S&P Strength and TRIN: Another Look at Efficiency

The S&P 500 (SPY) has approximately 2.2% in the past 3 days on an average TRIN reading of .69. I went back to March, 2003 (N = 759) and found 51 occasions in which we had a three-day period with a rise of more than 2%. The average three-day TRIN for those occasions was .80, indicating that we have had an above average concentration of volume in rising issues.

Six days following the three-day rise, SPY has averaged a gain of .70% (35 up, 16 down), much stronger than the three-day average gain for the overall sample of .36% (446 up, 313 down).

When we break the strong occasions in half based on three-day TRIN readings, however, a distinct pattern appears. When SPY is strong and the TRIN is lower than average (higher concentration of volume in rising issues), the next six days average a gain of only .03% (13 up, 12 down). When SPY is strong and the TRIN is higher than average, the next six days average a gain of 1.35% (22 up, 4 down).

Once again we can understand the results in terms of market efficiency. When it takes a greater concentration of volume in rising issues to achieve a particular rise, the market is less efficient and subsequently produces subnormal returns. When the market is more efficient--able to generate a given rise without a commensurate concentration in volume--returns following strength tend to continue the strength.

The results suggest that it may be difficult to generate the upside followthrough normally associated with a strong three-day gain.

TRIN and Market Efficiency

This will kick off a historical look at the TRIN (Arms Index). Today we had a gain of .19% in SPY and TRIN was .75. I went back to March, 2003 (N = 761) and looked at all one-day occasions in which TRIN was between .70 and .80 (N = 87). What we find by making TRIN the independent variable is that such a TRIN value can be associated with very different market outcomes. For example, in the sample, the one-day SPY readings associated with TRIN between .70 and .80 range from a loss of -.15% to a gain of 2.13%. The gain of .19% on Monday was definitely on the lower end of the spectrum. It's saying that, although volume was relatively concentrated in advancing stocks, such concentration could not generate much upside in the large cap market.

I divided the sample in half based on the SPY outcomes to see what happened the next day. When TRIN was between .70 and .80 and SPY was weak, the next day in SPY averaged a gain of .01% (21 up, 22 down). This was weaker than the average rise of .06% for the full sample. When TRIN was in the same range and SPY was strong, the next day in SPY averaged a gain of .08% (25 up, 19 down). It thus appears that when the market is relatively inefficient--a concentration of volume cannot generate much price gain--short run outcomes are weaker than if we see volume concentration associated with price strength.

I'll be investigating this idea of market efficiency further. For now, we'll count this one very modestly in the bear's column.

Trendiness on a Weekly Basis: Unexpected Findings

Here is a weekly view of the issue of trending markets using S&P 500 and NASDAQ 100 data. (By the way, a different perspective is posted to the Trader Performance page on my personal site). I went back to January, 1997 (N = 479) and just looked at weekly closing prices. Once again, I focused on occasions in which either a rise was followed by a rise or a decline by a decline.

In all, for the S&P, there were 239 occasions in which we saw a two-week trend and 240 occasions in which rises were followed by declines or vice versa. This is exactly what we'd expect by chance. In the NASDAQ, we had 246 trending occasions and 233 non-trending two-week instances. (Interestingly, the NASDAQ also showed up on my Trader Performance analysis as performing better vis a vis momentum/trend trading).

Looking only at 2005/6 data (N = 62), we had 31 trending two-week periods in the S&P and 31 non-trending ones. In the NASDAQ, we had 33 trending periods and 29 non-trending ones. Interestingly, we see neither evidence of persistence (trending) or anti-persistence (reversal) in the weekly data--even the most recent weekly readings.

What this says to me is that the market's loss of trendiness is occurring more at shorter time frames than at longer ones. My next Trading Markets article will address this.

Will Very Short-Term Momentum Trading of the S&P Work?

As we've seen from the recent Trading Markets article and yesterday's blog posting, the S&P 500 Index has been losing its trending properties across multiple timeframes, from daily through intraday. Anecdotal evidence suggests that scalpers are suffering in this market, as well. Let's see how the market is trending on their time frame. Please note that when I talk about scalpers, I am referring to liquidity providers in the electronic marketplace. These are participants who are in the market most of the time, working bids and offers and attempting to extract small, frequent profits from very short-term movement.

I went back to January, 2004 and looked at all five-minute periods in the ES contract (N = 44,110). As before, I calculated all instances in which the market was either up following a five-minute rise or down following a five-minute decline. What we see is that only 14,362 of the periods displayed such two-period trends. This proportion is *much* worse than what we saw in the daily or even the hourly data. In essence, it's saying that the odds of the market rising five minutes after a five minute rise (or declining after a five-minute decline) are worse than one in three.

The reason for this is that a number of five-minute periods close unchanged. The low volatility conspires to restrain very short-term trending behavior. This makes momentum trading near impossible. Either one must extend one's holding period well beyond five minutes--in which case you're really no longer a scalper--or one be willing to fade any movement whatsoever, risking those one in three occasions when the market can run you over.

Is there any hope for momentum and trend traders in different instruments? This will be the upcoming focus.

Does The Market Trend on an Intraday Basis?

I'm getting quite a few positive comments on my recent Trading Markets article that documents the decline in trending behavior in the S&P 500 Index over the past 40 years. What has been eye-opening, however, is that this loss of trendiness has occurred across all time frames that I've investigated thus far.

Here's an example. I took hourly readings of SPY since December 13, 2005 (N = 478). I once again looked for all occasions in which a rise was followed by a rise and a decline by a decline. If the odds of a rise or decline are 50/50, we should see half of all occasions by chance result in either two consecutive rises or two consecutive declines.

In fact, we see 225 occasions where either a rise was followed by a rise or a decline followed by a decline and 253 occasions where rises were followed by declines or declines by rises. During March alone (N = 64), we've seen 28 occasions where rises were followed by rises or declines by declines and 36 occasions where there was no continuation of a move.

Once again, we not only see an absence of trending--failure of rises to be followed by rises and declines by declines--but actually evidence of antipersistence. On average, rises are being followed by declines and vice versa. This is wreaking havoc with momentum traders in the ES and SPY markets.

Next I'll look at other indices and sectors and see where there might be opportunity for trend and momentum traders.

Small Cap Stocks: Five-Day Weakness and What Comes Next

It's been a rough week for the small caps, as the Russell 2000 ETF (IWM) is down over 3% on a five-day basis. I went back to March, 2003 (N = 757) and looked for all occasions where the Russell was down comparably and what has happened next.

In all, I found 58 occasions in which IWM was down 3% or more in a five-day period. Five days later, the S&P 500 (SPY) was up by an average .72% (36 up, 22 down) and IWM was up by an average (1.13%) 35 up, 23 down. Both are considerably stronger than the average five-day gain for SPY (.30%; 442 up, 315 down) and IWM (.50%; 445 up, 312 down). This will have me looking for upside setups today and Monday.

Going From Trade Ideas to Profitable Trades

Before the open tomorrow, I will post an analysis of market expectations based on events among the secondary stocks. That should prove informative, as we have interesting five-day patterns.

Here, though, I want to follow up on my post earlier today. The market provided a great example of how trade ideas developed through historical analysis are just that: trade ideas. They need to be confirmed by real time market action to become viable trades.

One of the ways I'll look at a market intraday is to scan for expectable events and setups for those events. An example would be a move back to the average trading price for the current or previous day. I know, based on research, that the market will return to its average price 3/4 of the time--and more often if volume/volatility are low. Knowing this, I will then look for a setup--a real-time event--to confirm for me that this historical tendency is likely to occur. Thus, for instance, I'll see that the ES has moved to the upper end of its range, but other indices haven't. Then I'll see volume lifting offers drying up. That will trigger my trade for a move back to the day's average price.

Today I was looking for a different expectable event: A boost in the NYSE TICK to 1000 or greater. Before acting on that, I needed a setup: Some real-time event to confirm for me that buyers were gaining the upper hand. Not only didn't we get the setup, we got the reverse: Bond yields after 10:30 AM CT rose and the market sold off, taking the TICK lower. At that point, our Cumulative Adjusted TICK began making daily lows and it was clear that traders were hitting bids, not taking offers. With interest rates again weighing on stocks and short-term sentiment negative--as shown by the negative TICK--the trade idea gained no validation.

All of this raises an important point: Coming up with good trade ideas is simple. The difficult part is knowing whether and when to act upon them. In my own trading, I need an idea, a setup, and then a framework for managing the trade once it's on. There is much more to trading competence than coming up with good ideas--and today provided a nice demonstration of that.

NYSE TICK and Descriptive Statistics

I'll amplify on this tonight: Sometimes descriptive statistics alone are of value in trading. For example, as I write at approximately 10:30 AM CT, we have not yet had an NYSE TICK reading of +1000 or greater. Since the start of 2005 (N = 296), there have been 262 days that have seen such elevated readings and only 34 that have not. I'm willing to bet that we will have just such an elevated spike sometime today, and I'll look for the setup. Then I'll trade the instrument best positioned to take advantage of that spike, which--right now--looks like SMH, the very instrument highlighted in yesterday's entry. Not something I'd bet the farm on, but a thought going through the head right now...

More SOX and Stocks

Keep your eyes on the Trading Markets site this weekend. I have a historical analysis scheduled for publication that examines the stock market's trending behavior over a 40 year period. It's eye opening. I'll have a very brief summary on the Trading Psychology Weblog tonight.

I thought I would update some of the modeling with the semiconductor stocks (SMH), given that we're down more than 5% over the past four trading sessions. One wrinkle I'm adding to the analysis is that I'm examining outcomes across three instruments: SMH, QQQQ, and SPY. This addresses the theme I've been touching upon lately of maximizing the instrument that you trade as well as the timing of trades.

Since March, 2003 (N = 753), we have had 53 days in which SMH has been down more than 5% over a four-day period. Four days later, here's how the outcomes looked:

• SPY: Average gain = .79% (38 up, 15 down). Average four-day gain for sample overall = .24% (435 up, 318 down).
• QQQQ: Average gain = 1.21% (36 up, 17 down). Average four-day gain for sample overall = .31% (421 up, 332 down).
• SMH: Average gain = 1.80% (35 up, 18 down). Average four-day gain for sample overall = .34 (400 up, 353 down).

What we can see is that there are distinctly positive outcomes four days out across all indices. When SMH is very weak over a four-day period, the next four days have been bullish on average. Of the three ETFs, SMH has milked this pattern the most, more than doubling the average gain in SPY. It thus appears that the greatest edge is not only trading to the long side over this swing period, but also trading the very instrument that has been weakest. Let's follow up on this shortly.

Two Days of Strong Downside Momentum: What Next?

The past two sessions, we have had broad downside momentum in the stock market, as measured by the Demand/Supply Index monitored on the Trading Psychology Weblog. I went back to March, 2003 (N = 754) and looked for similar periods in which the two-day Supply exceeded 270 (N = 15). Two days later, the market (SPY) was up by an average of .18 (10 up, 5 down), modestly stronger than the average two-day gain of .12% for the sample overall. One week (five days) later, however, the market was up by an average of only .02% (7 up, 8 down)--much weaker than the average five-day gain of .30% over the entire period (442 up, 312 down) overall. A bounce following strong downside momentum followed by subnormal performance thus seems to be the norm.

The recent market has been hostage to bond/interest rate movements, however, so these may well be calling the shots in the near term. I'm also noticing a lead-lag relationship between the DAX and the S&P, with strength in the former leading large cap strength yesterday and weakness leading overnight weakness in the ES.

Strong Dow, Weak Speculative Stocks

Hope you've been able to profit from the trend/momentum and interest rate/stock relationships we've been seeing. It's been a bit tricky if you've been trading the large cap indices such as the S&P 500, because there has been a huge divergence in performance between those large caps (stronger) and the secondary issues (weaker).

I noticed, for example, that the Dow Industrials (DIA) were actually up on the day, while my basket of Speculative Stocks was down by over 1.5%. That is quite a gap. Going back to January, 2003 (N = 797), I could only find eight occasions in which the Spec stocks were down by more than 1%, but the Dow was up. Three days later, the Dow was up seven of those eight occasions by an average of .91%--much stronger than the average three-day change of .10% (435 up, 362 down) for the sample overall. It appears that, when the Spec stocks have been down sharply, they--and the rest of the market--tend to rebound three days out.

Interestingly, if we just look at occasions in which the Dow is up, but Spec stocks are down (N = 60), the next day the Dow is down by an average of -.13% (26 up, 34 down) and the Spec stocks are down even more. This fits with other research I have done: When Spec stocks underperform the Non-Spec issues, the next day sees some follow through in weakness. A reasonable scenario for the market, then, might be weakness tomorrow followed by a rebound.

Tomorrow AM, I will post a different analysis, tracking market performance after several consecutive down days.

High TRIN and Next Day Open - Quick Note

Quick note on TRIN (NYSE Arms Index) that I will follow up later today. In response to the reader's question, when TRIN is 2.0 or greater and SPY has fallen by .75% or more (N = 37) since March, 2003 (N = 758), the market opens the next day up by an average .17% (28 up, 9 down), much stronger than the average overnight gain of .04% (415 up, 343 down). This is further evidence that downside momentum from the previous day has not tended to carry over to early the next day. More to come...

Big Down Day: What Happens at the Open?

Quick note before Tuesday's open. A reader asked me what happens at the open after the market has moved down sharply. I looked at SPY from March, 2003 to the present (N = 758) and found 106 instances in which the market was down .75% or more from open to close. The next day's open averaged a gain of .09% (72 up, 34 down)--considerably stronger than the average overnight move of .04% (415 up, 343 down). There is thus no evidence that weak markets during the day session spill over to the overnight and, in fact, we see a modest bounce more often than not in the overnight. Given the tendency of strong downside momentum markets to continue weakness in the short run, this may set up a trade of selling strength early the next day.

Thanks, BTW, for the kind comments sent to me re: the recent Trading Markets article.

Interest Rate Rises: What They Mean for Stocks--and Bonds

On 3/2/06, I wrote about the possibility of changing cycles vis a vis interest rates and stocks based on recent shifts in historical patterns. Today we saw more evidence of this, as stocks moved to multi-day lows and rates moved to multi-day highs. Specifically, we are down about 1% over the last five days in SPY and up about 3.22% in the 10-year Note interest rate ($TNX). I decided to take a longer look and went back to April 2000 (N = 1480) to see what happens when we get a rise in rates of over 2.5% in a five day period.

I found 251 such occasions. Ten days later, interest rates had risen further by an average of .61% (135 up, 116 down) and stocks had fallen further, by an average of -.11% (121 up, 130 down). Rates were stronger than the average change for the sample overall (-.07%; 639 up, 841 down). Stocks were weaker than the average change for the sample (-.03%; 762 up, 718 down).

I then looked at the results from 2003 to the present (N = 155). Ten days after the five-day rise in rates, stocks were down by an average of -.19% (74 up, 81 down) and rates were up by 1.10% (87 up, 68 down).

In short, a strong weekly rise in interest rates has been associated with further bond weakness (rate rises) and further stock weakness. We'll have to count this as one for the bears going forward.

Midday Price Patterns

After looking at opening and closing hours, I decided to split the difference and investigate midday hours: the time in the market from the end of the first hour of trading to the start of the last hour. Going back to March, 2003 (N = 757) in the Dow Industrials, I find 78 occasions in which we had a gain of .50% or greater during the midday hours. The average change to the same day's close was .09% (46 up, 32 down), modestly stronger than the average gain of .01% (403 up, 354 down) for the sample overall.

We had 89 occasions in which the midday hours lost .50% or more. I did not find a meaningful edge by the close of the same day. By the close the following day, however, the Dow averaged a gain of .11% (50 up, 39 down), stronger than the average gain of .06% (393 up, 364 down) for the sample overall.

My overall impression is that intraday traders, like the protagonist in Neil Young's "The Needle and the Damage Done", are milking blood to keep from running out. While there are some historical patterns in intraday markets and over very short time frames, they are modest compared with those that we've seen over swing periods. I'm not sure that's well appreciated among active traders.

Last Hour of Trading: What It Means

After looking at the opens recently, I decided to investigate market performance following strong vs. weak closes. I went back to March, 2003 in the Dow Industrials and examined the last hour of trading and its effect upon the next day.

Overall for the sample, the average daily change in the Dow was .05% (401 up, 356 down). When the last hour of trading was up by .40% or more (N = 82), the market the next day averaged a loss of -.08% (40 up, 42 down). When the last hour of trading was down by .40% or more (N = 80), the market averaged a gain of .20% (48 up, 32 down).

We thus see modest evidence of reversal following weak and strong last hours of trade. As with market opens, we are seeing little evidence that strong or weak closes carry over to the next day of trading. If anything, such moves are more likely to produce retracement than continuation.

Weak Opens: What Comes Next

When the market opened lower today, a couple of traders asked me, "How low do you think we could go?" As Ayn Rand would say, "Check your premises!" Does a down open usually lead to a down trading day?

I went back to March, 2003 (N = 757) and found 42 days in which we opened down more than half a percent in SPY (as we did today). From the open to the close, the market was up 22 times, down 20, for an average gain of .18%. That is better than the average open-to-close change of .02% (403 up, 354 down) for the sample overall.

When, however, the down open follows a day's trading session (previous open to close) that is weak, the average change from open to close is .32% (11 up, 10 down). When the down open follows a day's trading session that is strong, the average change from open to close is .03% (11 up, 10 down).

A down open after the previous day is weak is thus not more likely to be strong, but its gains are larger than its losses. Under no circumstances, however, could I find evidence that a down open produces subnormal market performance from open to close. It pays to check those premises!

Rising Gold, Rising Rates: Evidence of Changing Cycles?

The past two days we've seen interest rates on the 10-year Note rise by about 2% and gold stocks ($XAU) rise by over 4%. So I decided to take a look at what happens after two-day periods in which both rates ($TNX) and gold stocks rise by 2% or more.

Since March, 2003 (N = 749), we've had 32 days that meet the 2% criteria. Interestingly, four days later we see an average change in SPY of .70% (23 up, 9 down). That's quite a bit more bullish than the average four-day change for the sample of .25% (433 up, 316 down). We have to count that one for the bulls going forward.

One would think that an inflationary environment (rising gold, rising interest rates) might weigh on stocks. At least in the near term since 2003, that hasn't been the case. If we started seeing that pattern emerge, it would represent a key market shift. We've had 10 instances of 2% rises in rates and gold since 2005 and 6 have resulted in positive change in SPY four days later (average change = .22%). Thus the bullish pattern really isn't manifesting itself recently. This is worth keeping an eye on, as we might be seeing market sentiment re: inflation changing before our eyes.

Up Day, Down Open: Real-Time Update

At the time I'm writing this (7:56 AM CT), we're looking at a down open in SPY after an up day yesterday. I found 123 days since March, 2003 (N = 755) in which we were up more than .75%. When the market opened down the next day, the average change from open to close was .15% (26 up, 23 down). When the market opened up the next day, the average change from open to close was -.08% (35 up, 38 down). It thus appears that a down open does not necessarily mean that the day has a bearish cast after an up day and, in fact, may even be slightly bullish.

When the market opens, I then look to see which sectors are strong and weak and conduct lead-lag analyses to further update forecasts. This is the essence of dynamic modeling, as opposed to trading fixed models.

----------------------------------

9:54 AM (CT) Update - Notice how the semiconductor strength and then the breakout in the NYSE TICK with the buy programs in the Russell stocks preceded the S&P 500 move above its open. That made buying weakness in ES per the analysis above easier, knowing that these are strong lead relationships. I hope this opens readers' minds to different ways of looking at real time market action.

Up Open After Down Day: Sequence Analysis in Action

This is an important blog entry; you may want to review the posting from 2/26 (Sequence Analysis) before reading what I have here. The following is an example of sequence analysis at work.

Let's set the stage. After an up day on Monday which made a five-day high, we sold off on Tuesday and made a five-day low. Several of my analyses suggested a high likelihood of a down day today.

In sequence analysis, you always update forecasts with the most recently available data. This morning, the market opened up by more than a quarter of a percent. The updated forecast thus asked the question: What has happened historically when a down day has been followed by an up open? Specifically, I looked at occasions in which the market was down more than .75% on the day (N = 118) since March, 2003 (N = 755) and then divided the sample in half based on the following day's open.

When the next day's open is strong (as was the case today; N = 59), the average move from the open to the close has been .19% (38 up, 21 down). When the next day's open is weak (N = 59), the average move from the open to the close has been -.01% (33 up, 26 down).

Looking a bit further out, when the market opens strong after the down day, the move to the *following day's* close has been .41% (36 up, 23 down). When the market opens weak after the down day, the move to the *following day's* close has been -.06% (29 up, 30 down).

In short, a strong open following a weak day affects the market's short-term trajectory. This is an example of how updating forecasts with real time data can greatly aid trading. Traders need not wait for real time events to occur, however, to conduct these analyses. They can prepare for real time possibilities by conducting "what-if" scenarios with the historical data. Such sequence analysis would look at all historical occasions of market declines such as Tuesday's and then investigate what happened when the following day opened strong.

This captures the difference between mechanical system traders and historical pattern traders. Mechanical traders trade a model. Historical pattern traders conduct ongoing modeling as a dynamic process. Very, very few traders understand this and appreciate its potential.