* Ron Griess of The Chart Store provides even more color on the history of the Dark Cross:*

>

The Financial Press has been full of stories about so called “Dark Crosses.”

We did some analysis on our S&P Composite data from 1930 and present the following two tables.

The first table shows the performance of the S&P Composite for the time periods listed when the 50 day moving average is falling and crosses the 200 day moving average while the 200 day moving average is still rising.

The second table shows the performance of the S&P Composite for the time periods listed when the 50 day moving average is falling and crosses the 200 day moving average and the 200 day moving average is also falling.

>

### 200 day moving average is still rising

>

### 200 day moving average is falling

Category: Technical Analysis

Please use the comments to demonstrate your own ignorance, unfamiliarity with empirical data and lack of respect for scientific knowledge. Be sure to create straw men and argue against things I have neither said nor implied. If you could repeat previously discredited memes or steer the conversation into irrelevant, off topic discussions, it would be appreciated. Lastly, kindly forgo all civility in your discourse . . . you are, after all, anonymous.

Very good…thank you.

What I get out of this is: the market could go down, or it could go up.

I’d love to see the same breakout but just for 1-6 mo, 1 yr returns after the 200 ma turns negative (as it did yesterday), ignore the dark cross or not.

Currently the 200-day moving average appears close to be going sideways — going neither up nor down to any significant degree. Also, the arbitrary time periods following a signal (1 month, 2 months, etc.) ignore that a contrary “Golden Cross” buy signal might have intervened in the interim. For example, in 1933, obviously the 50-day SMA crossed back above the 200-day SMA sometime well before the trader who took the “Dark Cross” sell signal lost 83.52% of his stake.

Seems pretty inconclusive.

Regards,

TDL

@seneca Says:

“Also, the arbitrary time periods following a signal (1 month, 2 months, etc.) ignore that a contrary “Golden Cross” buy signal might have intervened in the interim. ”

Make sense to me. Any way we could mark the moment when there was a Golden Cross?

I color printed these two tables, cut out each cell, and put all the little pieces in a cup. I shook the cup and dumped the pieces on the table.

The resulting pile of red and green bits vaguely looks like the ghostly image on the Shroud of Turin. So I think the market might go down. No up… :)

The 200 day MA is pretty close to dead level these days.

Telling the future from the past is not a terribly reliable practice, as the present is composed of the confluence of a bazillion different things, all moving in different directions.

Yet this is the basis for the entirety of Technical Analysis — the divination of an uncertain future from the “patterns” formed by a tiny, tiny subset of the things that make up the Present.

Have at it, fellas. Best of luck.

This sis not to imply that Fundamental Analysis has any greater accuracy.

I’d like to see a double-blind study of monies made using fundamental analysis, technical analysis, and the Magic 8-ball. I’ll bet there is not a significant difference between any of them.

Economic forecasting is like a weather forecast or reading from an almanac. Its interesting stuff and ultimately proves you were on the right track or not.

This is the beginning of hurricane season….thus hurricanes are likely to form from now until the end of the summer. I cant tell you exactly what day they will form but there will be several of them throughout the summer. I cant tell you what category, but they can range.

constantnormal,

Then why bother reading financial blogs? After all if markets are so efficient, you are not maximize your productive time by consuming information that has already been priced in.

Regards,

TDL

Nice rebound today.

I think all of this type of information is useful for anyone who thinks they have it all figured out.

It clearly illustrates that we have markets which move, in all directions, and that to be an investor in those markets and make money, not lose, one must be very nimble and agile. As well as involved. This is a story that can never be told to 401k participants, because they are taught to “buy and hold”. A true investor knows that markets can move quickly and violently, and that you’ve got to be ready to move with them. I’ll even go as far as to say that the traditional question of investment time frames or horizons is a notion that bears far less significance to most investors than perhaps they have been told. The market doesn’t care what your time horizon is; it will do what it is going to do. You best adjust your investments and how you react at different times.

Throw some fundamental analysis on top of those dates and then we’re getting somewhere…

Well, S&P breached well below 1038 today hitting 1028 before closing at 1030-1031 with volume ROC spiking. We heading down to 850-865, or do we need to see a follow through tomorrow? Where were the HFTs to force it back above 1040 at all costs in the last 30min of the day?

@TDL — I never said the markets were efficient, only that we cannot hope to predict them over the short run with any accuracy. Chaotic would be a somewhat better description, I think. Over the longer term, one can make the (unproven) assertion that over time, the markets will tend to reflect the economy they represent (via their proxies, corporations), and thus expanding economies have better stock market opportunities than contracting economies.

One can take a gander around the economies of the world and confirm this pretty easily. This is the basis for index funds, as well, but the purveyors of those funds never point this out, as they would then be advising their stockholders to switch to bonds/cash during significant market downturns — although I like to think I can try to do better than the mean by omitting those stocks from my portfolio that are almost certainly going to do poorly over the long haul. That is, if I had enough in the game to be able to afford a portfolio made up of a few hundred stocks, instead of the handful I must satisfy myself with.

It is in seeking out the obvious (or not-so-obvious) inefficiencies in the markets that one can hope to do better than the mean, and a part of that lies in investigating the mental state of the markets, the beliefs and opinions of the players within (at least the ones that are seeking to figure it all out, and not those who have a dogma-driven methodology they ascribe to).

But the real reason I participate in the financial blogs is that 1) I get informed about what others are thinking (not infrequently I even manage to learn something), and 2) it’s fun.

constantnormal,

Fair enough and thank you for a rather concise and articulate view of the markets.

Regards,

TDL

So is this statistically significant? If anything, this data seems to imply that dark crosses have very little if any correlation with subsequent market moves. Has anyone done the statistical calculations?

@Rescission 3:25 pm — “Nice rebound today.”

I just got in from mowing the lawn … and wtf? A drop-off in the last half-hour is a whole lot different picture than I think anyone expected after yesterday’s last-10-minutes rally.

So, was your comment sarcasm, or did you jinx the markets?

I’m seeking explanations for this, in the light of yesterday’s late rally. This IS the last day of the quarter, correct?

Are there not a bunch of fund managers eager to see their portfolios up on the last day of the quarter? Or are those same portfolio managers all rushing to get into cash so that they can show they were safely in cash at the end of this quarter?

Explanations/theories/humor?

last week David Rosenberg said “DOW 5000, GOLD 5000″.

And, if you crunch all the above data in ounces of gold instead of federal reserve notes, it looks much much worse.

@foxyrabbit :-)

By statistical calculations, you could do a chi-square, or a binomial distribution with pairs of before, after.

If you do a chi-square you need deal with actual values, whereas if we do binomial calculations, we can quantify everything based on only positive and negative, gain and loss.

If we do the pairs of before, after, this means that the occurrence of a gain or a loss is not any better that what we would expect if the results were random — ie., a gain or a loss is equally likely. Since their are only 2 choices, each 50%, we call this a Binomial distribution.

Now just like flipping a coin, their could randomly occur more heads than tails if you flip it 20 times, say 12 heads and 8 tails. You wouldn’t consider the coin to be overly biased towards heads. If however, you got 19 heads and 1 tail, then you would find it very hard to believe the coin is not biased. You can use the Binomial probability model to calculate what the expected percent of occurrences this would be if each order of 20 tosses is equally likely.

If you look at the data, each column has green positive numbers (heads) and red negative numbers (tails). Most columns have roughly the same number of greens and reds, except the 1 month and 1 year in the 2nd table (when both 50day and 200 day are going down). Both of these have 5 versus 12.

The probability that 5 versus 12 would occur randomly is 20%.

Which means the potential correlation between “dark crosses” and gain or loss is not better than randomness.

Someone pointed out that the intervals are too cumbersome and might overlap other “dark cross” events. This would not change the analysis because any subset of the sample space would have the same overall systematic properties. For instance, if we refined the 20 tosses of a coin example and only looked at a sequence which followed a previous toss of heads (or followed by two prior tosses of heads, or a heads-tails-heads on the prior 3 tosses, or et cetera). The probability of randomness would not be affected, nor would the potential dependent relationship be concealled, because (returning to the coin example) if the coin was biased to pick heads (19 to 1) than you would see a lot of tossing heads no matter how you refined the 20 tosses schema.

If you look at the top chart however, recent experience (from rows 21 thru 28) would lead you to believe that positive greens always occurred. Which makes my second point. Subsets of sample spaces can give you a false sense of dependence.

If you want to do a chi-square, you could use an average of zero as the expected value. The numbers in the columns are then the observed values. For each cell do (observed – expected)^2 divided by the expected and sum up the results. For the top table, the degrees of freedom is (28-1)*(5-1) = 27*4 = 108. For the top table, the degrees of freedom is (17-1)*(5-1) = 16*4 = 64.

A rough rule of thumb, if the sum of the cells is more than the number of cells then the results are not due to randomness and percentages of each column are not consistent from column to column, ie. the rows are not in the same ratio (for example, 3:2:1 = 6:4:2 = 12:8:4).

I should have proofread.

If the expected value is zero, you obviously can’t divide by zero.

You need to convert the data cells into frequencies, which is what a percentage is, ie. 42% means 42 times out of 100.

In this case, the percentages mean a portion of a monetary value, which isn’t the same thing as a frequency of occurrence. Also, the percentages in the cells are not cumulative, in that if you added the total for each row, or each column, or even the entire table, the sums will not equal 100%. That is because the percents are due to portions of value at a time moment when the event “dark cross.”

Hence, this data set is a time series, not a sample space of a population. So you can’t do a chi-square analysis, unless you measure these observed data points against a model of data points.

You can do this, but the math is actually useless because the parameters are too inconsistent. For example,if you measure this data set against a Poisson distribution, you need an average gain/loss. Look at the data above and see if you think an “average” of anything is very helpful.

Just because I have to be somewhat thorough.

You could also do a chi-square and use the current rate of interest as the expected value. If you do this, then the negative percentages will weigh more than the small positive gains because, assuming the rate of interest is 1% … then (3% – 1%)^2 = 4 but (-3%-1%)^2 = 16 . (^2 means raised to the second power, or squared, or multiplied by itself).

Which is actually useless information, because quite simply … you can’t predict the future of complex interactions based upon the past. Each interval of time through which we perceive events is …

1) subject to imperfect data collection (which is usually changed at a later date) ;

2) based upon a scale which can only use the past as a guide; and,

3) subject to complex interactions and random events that cannot be understand until they are analyzed and understand many intervals of time AFTER they occur.