Data Mining with $NDXA50 Charts

My recent post looked at market extremes and charts depicting the percentage of stocks of the S&P 500 above their 50-day moving average.

We're approaching some market extremes, especially with the NASDAQ's recent tear.

Then, on Monday, the following post from Barry Ritholtz lodged itself in my head:

Peter Boockvar notes:
Today's consolidated volume for NYSE names was a
low for the year to date.

Helene Meisler notes:
It's a bear market -- and these low volume rallies
are typical of short covering. (unless you believe all those dark pools are
swallowing all the volume!).
As long as there are shorts in the market we
will get rallies; Once the shorts cover we get a decline like last week.

Doug Kass goes even further:
I am starting to scale into shorts
now. My favorite? PowerShares QQQ (QQQQ). My catalysts? The VIX and
volume.

That's our surprising data point of the day.

Kass was looking at shorting the NASDAQ 100 back on Monday. It's moved up almost 2%-3% since then. 83 components of the NASDAQ 100 are now trading above their 50-day moving averages. When that number hits 85 (or 15) on the $NDXA50 chart, is that a timely trading opportunity?

I didn't have an immediate answer, apart from Trader's Narrative's opinion that a low number of stocks trading above their 50-day moving average is a superior signal to its converse. So I put on my virtual spelunking hat, turned on that light on the front, and mined some data.

Here are the $NDXA50 charts from stockcharts.com:

How one defines extreme levels in these charts is somewhat arbitrary. At first, readings at or above 80, or at or below 20, seem like a good place to start. For me, that's where these charts begin to grab my attention, but if you look over the six-and-one-half years of information, you'll see that the 80/20 levels offer up too many data points. 80/20 is just not extreme enough.

85/15 are better extreme levels.

Here were my rules for this bit of mining:

I poured through daily $NDXA50 charts (not shown above) and noted the first trading day that the chart produced a value of 15 or below, or 85 or above. I recorded the price range for QQQQ shares for the following trading day, and then looked at the QQQQ trading range five, ten, 20, and 40 trading days into the future. Once the index was at an extreme level, I would not record a new trading day with a value of 15 or below, or 85 or above until the chart value returned to the normal range (16-84) for at least one trading day.

I also arbitrarily decided to start at 1/1/2004 until today.

Using these criteria, I found seven dates when $NDXA50 fell to 15 or below:

3/22/04
8/6/04
6/8/06
7/14/06
1/4/08
2/6/08
3/10/08

Six out of seven turned out to be profitable trades generally between five and ten trading days later. Three became significantly more profitable 20 and 40 trading days out. Only one trade required more than 40 trading days to turn a respectable profit (1/4/08).

Yes, this is a small data set, but the downside is quite limited if you take advantage of this trade by buying QQQQ shares. Buying QLD shares adds some leveraged juice to this trade, but also increases the downside risk. QQQQ call options are another interesting alternative, too, but as this data set implies that you need five to ten trading days for the trade to work, make sure that the options decay over that time doesn't cancel out the gains.

On the flip side, I found 14 discrete events when the $NDXA50 touched 85:

10/4/04
10/6/04
11/4/04
12/1/04
12/9/04
12/14/04
12/29/04
5/20/05
5/26/05
12/15/05
9/13/06
10/12/06
10/8/07

Here, the trading signal to go short on QQQQ was far from compelling. Six trades would have been successful, four would have failed, and four offered insignificant to small gains.

That's disappointing, as the NASDAQ is almost there, at 83. I agree with Kass, and may follow him in shorting the QQQQ's (I'll do it via adding to my QID position), but I'm not using the $NDXA50 data as a trigger. The data doesn't compel a trade.