May

12

 UPDATE:

The Winners of the least effort contest were jointly in a tie. Mr. Gary Rogan and Mr. Steve Ellison. I will split the prize between them. The creative and physical ideas of Mr. Rogan were very excellent and best of all, but there was no testing. Mr. Ellison gave a great test, and a complete answer, but Rogan can't be denied his place either. vic

I'll give a prize of 1000 to the person or locus of his choice that comes up with the best way to test the principle of least action or a related principle of least effort.

It's in honor of my grandfather. Whenever I'd ask him which way he thought the market would go he'd say, "I think the path of least resistance is down" starting with Dow 200 in 1950. We need some more quantification around here.

You might consider max to min or a path through a second market back to home. Or round to round? Or amount of volume above or blow. Or angle of ascent versus angle of descent. Or time to a past goal versus the future? Or some mirror image or least absolute deviation stuff?

Sushil Kedia writes:

With utmost humility and clearly no cultivated sense of any derision for the Fourth Estate, I would submit that since it is the public that is always flogged and moves last, the opinions of all media writers, tv anchors are the catalysts, the penultimate leg of the opinion curve. A test of the opinions of the fourth estate on the markets would provide the most ineffective wall of support or so called resistances. Fading the statistically calculated opinion meter (if one can devise one such a 'la an IBES earnings estimate a media estimate of market opinion) and go against it consistently over a number of trades, one is bound to come out a winner. Can I test it? Yes its a testable proposition, subject to accumulation of data.

Alston Mabry writes:

The following graph (attached and linked) is not an answer but an exploration of the "least effort" idea. It shows, for SPY daily since August last year, the graph of two quantities:

1. The point change for the SPY over the previous ten trading days.

2. The rolling 10-day sum of the High-Low-previous-Close spread, i.e., "max(previous Close, High) minus min(previous Close, Low)". This spread is a convenient measure of volatility.

Notice how these quantities move in tight ranges for extended periods. These tight ranges are some measure of "least effort", i.e., the market getting from point A to point B in an efficient fashion. As one would expect, the series gyrate when the market takes a temporary downturn. Also note how when one of the quantities swings above or below it's mean or "axis", it seems to need to swing back the other way to rebalance the system.

Bill Rafter writes:

 This nicely illustrates how relative high volatility is bearish on future price action.

Jim Sogi writes:

The path of least resistance would be the night session. Low liquidity allows market mover to move market. Every one is asleep. Dr. S did a study some years ago. Updating shows total day sessions yielding 94 pt, but night session yielding 232 points. Don't sleep…stay up all night or move to Singapore. Recent action is in line with hypothesis.

Bill Rafter writes:

Haugen's "The Beast on Wall Street" (i.e. volatility) came to the conclusion that if you want less volatility in the markets, keep them closed more, to essentially force the liquidity into specified periods. That is, 24 hour markets promote volatility. Or a corollary was that a market is never volatile when it is closed. [this is from memory and I may also be regurgitating from a personal conversation with him]. An oft cited example is the period in the summer of 1968 when equities were closed on Wednesdays to enable the back offices to get up to date with their paperwork and deliveries. During that time the Tuesday close to Thursday opening was less volatile than expected (twice the daily overnight vol).

One could take this thought and stretch it to say that the periods of least resistance would be those without heavy participation. One could easily compare the normalized range (High/Low) of those periods versus the same of the well-participated periods.

Craig Mee writes: 

Hi Bill,

You would have to think that in 68 there was sufficient control of price and news dissemination. In these times of high speed everything, that this could create bottlenecks and add to the volatility. No doubt a bit of time to cool the heels i.e limit down and up for the day restrictions, is a reasonable action, even if it goes against "fair open and transparent markets" but unfortunate it seems little is these days.

Bill Rafter replies:

I should have been more specific about the research: take the current normalized range for those periods of high liquidity (when the NY markets are open) and compare that to the normalized range of the premarket and postmarket periods. Do it for disjoint periods (but all in recent history) so you don't have any autocorrelation. My belief is that you will find there is less volatility intra-period during the high liquidity times. While you are at that you can also check to see during which period you get greater mean-reversion versus new direction.

If that research were to show that (for example) you had greater intra-period volatility during the premarket and postmarket times, and that those times also evidenced greater mean-reversion, you could then conclude that those were the times of least resistance. That would answer Vic's question. Okay, now what? Well you could then support an argument that with high volatility and mean reversion you should run (or mimic running) a specialist book during those times. That's not something I myself am interested in doing as it would require additional staff, but those of you with that capacity should consider it, if you are not yet doing so.

Historical sidebar: '68 was a bubble period caused in part by strange margin rules that enabled those in the industry to carry large positions for no money. The activity created paper problems as the back offices were still making/requiring physical delivery of stock certificates. The exchanges closed trading on Wednesday to enable the back offices to have another workday to clear the backlog. The "shenanigan index" was high during that time.

Phil McDonnell writes:

Bill, you said "During that time the Tuesday close to Thursday opening was less volatile than expected (twice the daily overnight vol)."

For a two day period and standard deviation s then the two day standard deviation should be sqrt(2)s or 1.4 s. So the figure of twice the volatility would seem higher than expected.

Or am I missing something? 

Steve Ellison submits this study:

The traditional definition of resistance is a price level at which it is expected there will be a relatively large amount of stock for sale. 
Starting from this point, my idea was that liquidity providers create resistance to price movements. If a stock price moved up a dollar on volume of 10,000 shares, it would suggest more resistance than if the price moved up a dollar on volume of 5,000 shares.

To test this idea, I used 5-minute bars of one of my favorite stocks, CHSI. To better separate up movement from down movement, for each bar I calculated the 75th and 25th percentiles of 5-minute net changes during the past week. If the current bar was in the 75th percentile or above, I added the price change and volume to the up category. If the current bar was in the 25th percentile or below, I added the price change and volume to the down category.

Looking back 200 bars, I divided the total up volume by the total up price change to calculate resistance to upward movement. I divided total down volume by the total down price change to calculate resistance to downward movement. I divided the upward resistance by the downward resistance to identify the path of least resistance. If the quotient was greater than 1, the past of least resistance was presumed to be downward; if the quotient was less than 1, the path of least resistance was upward.

For example:

                           Previous 200 bars
                                                        Up
   Date     Time     Up Points Volume  Down Points Volume Resistance

3/25/2011   15:50   53   6.49  99431    61  -7.38  149867     15311


   Down       Resistance     Actual
Resistance      Ratio      net change
     20310       0.754       -0.03

Unfortunately, the correlation of the resistance ratio to the actual
price change of the next bar was consistent with randomness.


Comments

Name

Email

Website

Speak your mind

7 Comments so far

  1. Brendan Dornan on May 3, 2011 11:58 am

    The number of units that it takes to move a stock one level turned into a running ratio. The truth about tape reading is you learn much more about the strength of a stock by how firm the bid is when people are selling, and considerably less when seeing offers being taken.

    I have academic papers showing this principle to be predictive, and i would use the 1000 towards having them replicated and tested. This task is considerably more complicated than it sounds in the age of dark pools.

  2. Ryan S. on May 3, 2011 2:57 pm

    1. Take percentage up or down for every market trading before start of U.S. market.
    2. Average these percentages weighted by capitalization…write down.
    3. At lunch time in the U.S., average in the U.S.market…write down.
    4. At end of U.S. day, average all again…write down.
    5. See if the first number predicts the second (or third)…See if the second number predicts the third.

    Maybe you get a “lunch time path of least resistance.”

  3. Nick Grosvenor on May 3, 2011 8:50 pm

    I think that Amazon’s mechanical turk is the perfect lab.

    https://www.mturk.com/mturk/welcome

    Post a job description asking people to transcribe a short 30 second audio clip, filled with just a few short simple sentences.

    on the page that has the media player playing the audio, provide a link to a computer transcriber. Explain that if they download the program which takes approx 20 mins to download, it will quickly transcribe the audio automatically, (use a technique to distinguish between a human and computer transcriber, for later analysis eg; program saves to different file format or saves with metadata)

    pay them 5 cents per job.

    Faced with this circumstance, our worker could quickly, within 5-10 minutes manually transcribe the audio, or go through the trouble of waiting 20 or 30 minutes to download a tool that could repeatedly do the task for them automatically and earn more money over the long run.

    the logical thing would be to download a program and repeatedly use it to quickly transcribe audio.

    the path of least resistance would be to quickly and manually get the job done.

    what would transpire?

    how many variations could be made with the amount of payment, the time to complete the task manually, and the time to download the program.

    By using a $1000, you could get a 180,000 point data set. (amazon gets ten percent)

    If this isn’t what you’re talking about. If you’re referring to some technical market idea using trader speak, then I just wasted 30 minutes coming up with this idea.

    But i still like it.

  4. Robert McAdams on May 4, 2011 11:49 pm

    I’ve heard that a stock needs money to drive it up, but can fall under its own weight. Since higher price stocks require more money to move, once they are owned by everyone, the path of least resistance must be down. Best to wait for a positive news event and note the reaction.

  5. Ken Drees on May 5, 2011 10:52 am

    The path of least resistance to me means a trending market and not a chop, and therefore not a sideways market.

    I like a series of tests together to prove that the market was trending and therefore simply stay with the movement up. I only use this method for trades in a up trending market

    For the purposes of an up market:

    If today’s close is greater than the close of two day’s ago and this fact has occured at least twice out of 4 measurement tests, then it would be considered a positive. If ever 3 out of four tests are negative then you restart the count. Once you get 2 of 4, enter the trade long.

    Couple this with a layover of a simple bollinger band chart. I consider a trending market to be one that has already moved up off the lower band and into the top 30% or more and even touching or exceeding the top bollinger band.

    So I want the 4 day series counts positive and I want that trading to be occuring in the upper zone of the band. This trading concept does not call your bottoms.

    Triggers for getting out are harder for me to define:

    I look at band width compression and then a trading day close below the high 30% to mid point of the band (or even intraday violation). A close or an intraday price exceeding the top band sometimes occurs which indicates a short term or long term high for the trending move. A combo of TWO negative count tests back to back AND with the price of the item coming down from the high bollinger band exceeded is usually a good exit. Obviously, too stringent an exit test and you are pulled down from the trade maybe too soon and a lax exit test may get you out already near the lower bollinger band which may be a short term oversold.

    I find that using the count — is the closing price truly higher then 2 days ago and has this fact occured in a series at least 2 times out of 4 days in a row shows an uptrend because you are testing if the market is stronger than the recent past and you can give the market its time to pause and back and fill.

    For example — 1 being “yes” to the count test and 0 being “no” for the count test.

    Here are the possible favorable patterns for any 4 day series:
    1111, 1110, 1100, 0011, 0111, 0110, 1010, 0101, 1001.

    I also like the fact that any back to back 00 tests put the trader on high alert. And after a run of 111 for example with some price distance up the chart, the market can correct for a 00 without distubing the trade. The bollinger bands give you a visual that the raw counts do not–the strongest and sometimes most boring markets are ones that plod along in that high zone, never really falling down much from the high band. And sometimes a 1000 count series will occur but the trading is still in the high zone of the BB. Thats when I squirm because the count gave a sell but the BB is still in the good zone.

    ****

    I have combined these simple learnings from two giants and want to acknowledge Tom Demark and John Bollinger who influnced me very early on in my study of markets.

  6. AK on May 6, 2011 8:29 am

    How About:

    1. Grab a couple of your favorite historical datasets from Yahoo or Google Finance for example (making sure one is lagged for Predictive Power!)

    2. Use your data to populate the Eureqa machine set it and forget it.

    http://creativemachines.cornell.edu/eureqa

    3. Paper trade your new “system” until a feeling of invincibility overwhelms you.

    4. Profit!

  7. Nick Grosvenor on May 6, 2011 12:06 pm

    I like AK’s idea!

Archives

Resources & Links

Search