tools you mentioned (Shiller, SP500 forward earnings etc).
The current violent and quick downturn cannot be predicted imho. Short moving averages could (in theory) do the trick but then again, will be whipsawed to death in a more sideways market.
Therefore, we have to live with it. Drawdowns are inevitable and they cost nerves, money and unpleasant feelings. It’s the cost of holding stocks.
If someone does have an idea that I am not aware of, I’d be gladly listening.
Werner

4

“The current violent and quick downturn cannot be predicted imho.”

Yet we know that October is the most volatile month of the year by far. And September has the worst record of any month of the year (by far).

All I can say is “you can lead a horse to water but you can’t make it drink.” Sorry about the cliche.

Steve

5

Steve,

I do agree with you about volatility in October and Sept. There’s a lot of research not just on equity markets but on options vol. that shows this historically.

Most professional ‘index timers’ take very similar approaches. For the most part they -
a) create a set of sub-systems (from 4-6 to thousands).
b) sum the signals from the subsystems to get a weighted signal ranging from -1 to +1 (but often in the .5 to .8 range).
c) (sometimes) adjust the weighted signal for trailing forecast accuracy
d) (usually) adjust the total exposure based on 1. the weighted signal score modified by 2. the trailing period system vol. vs. the target system volatility budget.

Parts a) and b) above can be easily done in P123, but can’t be offered to sub’s as R2G’s as they require a book. Part c) is outside the scope of P123. Part d) I have posted feature requests on. If we could adjust amount invested with a formula would allow for this (especially if book leverage could be adjusted based on risk budget).

The current problem is that too many people still view market timing as a binary event. When multiple systems are considered at a book level, market timing is a fairly normal (although not universal) part of how many of the best professional traders manage risk.

Best,
Tom

6

Fips,

What you are saying is that you don’t believe in any market timing, so you will engage in mean reversion based market timing. You are saying you will change your portfolio level allocations based on a belief in mean reversion. This is just another form of market timing.

Best,
Tom

7

Here is a good summary (since 1950): http://www.advisorperspectives.com/dshort/commentaries/SPX-Tops-and-Bottoms.php

8

Interesting link, thanks. It’s interesting that October had so many market bottoms. Aren’t market bottoms a good time to buy?

9

Chipper - only if you are sure they are bottoms! Oct 1987 and 2007 were down months but not bottoms.
Steve

10

How about the presidential cycle (2nd term)?

[url=http://www.mcoscillator.com/learning_center/weekly_chart/2nd_presidential_years_are_different/]http://www.mcoscillator.com/learning_center/weekly_chart/2nd_presidential_years_are_different/[/url]

Perhaps it is all up and away from here.

Steve

11

Tom,
I know that what I proposed is just another way of market timing. What I meant to say was that I am trying to alter my way of handling market signals (based on mean reversion). I am trying to find different indicators for when to exit the market and when to come back. I guess it’s similar to the approach you are talking aboug.

Steve,
thanks for the intersting links. Here’s another one on the presidential cylce:
Washington Meets Wall Street, as appeared in Journal of International Money and Finance.

Best,
fips

12

Fips,

‘Index timers’ tend to use some or all of the following:

  1. Breakout
  2. Trend
  3. Mean reversion (Bottoms)
  4. Trend exhaustion (Tops)
  5. Sometimes some fundamentals are used, but they haven’t been a big part of any program I’ve ever looked at.
  6. Other pattern (i.e. trading long-short on two highly correlated instruments based on expectations for them)
  7. Index arbitrage (i.e. all the stocks in the SP500 vs. the Index…usually hi-frequency based).

People trade a variety of ‘cycle lengths’. From intra day, to a few days to weeks or months.

On the more ‘tactical asset allocation side’ (like once every year or once every 3 years), people do make tactical allocation shifts. This is where CAPE levels and the larger fundamentals are more often utilized.

Each type of system has an environment it will naturally do well in and others it will ‘fail in.’ Mean reversion and calling tops type systems do very well in choppy markets without clear direction. And mean reversion also does well when market set-backs are small’ish and temporary. Trend tops do well in markets near exhaustion and reversal. A long up bull will kill this strategy. Trend (with longer cycles) does very well in violent and/or sustained up and down moves. An oscillating and/or sideways market can kill these systems.

So…they are all put together into a weighted signal with final weighting driven by vol. vs. risk targets precisely because they all do well in varying market types.

Best,
Tom

13

Doubling down