Do trends exist?

Research
21

Auto-serial correlation, I see. A fascinating topic.

I wrote a formula to test the rolling 21-day serial auto-correlation of daily returns for the fundamental chart. I am not sure if I can share the chart directly, so here’s the formula:

(LoopSum("Ln(Close(CTR)/Close(CTR+1))Ln(Close(CTR+1)/Close(CTR+2))",21,0,1,0) - 21LoopAvg(“Ln(Close(CTR)/Close(CTR+1))”,21,0,1,0)LoopAvg(“Ln(Close(CTR+1)/Close(CTR+2))”,21,0,1,0) ) / (21LoopStdDev(“Ln(Close(CTR)/Close(CTR+1))”,21,0,1,0)*LoopStdDev(“Ln(Close(CTR+1)/Close(CTR+2))”,21,0,1,0))

It provides a normalized Pearson coefficient between -1 and 1 corresponding to how likely previous daily returns are indicative of current daily returns – on a rolling monthly basis. The squared-value (i.e., r-squared) could be used, I suppose, to indicate whether an observed 21-day trend is significant or just noise.

My suspicion with this is the same as Yuval’s:

A perhaps more surreptitious way I used to test the “illusion” theory was to make spreadsheets using randomly generated data (generated usually by Geometric Brownian Motion – a-la Weiner Process – or some comparable stochastic process). Then I would draw trends line using Excel’s charting features. Then someone would ask me what I was doing whereupon I would explain that I was using technical analysis to uncover the hidden patterns in charts. Usually someone would ask, “but can you make any money off of it?”. I would respond, “absolutely!” Often they would reply back, “cool! Can you show me how to do it?”. – “Of course!”, I would oblige.

About 5 or 10 minutes into it, I would break it to them that all the charts we had studied were randomly generated. All the patterns we had meticulously observed were purely figments of our own imagination trying to make sense of noise. So much for charts…

Or as my old friend Bill on the CBOT used to say, “Charts? I’ll tell ya’ about charts. All ships at the bottom of the ocean had charts on them. Some good it did them!”

21-day ASC of daily log returns.PNG.jpg
21-day ASC of daily log returns.PNG.jpg1768×1100 316 KB

22

In my opinion trends exist due to self-fulfilling prophecy of market participants. If a good number of market participants draw the same trend line, then you would see price bounce off the trend line. A test to assess this could be designed as follows:

  1. An objective way to draw a trend line (e.g. at least 3 points should touch and it should not be pierced) and score it (more touches and more days = higher score)
  2. Now as price approaches closer (based on some measure like 0.25 * average true range) to this, does the proportion of days that close in the direction of the trend higher than 50% (e.g. price approaches from above, then higher positive closes, and price approaching from below, then higher negative closes)

Step 1 is not trivial (at least to me as a non-programmer). However, it would be an interesting exercise if someone has the programming pedigree.

Cheers,
Mukesh

23

Mukesh -

Can price be influenced by a number of market participants all wanting the same price point? If that were the case, prices would never fall, as every holder of a share wants its price to rise.

Every purchase of a share on the secondary market has to be matched by an equal number of sales, and prices only rise when the number of buyers exceeds the number of sellers. So this would assume that when a price is below the projected trend line the number of buyers will exceed the number of sellers and when it’s above the trend line vice-versa. If a huge number of market participants followed trend lines quite religiously and all used exactly the same trend line, which seems very unlikely to me, you’d have to find an equal number of people who were willing to sell at those prices, which seems even more unlikely. Changes in stock prices are driven by all kinds of market behavior, but I can’t imagine a world in which trend lines are a major factor. Nor can I conceive of a mechanism by which market participants’ prophecies about individual stock prices can be self-fulfilling. If that were the case, market participants would group together to make profitable prophecies.

  • Yuval
24

Mukesh,
This is a reasonable hypothesis that I have heard many times. Denny has mentioned this. But it is everywhere, including some of the books I have read and the traders on CNBC program “Fast Money” say this often. I could probably find the reference for the same idea regarding Fibonacci retracements in my (Kindle) library. But I am not telling you something that you do not already know. People’s perceptions can affect the supply and demand curve, obviously. Indeed, I think it is now almost universal that technical analysts try to explain why their idea works (or is plausible) based on people’s perceptions and biases and the effect this has on supply and demand.

It is my personal hypothesis that much of this remains true but that there a people like D.E. Shaw who are a few steps ahead of me on this. Possibly fading whatever trend I buy into. I have no proof of this. DE Shaw is just a example of someone who would be successful at following and/or fading any trend and would have enough money to have an impact on the market. But he does not write books about trading. The traders who do write the books are certainly fading some trends but I do not know how this is working for them.

Specifically, I no longer want to place any bets on what will happen close to a trend-line or a breakout point. But none of this would argue against some (or many) trends being due to people’s expectations—including the expectations of technical analysts.

I would be interested in seeing any studies on this also. If you haven’t checked it out already, I can recommend David R. Aronson’s book called “Evidence-Based Technical Analysis.” But you have already probably looked at it and I certainly would not use this book to argue for or against any particular method.

Best,

Jim

25

Yuval,

Yes, that is true. However, if you follow the market profile paradigm, markets spend time at an equilibrium level where the net buyers and sellers are in balance. For price to break out of this equilibrium, you need a market participant from a higher time frame to intervene or something about the current belief changes (e.g. fed decides to raise the rates, when expectation was to not raise them). The reason for price to show some temporary reaction at points of confluence has to do with the risk / reward profile and crowd behavior. If there is a price point where view points of majority of participants across time horizons aligns, the short-term participants will take a shot as the risk is well defined and low vs. the high reward if the price were to bounce off. I am not sure if the reason this works has to do with premium for providing liquidity by leaning against the intermediate price action.

Then, there are other arguments about market structure (asymmetrical returns) - e.g. channels form because there is not much change in information for long-term participants and short-term market participants start to take positions where the risk is defined (price does not break the channel = low risk) and retraces back to the other edge of the channel = high reward. Some other examples of generally accepted price points are 50% Fib retracements, Victor Sperandeo’s 1-2-3 trend change, 10-day sma - any situation where the risk / reward is skewed (defined risk vs. high reward if they hypothesis worked)

Of course, these are all theories that sound logical on the surface. The true test would be to run this on a enough observations across different stocks and time frames.

Based on my personal experience with day trading, market structure exists on the day time frame. It does produce positive returns if have good discipline to take losses quickly when wrong and hold your winners as long as conditions support (if you can keep instant gratification in check)

26

No. Actually, the sentence is self contradictory. And gets to the heart of the issue of why the prices might rise as Mukesh suggests.

27

Jim,

Thanks for the suggestions. I have moved away from pure technical analysis but use it to improve my entry and exit points (e.g. buying closer to lower band of Keltner channel) for stocks already picked by P123 models. I also initiate my positions using low delta strategies (e.g. credit put spreads) and then add (in the money call credit spreads) if the stock moves in my direction. With this approach, my absolute returns are lower but because they are smoother, I can use higher leverage with options.

Thanks,
Mukesh

28

Mukesh,

Cool. Now I know that it is not just D.E. Shaw that I do not want to trade against! Meaning, I think you would be a step ahead of me if we were trading the same stock.

-Jim

29

I’ve now published the results of my studies of price trends using the S&P 1500 since 1999. You can read it here: https://seekingalpha.com/article/4078738-price-trends-exist and here: http://backland.typepad.com/investigations/2017/06/do-price-trends-exist.html

Thanks to those who helped me work out this knotty problem with your responses to my posts.

  • Yuval
30

Yuval,
I agree with your findings:
In every single case, a stock that was beating the benchmark had a greater chance of not beating the benchmark the following week than of continuing to beat the benchmark, and an underperforming stock had a greater chance of outperforming than of continuing to underperform. In every single case, stocks were more likely to “mean-revert” - to reverse direction - than to continue along their “trend.”

One can demonstrate this easily by using a single factor ranking system.
The one factor system is based on the price changes over a short period. The idea being that ETFs which have experienced a decline over a short period will bounce back, reverting and doing better than ETFs which have not declined in this way.
Here is a model with a 35% annualized return using this principle.

http://imarketsignals.com/2017/im-5-etf-trader/

31

Nice discussion, and nice backtested etf model, Georg. I have a theory to explain at least partly the tendency of some underperforming stocks and perhaps some etfs and funds to outperform going forward awhile that has nothing to do with their fundamentals and all to do with port and fund rebalancing. Most if not all funds of funds and private or institutional portfolios operate with periodic rebalancing. They probably rebalance on widely varying schedules, creating a rebalancing continuum. If they are not cap weighted but instead attempt to maintain equal or proportional weights in their assets, they will sell a percentage of their best performing assets and buy a percentage of their worst performing assets periodically. This could cause at least part of the price action seen. With the expanding use of etfs, funds of funds, and robo-investing in the recent past, this effect could have increased but I have no proof. If true, the effect will tend to act counter to an efficient market theory based on fundamentals. Good stocks tend to get penalized and bad stocks get a boost, even if just temporarily. This could help provide a little market inefficiency for stock picking.

Just a theory at this point. Any thoughts?