Forecasters of the stock market

Manage
21

I’m going to call upon a phrase often used by a professor with who I once studied: “That sounds good if you say it fast enough.”

Have you ever made a truly random investment decision? Have you ever heard of anyone having done so? What might a truly random investment decision look like?

“I’m buying XYZ because of good fundamentals.” That’s obviously not a random decision.
“I’m buying XYZ because of such-and-such technical signals.” That, too, obviously is not a random decision.
“I’m buying XYZ because the company’s products are cool.” That may be naïve, but it’s not random.
“I’m buying XYZ because my neighbor said it’s hot.” That may be foolish and reckless, but it’s not random decision.
“I’m selling XYZ because somebody in the background coughed too much during the conference call.” That’s bizarre, but not random.

To us, the sum total of all these decisions, and a gazillion like them, may appear random. But are they truly random, or is that a function of our inability to create and implement, say, a model that looks like this:

  1. Good fundamentals according to the criteria of Investor A (something%) higher is better
  2. Good fundamentals according to the criteria of Investor B (something%) higher is better
  3. Good fundamentals according to the criteria of Investor C (something%) higher is better
  4. Good technicals according to the criteria of Investor D (something%) higher is better
  5. Good technicals according to the criteria of Investor E (something%) higher is better
  6. Products Investor E will like enough to cause him to buy the stock (something%) higher is better
  7. Products Investor F will like enough to cause him to buy the stock (something%) higher is better
  8. Bullish gossip from the neighbor of Investor G AND Investor G’s willingness to act on it (something%) higher is better
  9. A nagging cough on the part of Mr. X, the CFO AND Investor H’s inclination to be freaked out by this sort of thing (something%) lower is better
  10. Etc. etc. etc. for a gazillion other factors involving a gazillion other investors, some silly, some brilliant and a whole bunch in between

We can’t create such a model today. But is it not possible we may be able to do it by, say, 2715?

As to why p123 folks can do well now . . . they don’t have models like this, but they may recognize, for example, that factor one involves criteria can be articulated and given that it’s likely to be shared by many investors, that it’s overall weight in the sort of models we are able to build ought to be somewhat high. Ditto #4, which implicitly may capture many unidentifiable instances of #6 and #7, however imperfectly as well as even #8. As for #9 and #10, well, that’s why even the best p123 models can’t be expected to work perfectly 100% of the time.

Who are the folks who believe in randomness? They are those who don’t even try to comprehend the sort of dynamics of the above model, ignore the whole thing, throw up their hands, and say it’s random.

22

Marc - perhaps you should read this article: http://investingcaffeine.com/2012/07/08/experts-vs-dart-throwing-chimps/

The article pertains to a Nobel prize winner’s theories so I’m sure you will respect his point of view :slight_smile:

Are you sure that we have learned from 1929 or have politicians learned to spin the truth? Now recessions aren’t declared until after they are over. The “D” word is never mentioned. Why? Because the politician in power doesn’t want it known that he is doing a poor job. The definition of GDP and inflation are constantly tinkered with. Employment statistics are meaningless.

1987 was the first program trading related market crash (not economic driven). Yet we now have had a new breed of program trading crash - the FLASH crash. Have we learned from the FLASH Crash? There is no doubt in my mind that it will happen again.

Have we truly recovered from 2008? Politicians would like us to believe so but full-time meaningful employment hasn’t recovered, only corporate profits.

Saying the markets are deterministic is like saying the game of chess is deterministic. Can the person playing the white pieces win from the first move? Who knows. Maybe in 100 years Big Blue will have that figured out. But then Chess will no longer be. Just like the stock market will no longer be if there is no uncertainty of outcome.

Steve

23

The article was standard rehashed gibberish about how experts suck, the one exception apparently being the author of the article, his money management firm and his own how-great-I-am books which aren’t even available on Amazon. The ideas of the Nobel Prize winner he pretends to write about, however, look fascinating. (I picked up the latter’s book, only $2.99 on Kindle.) Thank you for the suggestion.

How many people do you know who’ve lost their entire savings (supposedly basic plain-vanilla bank accounts) because of bank failures? Talk to people who actually lived through the depression. I think you’ll find that we have, indeed, learned a heck of a lot.

Yeah, that’s the rhetoric being tossed around, mainly by reporters and pros who try to dodge accountability for having invested stupidly back in the day. But it’s very much inconsistent with the way I remember the world. For one thing, valuations back then were insanely high. I still remember the thick August 1987 Goldman Sachs report on their then-standard blue paper blue penned by Abby Joseph Choen (sp?) explaining how traditional valuation metrics are obsolete. My instantaneous reaction was “Oh sh**, JP Morgan’s shoeshine boy is speaking again!” What I also remember was the U.S. trade deficit soaring badly ion the context of the economy as of that point in time with the penultimate crash-day catalyzed by a spectacularly worse-then-expected trade deficit number. (And we know, of course, that besides the implication on currencies, inflation rates, and interest rates, trade deficit is a direct subtraction from GDP (as was a the case with GNP when we used that).

Don’t get hung up on derivatives, algorithms, etc. Those just provide individual flavor, and blaming them for what happens is rhetorical bullsh**. (However much many distrust politicians, the latter are, at least, way above the media when it comes to getting thing right.) The ultimate reality is that whatever the mechanics, bad fundamentals mean bad stock prices, lending to borrowers who can’t repay means loan crisis, etc., etc., etc.

24

Along Marc’s thought of modeling everything:

Is anyone interested in building a database of text data, such as from twitter, letters to shareholders, press releases, analyst reports, etc. and using it to backtest and build models, and combine with the existing numerical financial data? I’ve read papers that showed daily tweeter mood is correlated with the stock market. This is very fascinating to me, but I have neither the time nor programming expertise to do this. I would love to test the predictability of text strings such as, just off the top of my head: “removed going concern”, “exceeded expectations”, “lawsuit settled”, “announced contractural agreement” etc.

25

Marc - my grandfather bought a farm in the summer of 1929. Two mortgages and no experience farming. In the early 30’s the banker came along and ripped up his mortgage as he knew he wouldn’t get anything for the farm. Today we have robo-forclosures. I’m not sure of the current state of affairs, but it wasn’t that long ago when 6? million people were on the streets in Florida. If we have learned so much then why wasn’t the issues of 2008 prevented? And why haven’t issues really been fixed?

In any case, dart-throwing regularly outperforms the wall street so-called experts. It is not difficult to argue that dart throwing is a legitimate stock picking technique.

Steve

26

Steve and all,

I recently read a book called: “Fragile by Design: The Political Origins of Banking Crises & Scarce Credit” by Charles W. Calomiris and Stephen H. Haber. Steve (and others from Canada) may find this particularly interesting because it covers banking in the U.S. and Canada. It explains why Canada has never really had a banking crisis (I do not know if Canadians agree with this). More important for me: it clearly explains why we had a crisis in the U.S., in my opinion. And it supports Steve’s view that things are not likely to ever be fully fixed due to politics.

Downsides: 1) It does not give one much hope that the banks will ever be separated from politics and so banks will always remain fragile–a definite prediction (so probably appropriate for this post. 2) It won’t help you pick stocks. These are business economists from Stanford University. The book is thick and heavily referenced. However, the parts I read (mostly the U.S. stuff) was an easy, entertaining read–doubtful that it is meant as a textbook.

27

YEP! :sunglasses:

28

“It is not difficult to argue that dart throwing is a legitimate stock picking technique.”

Steve, Yes you are right, but we have to ask why this is the case.

Those random systems usually equal weight and that itself is a model, because it (usally) weights small caps and mid caps the same as big caps and
small caps and mid caps outperform big caps…

Regards Andreas

29

Jim, thanks for recommending a very informative book! I just read through the reviews on Amazon (there are already a few good ones) that summarized the book and I learned more about the US banking system in 5 minutes than I have learned in the past five five years–and I have read up on it quite a bit since the financial crisis.

30

Andreas - In some lotteries, you can buy what is called a “Quick Pic”, or random set of numbers generated by a computer. The odds of winning aren’t any better than using any other strategy for playing lotteries. The odds aren’t better, but the payoff generally is, simply because if you win then you probably don’t have to share the money with too many other players as people tend to focus on certain numbers and patterns.

I view the stock market in a similar light. Throwing darts doesn’t improve your chances of picking good stocks but you are more likely to pick stocks that aren’t as popular, and certainly not as popular as wall street darlings. Thus the payoff tends to be better.

Steve

31

The world is NOT deterministic, ask Heisenberg http://en.wikipedia.org/wiki/Uncertainty_principle

As as dwpeters pointed out correctly „our predictions influence the market „ as soon as we or someone trades based on them. http://en.wikipedia.org/wiki/Feedback#Economics_and_finance

As soon as we start observing, we change the system. http://en.wikipedia.org/wiki/Observer_effect_(physics)

Therefore markets are not and will never be deterministic. http://en.wikipedia.org/wiki/Determinism

So the market is NON deterministic http://en.wikipedia.org/wiki/Nondeterminism
The market is a stochastic process http://en.wikipedia.org/wiki/Stochastic_process

But NOT in a smooth sense like the random walk (independent identical distributed, http://en.wikipedia.org/wiki/Laplace_distribution), which results in (CAPM, Black Scholes, Efficient Market Hypothesis, Modern Portfolio Theory, Indexing, ETFs)
This would mean we can not beat the market, we just buy ETFs and follow it.

Rather the market is much more complex dependent, fragile, fractal, non linear feedback system, with NON DETERMINISTIC / STOCHASTIC elements. http://en.wikipedia.org/wiki/Chaos_theory#Distinguishing_random_from_chaotic_data

So eventhough the market is non deterministic/random/stochastic, there are certain dependencies, we can exploit.
Certain stockfactors and formulas, which in the past have and might in future outperform the market.
But we can not predict the market with certainty since it is NON Deterministic.
Therefore we have to distribute our bets over many stocks, models and factors to increase the probability of winning.

WE CAN OUTPERFORM THE MARKET!!!

But predicting (in the sense of market timing) a

non linear feedback system, with NON DETERMINISTIC / STOCHASTIC elements and complex dependencies

is NOT possible.

Instead we have to expoit the dependencies and ride our stock factors through up and down markets. No excess return without riding the Drawdowns.

32

“And the lesson for us all is, market timing does NOT work!!!”(Tobiasberr)

I totally agree.

Maybe it is because the future is random, or maybe because we are not living in 2700 and we can’t use, or know all the data.
It is really not important why we can’t predict the future, it’s a fact that we can’t.

Some ready2 go systems performances, are base and depend upon market timing and hedge them
Some past simulations look beautiful, but the future is what really important and I be very surprise to see them working forward in the future, without “fixing” the past all the time

That exactly why all my “Sherman’s Way 2 go” systems (and other designers) are not using a market timing

Because at the end of things,
It is not so important predicting the market future; only beating it

Amiran

33

Are we timing the market or just trading the market?

We are given many good old sage advice like;
“sell the losers and let the winners run”
“Follow the trend”
“the trend is your friend”
Does any of that advice work?

And there are the technical indicators that are so often quoted.
Can a system that trades between SPY and SH work?

And we follow our trading systems to pick the best stocks to buy.
We use rules to buy into stocks and sell out of them. Do they work?
If they don’t, why are we here?

I contend that if any of those approaches works for stocks and/or ETFs, then there is NO justification to say that they won’t work for the market.
If you can’t figure out how to use the P123 tools to trade in and out of the Market, or use hedges on a timely basis, then don’t try to tell us it can’t be done.

What determines whether or not timing has been successful?
I say that it is successful if: 1) you avoid some of the drawdown, and/or 2) get back in before the market recovers to a value higher than you got out. If we can achieve both 1) and 2), we are doubly successful.

I have been successful with either 1) or 2) or both since the oil crises in 1973>1974. Don’t try to tell me it can’t be done.

Denny :sunglasses:

34

Hi Denny,

Outperforming the market by
Market timing is much more difficult, than riding the right factors through bull and bear markets.

It is much more difficult building robust market timing models, compared to robust ranking systems.
The 14 year Backtest on a ranking system looks back on a few thousand trades, where as the 14 year backtest of a market timing or hedge rule has only 10 or twenty entries/exits.
Therefore it is far less statistically reliable.

I have never succeeded with market timing / hedging.
I always sacrificed some overall performance, to get smaller maximum drawdowns, compare to using the model without.
And as a result, especially based on the low statistical backtest reliability, I have always gone live without.

Denny’s R2G MidCap & LargeCap, 10 stks, AvgDailyTot >$10Mil did have some bad market timing since launch.
https://www.portfolio123.com/app/r2g/summary/1045249
So even you as a Pro are having difficulties.

P.S. Please do not take this too personal, you are definitely the much more experienced and better model designer and trader. I really respect you.
I am with P123 since one year, you are a member since 2004.
In 1973 I was not even born or in the womb, so I am definitely not in the position to critizise you in any way.

35

In addition to Denny there is Ray Dalio.

I admit that I cannot time the market very well. But Ray Dalio (Bridgewater Associates) who manages the largest hedge fund in the world changes his holdings according to the type of market he is in. He predicted the financial crisis in 2007. He is well diversified and a lot of his holdings are uncorrelated but he decides what is correlated and not correlated according the the type of market he is in. He has a long successful track record.

It is hard but I would not go so far as to say it cannot be done.

36

I would love to think it is possible to time the market, but I am suspicious. I have seen respected R2G designers launch models, then ‘tweak’ the timing rules when there are problems.

37

Here is a study starting in 1871 which shows that a simple moving average crossover system has handsomely out performed “buy-and-hold”.
http://www.advisorperspectives.com/newsletters09/Moving_Average-Holy_Grail_or_Fairy_Tale-Part_1.php
http://www.advisorperspectives.com/newsletters09/Moving_Average-Holy_Grail_or_Fairy_Tale-Part_2.php
http://www.advisorperspectives.com/newsletters09/Moving_Average-Holy_Grail_or_Fairy_Tale-Part3.php

My own MAC system http://advisorperspectives.com/newsletters12/Beyond_the_Ultimate_Death_Cross.php was backtested from 1965 onward and uses daily values of the S&P500. A simulation in P123 using SPY and IEF from 1999 shows an annualized return of 11.0% with a max drawdown of 17.1% with 10 completed trades and one open position, all of them winners. This return is very similar to what the model provided from 1965 onward. The transaction list is attached.

So it would appear that slow market timing works quite well, but I would be suspicious of rapid market timing models.
Georg

Transactions MAC basic.xls (13.5 KB)

38

How many test points are required to be statistically significant? Some experts say 30, or for a higher accuracy, maybe 100. How much accuracy do we need to be comfortable? How many market events have occurred in the P123 data since January 1999? There have been only 2 recessions, over 20 “corrections” of 5% or more, but there are hundreds of losses over 2%. So what defines a data point? If your analysis looks at every S&P 500 loss over 2%, but rejects those that don’t continue “trending” lower below some additional drop, is your system using hundreds of data points?

There are many other sources of market technical data that goes back many decades, and contains a sufficient number of recessions to be statistically significant. Are we not allowed to use that data to develop a system and then use it with P123 sims & ports?

And what about Fundamental data? Market average earnings comes to mind. Has it not been shown over decades that when the average market earnings declines quarter after quarter, it is not a good time to be in the market?

How about the leading economic indicators. Is it useless to look for meaning and direction in them? Why do traders move from growth stocks to defensive stocks when the warning signs say to do so? Is that not market timing?

Instead of agreeing with the talking heads that say “you can’t time the market” and therefore not try to figure it out, is it not instead possibly wise to spend significant time and energy trying to figure it out?

What defines whether a timing system “works” or not? My approach is to try and avoid at least 50% of the market loses during recessions to protect my capital. I don’t try to avoid corrections. However, an approach that avoids 50% of the recessions’ losses, but may also have a few false starts including small whipsaws, is a very good system in my mind, and has met my needs.

Denny :sunglasses:

39

@Amirans. Congrats on your fine R2G performance since launch. But…there are many ways to invest and construct a portfolio.

Here’s some fun reading (Howard Mark’s latest memo):
http://www.oaktreecapital.com/memo.aspx?AspxAutoDetectCookieSupport=1

Any decision in investing involves risks. And potential rewards. Do we understand them?

Every 6 to 12 months, a new thread starts about why market timing is ‘impossible.’ Or, more recently, about why some R2G designers are ‘tricking’ potential sub’s. It’s amazing in its regularity. The old threads exist. Read them if you want.

As important, the data sets exist. Hundred plus year data sets for many indexes. Do the work. Do your own research. Do the tests and come up with timing rules or not.

But know that…Hedge funds and future traders hedge and make big bets and ‘market time’. So do many family offices. And most professional quant funds. Many of these use futures index overlays with mechanical trading rules to try and more positively skew the risk reward of the portfolio. And nearly all CTA’s use mechanical trading systems - computers following rules to make trades on major macro indexes…most prominently the SP500, QQQ and US bonds. At various time frames…from milliseconds to weeks typically. They have multi-decade long out-of-sample, audited histories. And are mostly, if not completely, ‘technical’ traders.

And…nearly all of the biggest discretionary trades in history are big discretionary ‘market timing’ bets. John Paulson against sub-prime. George Soros against the pound. John Templeton at the height of the Depression. Sometimes they end badly.

So…Know that if you ‘market time’…or actively trade, or post models on R2G…sometimes you will ‘look bad’, even for long periods. It doesn’t mean your thought process or research or decision making was wrong. (although it could). Likewise, outperformance over a 6 month (or 3 year period) is not ‘proof’ that your way is best.

And individual stocks are no less ‘chaotic’ then the broad market. And, in fact, likely much more chaotic. A stock, for example, can be radically influenced by all the following:
a) death, illness or crime of a CEO or key executive
b) really unusual weather
c) key cost input changes due to wars, random events, etc.
d) political climate or legislative changes
e) changes in management or key positions at a major customer or supplier
f) unforseen actions by a major hedge fund player(s)…see Herbalife and tug-of-war going on currently between Bill Ackerman and Carl Icahn. A mid cap that has major players playing an ego game…that could lead to 100% drop or 100% gain.
g) Major actions by competitors (see Game Stop and Walmart’s decision to enter the video game business…)
h) Technological changes (see Blockbuster and it’s death by the ability to download video).
i) Etc.

Many more stocks go bankrupt annually than entire economies. So…most people don’t put all their money in 1-2 stocks. Likewise, timing systems are best diversified and blended.

And, so…Some of the major questions in using market timing are:

  1. Have we put in the time so that we likely understand the risks well and deeply?
  2. Are we likely to be compensated for the risks we are taking?
  3. Are we ‘sizing our bet’, and constructing our portfolio, appropriately?
  4. Can we live with the ‘worst possible’ outcome…if that comes to pass?

And also for what it’s worth. I have and trade weekly ‘market timing’ systems. And monthly. And quarterly. Many have made me real money while having zero to neg. market correlation. Believe me or not. But, I trade fast and slow moving signals across a variety of major indexes and markets. Even just using the past 14 plus years in P123, there are nearly 800 weeks in the data set. I’ve built many 2 and 4 factor rankings on SPY, TLT, QQQ, etc…systems that make, on average, bi-weekly moves. For me, that’s plenty of data for some real level of confidence for that number of factors for say, a 1-2% portfolio allocation.

I know that sometimes every timing system will be wrong. So what? So is every factor and rule that can be used. In some market cycles high flying micro cap systems will get crushed. They still can be part of a balanced portfolio approach. Or…the entire portfolio if you want those risks.

I use market timing to control portfolio exposure in market conditions that I don’t like. Other people may prefer strict ‘asset allocation’ risk rules/management. Fine. They’re allowed.

But…I think what ‘bothers people’ (at least me) is when people post threads that ‘all market timing is bad’ and no one should do it. Or that any posted R2G using any hedging or market timing is somehow flawed or will implode. Or any system that is revised.

I would, instead, welcome more focused posts on the merits, strengths and weakness of specific market timing ideas. With research supporting people’s positions.

Best,
Tom

40

Thank you for your posts, Denny, geov, Tomyani.
You have convinced me.
I really have to learn a lot more about market timing.