Another Week, Another Record.  Again.

11/14/08 4:30 PM EST

As of: SPX 1045 HELP  ARCHIVE

It seems every week brings a new record.  Last week, we got those double-whammy down days that hadn't been seen in a generation, and this week we got a new record low in the NYSE TICK.

On Monday, we got a large gap up open to start the week.  I generally detest seeing gaps in the direction of my trade, but I couldn't find much of a reason to distrust the one we got.  That was unfortunate, as it failed miserably.

By Tuesday, the selling just kept coming, which was unusual given the studies we looked at last week.  We were simply not getting any upside follow-through from a setup that should have triggered some.  It was enough to stop me out of a modest intermediate-term long position I was carrying at the time.

That led to Wednesday's session, which was the ugliest one yet.  Breadth measurements showed exceptionally lopsided selling pressure, which was bordering on the "so bad it's gotta be good" side.  That morning, we looked at a handful of short-term guides that weren't as oversold as they had been in October, bringing up the potential for a price/sentiment divergence if the S&P broke its October low.  If so, it would look very much like March 2003 in that respect.

We got that break on Thursday.  That morning, it seemed like the best chance for the bulls would be a drop below the October low that triggered the sell stop orders that were surely resting just underneath, then a reversal back above.  We got that very thing, and no question what we saw was stops being run - the TICK on the NYSE set a new record low at 1:00 on Thursday, by a huge margin.  About the only way that's going to happen is if stop orders were hit.

We got the reversal too, and how.  The indices put in one of their largest three-hour ranges ever, spiking us right back up into the trading range of the past month.  That formed a picture-perfect reversal bar.

This morning, we looked at prior such Key Reveral Bars over the past 44 years in the S&P 500.  They have had good success in the past at identifying intermediate-term lows (really only one failure out of nine occurrences prior to this year), however there were two of them in September and they both failed miserably.

Some have suggested that in order for it to be a "true" reversal bar, it should also be an outside day, meaning that the day's low is lower than the previous day, and the close is higher than the previous day's high.  If we require that parameter, too, then we're only left with 3 instances in the S&P:  06/18/84, 07/24/02 and 01/23/08.

All of those were good for intermediate-term rallies, but it's hard to rely on just three instances.  So I ran the same test, except this time instead of using the S&P 500 index itself, I used the component stocks.  Since 1996, there have been 384 of these reversals, and it led to three-month gains 56% of the time, averaging +3.8%, and with an average maximum gain that was 35% larger than the average maximum loss.  That's good, but not great by any means.

Bottom line, yesterday's reversal was impressive, and many factors suggest it is a sign that the worst is probably over (i.e. seasonality, a multitude of historic oversold readings in mid-October, and subsequent signs of explosive buying interest - two 7% rallies within a month of each other, an exceptionally low Arms Index reading and the quick flip in breadth).

This morning I noted that it all seemed too perfect, that rarely the market will come out of its slumber in such a clean, classic fashion.  It's too easy.  But objectively the evidence remains positive over the next one to three months, so while I wasn't ready to rush in and buy (not after the disappointing behavior earlier this week), I was still looking to get back in on the long side and try once again for that intermediate-term rally.  One tranche of our buy stop orders was hit right in the closing minutes today, so we've dipped a tiny toe back in, and we'll just have to see if we can continue to hold and make progress above 850ish on the S&P.

Have a save and relaxing weekend and we'll see you next week.

Key Reversal Day, Tested

11/14/08 9:25 AM EST

Good Friday morning...We begin the day with some weakness in the pre-market futures.  We were recovering a bit before the Retail Sales number, but the worst reading ever recording in that series helped put a damper on that.

Anyone who's read this site for awhile probably knows that I don't put any faith in some common technical myths, such as "accumulation" and "distribution" days.  They just don't hold up in testing, at least for the broader equity indices.

Another similar question I'm hearing a lot concerns yesterday's "key reversal" day.  This is another common technical concept, though the definition differs depending upon whom you ask.  The one I'm getting most frequently uses these terms:

The Key Reversal Day (KRD):

*  Is the latest in a decline of at least several days.

*  Sets a new low sometime during the day.

*  Rallies to a substantial positive close (relative to yesterday).

*  Closes in the upper half of its intraday range for the day.

*  Shows an increase in volume.

There is some gray area in that definition, but I went back to 1965 looking for any time the S&P 500 hit a new 52-week low after having declined at least 2% over the prior week, then rallied on at least 10% greater volume to a 1% or greater positive close, closing in the upper half of its range.  Some of those criteria are subjective, but I think they're reasonable and fit the spirit of the definition.

By this definition, there have been 11 KRDs since 1965.  The next day was mixed, up about half the time.  Over the next week, the S&P was positive 8 of the 11 times, averaging +1.1%.  The risk/reward wasn't great, though - the average maximum loss during those weeks was -2.6%, compared to an average maximum gain of +3.9%.  Positive, but not exceptional.

These usually aren't meant as short-term signals, they are more appropriate for the intermediate-term.  In that case, a month later the S&P was up 5 times out of the 11, and the average return was -2.2%.  The maximum loss averaged a sickening -10.0% compared to an average max gain of +5.9%.  Not good at all.

If we look out three months, it gets a little better, with a caveat.  Then the S&P was up 8 out of 9 times with an average of +4.6%.  There were only 9 occurrences because the last 2 haven't completed yet.  We saw a KRD on September 16 and September 18th of this year, and the S&P is down more than 25% since those occurrences.

Thus the caveat - these signals have been relatively successful in an intermediate-term time frame with really only one exception prior to this year (in April 1970), but the ones we've seen this year did not lead to a low.  For those curious, here are the dates:  10/10/66, 04/29/70, 09/16/74, 03/09/82, 06/18/84, 09/28/90, 03/13/01, 07/24/02, 01/23/08, 09/16/08 and 09/18/08.

I've written a bit about seasonality over the past couple of weeks, in terms of October sell-offs leading to poor early November returns, before more of an upside kick into the end of the year.  For the short-term, there is an odd little dip in seasonality in the very middle of November, though - if you look at the seasonality charts for the S&P, NDX and DJIA on the Daily Overview page, you can see the negative red bars on the 11th and 12th trading day of the month.

I take these biases with a grain of salt, but it has been pretty consistent to see a bit of weakness around here.  For some reason, it really kicked in from 1981 onward - since that year, the S&P 500 has shown a positive return over the next two days only 29% of the time, averaging -0.5%.  After that, though, it was up 81% of the time through the first few days of the new year, averaging +3.0%.  That didn't suffer much at all even when we include the 1970's, with the same average return and a winning percentage of 79%.

Yesterday before the open, it seemed as though the best bet for the bulls would be a drop below the October low that washed out the sell stop orders surely resting just underneath, then a rally back above the breakdown area.  We got that exact scenario, and it was perfectly clear that stops were washed out.  The NYSE TICK hit a new record low, by far, at 1:00pm, telling us that a whole lot of sell orders were executed at the same time.

The reversal was obviously impressive, and many factors suggest it is a sign that the worst is probably over.  Over the past month, we've looked at a number of positives including seasonality, a multitude of historic oversold readings in mid-October and subsequent signs of explosive buying interest - two 7% rallies within a month of each other, an exceptionally low Arms Index reading and the quick flip in breadth.  Now we have a classic washout below a prior panic low, and a major upside reversal.

It seems too perfect, but objectively the evidence remains positive over the next one to three months.  I'm not ready to rush in and buy, not after the disappointing behavior earlier this week, so I'm still holding back here.  I think it's more likely we have some backing-and-filling to do, but as long as we can hold over 850, I'll be looking to add back some intermediate-term long exposure that I was stopped out of early this week.