August 31, 2005 - Brett Steenbarger



August 31, 2005

More on intermarket connections and their relevance for trading.  Crucial for trading the last couple of days.  I've been posting intermarket info to the TradingMarkets site recently.

The Importance of Micropsychology:  One bullish factor that might have impacted Monday's session was sentiment, as measured by the equity-based put/call ratio.  From January, 2003 through August, 2005 (N = 666), we have had 129 days in which the ratio was greater than .80.  Three days later, the market (SPY) was up by .53% (84 up, 45 down), outpacing the sample as a whole (.14%; 386 up, 280 down).  When the put/call ratio for stocks was less than .60 (N = 146), the next three days' price change averaged only .02% (86 up, 80 down).  It appears that this effect is further magnified by total option volume.  When option volume is above average (signifying high speculative interest), returns 3 days out are significantly lower than when option volume is below average.  I will be monitoring this for yet another addition to the Modeler.

Stocks took an early tumble on Tuesday, retracing their previous day's gains.  This returned us to a neutral short-term trend, as we closed near the day's average price of ES 1206.5.  New lows were muted among the large caps, and a plurality of stocks traded in short-term downtrends through the day.  The adjusted TICK ended at -32; the Institutional Composite reflected continued weakness in large caps, ending at -283.  Demand fell to 65; Supply was 52.  New 20 and 65 day highs actually rose slightly to 494 and 301; new 20 and 65 day lows dropped to 947 and 345.  So far, the ES has found support in the low 1200 region; whether that level holds or not is likely dependent on the course of interest rates and oil prices.

August 30, 2005

[pic]

Here's a screenshot direct from the Doc's trade station.  The candlesticks are the ES; the blue indicator is the NYSE TICK, and the yellow bars are volume.  With about 30 minutes left in trading, do you want to be a buyer or seller at 1214.75?  Is the market gaining buying interest or losing it? 

If a picture is worth a thousand words, this one was worth even more in Monday's market.  It is not enough to run statistics to *predict* what is going to happen.  Ideally, you want to understand *why* the market is moving.  This is also my way of tipping the hat to , an excellent service that I have subscribed to for years.

[pic]

Early morning weakness due to the hurricane did not follow through after the open, and we broke to the upside on the most consistent buying we've seen in a while.  This turned the short-term trend bullish, as we closed well above the day's average price of ES 1209.  New short-term highs dominated among the large caps, and, by the close, a plurality of issues were trading in short-term uptrends.  The adjusted TICK ended at +546; the Institutional Composite finished at +206.  Demand was 87; Supply was 56.  New 20 and 65 day highs dipped to 492 and 279; new 20 and 65 day highs expanded to 1402 and 445.  We broke nicely above Friday's high; buying dips that remain above the day's average price is the strategy for the open.

August 28, 2005

Note: In the wake of the overnight drop due to the hurricane, I took a look at what happens after opening gaps down of < -.5% in SPY.  We have had 78 such occurrences since 9/2002 (N = 738).  The average move from the open to the day's high has been 1.01% vs. .64% for the sample overall.  The average move from the open to the day's low has been -.85% vs. -.76% for the sample.  This suggests that days following a large downward opening gap have been more volatile than average.  The average range from high to low on the gap down days has been 1.84% vs. 1.59% for all other days.  In all, 72/78 of the gap down days have had subsequent high/low ranges exceeding a full percent.  I show no statistical edge from the open to the close on gap down days, but there is a slight edge three days out, with the average gain of .28% (48 up, 30 down) beating the .15% (425 up/313 down) for the sample.

I am preparing an article for the TradingMarkets site on the topic of who is in the market (see yesterday's blog).  Here's another interesting finding from Friday's market: one and two-lot trades comprised 55% of all ES trades for the day, but only accounted for 3% of total daily volume.  Conversely, trades of 50 lots or greater were only 10% of the day's trades, but these accounted for over 70% of all volume.  Not surprisingly, large trades of 50+ lots were concentrated during the first 90 minutes of trading and during the last 30 minutes.  The frequency of large trades during a 30 minute period was significantly and positively correlated with the frequency of large trades in the next period--and was also significantly correlated with the average price movement during that period.  One implication is that filtering trade ideas based on the presence of large traders in the market could improve trading performance.

Friday's market opened weak, rallied in the afternoon, and then fell back toward its lows, as we closed below the day's average price of ES 1208 and continued the short-term downtrend.  This looks likely to extend on Monday AM, given the rise in oil prices and overseas equity weakness due to Hurricane Katrina's anticipated effects.  New lows dominated the short-term picture for the large caps and across the broad market, ending the earlier noted waning downside momentum.  The adjusted TICK was -445, and the Institutional Composite was -367.  The latter continues the pattern of heavy large cap selling that, as noted before, needs to turnaround if we are to put in a viable bottom.  Demand was 53; Supply rose to 92.  New 20 and 65 day highs rose to 561 and 309; new 20 and 65 day lows also expanded significantly to 1387 and 427.  Both of those new low figures are two+ month highs--not exactly a drying up of selling.  

August 27, 2005

Note: I will be on the road Thursday through Sunday, so Weblog entries will be abbreviated during that time.

Here's the second article in the series on "When to Stop Trading".  The first in the series can be found here.

Who is in the market?  I examined one-minute data in the ES from July 1, 2005 to August 25th (N = 39 days; 15338 minutes).  The average volume per trade during that time was 14.91 contracts, but this varied considerably as a function of time of day.  Taking 30 minute market segments, I found that the average volume per trade during the 9:30 - 10 AM (ET) period was 17.61, while the average volume per trade from 1:00 - 1:30 PM (ET) was 12.76 and the average volume per trade from 3:30 - 4:00 PM (ET) was 16.78.  There was significant variability from day to day.  For example, for the 9:30 AM period, the highest volume per trade was 22.27 and the lowest was 12.83.  For the 1:00 PM period, the highest was 19.12 and the lowest was 7.97.  

Note that this is volume per trade, which tells you something about the size that is in the market.  Why is this important?  The average volume per trade on a 30 minute basis correlates +.33 with the absolute magnitude of the price change for that period.  Moreover, there is a correlation of +.53 between volume per trade in the present 30 minute period and the volume per trade in the next.  Monitoring volume levels and the average volume per trade tells you if big players are in the market--and that is an important determinant of how much price movement the market will give you in the near term.      

August 26, 2005

The market traded in a very narrow range on Thursday, as an inside day followed an outside one.  This did nothing to change the short-term downtrend, although we're trading above the day's average price of ES 1212.5.  The adjusted TICK was +301 and the Institutional Composite was -387, continuing the recent pattern of heavier selling among the large caps and relative outperformance of midcaps and small caps.  I do not expect to see a turnaround in the large cap indices until we see some evidence of buying in the Institutional Composite.  Interestingly, we've had positive adjusted TICK readings for the past seven sessions, but negative Institutional Composite readings for six of these.  It is this divergent performance that suggests to me that this is a correction in the bull market and not the beginning of an outright bear phase.  Also underpinning that opinion is the continued lack of downside momentum among stocks during the recent pullback.  Demand on Thursday was 66; Supply was 43.  New 20 and 65 day highs dropped to 465 and 264; new 20 and 65 day lows were 795 and 278.  An expansion of new 20 and 65 day lows beyond the 1194 and 315 registered on the 18th would keep the intermediate-term picture bearish, as it would suggest a broadening of the correction rather than a drying up of selling.

August 25, 2005

Since September, 2002 (N = 737), we have had 94 outside days in SPY (i.e., higher highs on the day and lower lows) such as seen on Wednesday.  There's no real edge the next day, as the average change in SPY is -.02% (51 up; 43 down) vs. .05% for the sample overall.  If, however, you break down the outside days on a median split based on whether the close was nearer to the day's high vs. low, a small tendency emerges.  When the close is nearer the day's low (N = 47), the next day's average change is .12% (28 up, 19 down).  When the close is nearer the day's high (N = 47), the next day's average change is -.15% (23 up, 24 down).  

My continued investigations with money flow are finding that the variable does, indeed, possess predictive value above and beyond the contribution of price change.  For example, I looked at only those "neutral" occasions when the three-day SPX was up by less than .5% and down by less that -.5% from January, 2002 to the present (N = 252).  Overall, for the entire sample, the next three day change averaged -.14%, with 129 occasions up, 123 down.  When you split the sample in half by three-day money flow, however, the average change over the next three days is -.28% when money flow is in the strongest half (61 up, 65 down), but +.01% when money flow is in its weakest half (68 up, 58 down).  The effect is even stronger when you filter strong and weak three-day price changes by strong and weak money flows.  Accordingly, I will be revamping the Micropsychology Modeler to permit forecast using money flow, alongside the other major Weblog measures.

Wednesday's market opened stronger, but faded with the bond market strength.  Its decline accelerated below the important ES 1216 level, and we closed well below the day's average price of ES 1218.5.  This reinstated a short-term downtrend.  Interestingly, only a narrow majority of issues traded in short-term downtrends; this was the first of several indications that the decline was more severe among the blue chips than across the broad market.  Among the large caps, short-term new lows dominated new highs and there was considerable selling momentum.  The adjusted TICK actually closed positive at +135; the Institutional Composite was quite weak at -734.  Demand was 74; Supply was 69--unusual lack of downside momentum for such a weak day.  New 20 and 65 day highs rose to 548 and 337; new 20 and 65 day lows also expanded to 923 and 278--still below the level of new lows seen a week ago.  We have broken below intermediate-term support and, as long as we stay below there, selling rallies is the preferred strategy.  A move back into the recent trading range however, coupled with the numerous signs of waning downside momentum, would suggest a bottoming process.

August 24, 2005

[pic]

Updated note: I calculated my money flow index on a 25 day basis for the basket of stocks and found that the current reading of +2215 is the third lowest reading since 2/2002 (N = 869).  This suggests very significant selling in the large caps, even though the S&P 500 is only down about 1% over those 25 days.  When the money flow has been below +3000 (N = 43), the market has been up over the *next* 25 days by an average of 1.73% (28 up, 15 down), handily outpacing the average gain for the sample of .40% (498 up, 371 down).  As a rule, selling the market after large money flow drops in the basket of large caps has not been a good strategy.

Here are a couple of new twists in my investigations of money flow.  I mentioned yesterday that extremes of SPY money flow tended to have positive returns over a three-day period.  It appears that this effect is greatly amplified by high volume.  In other words, when SPY volume is high *and* we have extremes in money flow, three-day returns beat average returns handily.  Moreover, it appears that there is *only* an edge when money flow extremes occur on high volume days.  Similar extremes on low volume gives no edge 1-3 days out.

Because I was curious about the finding of zero correlation between SPY money flow and SPY price change, I decided to see if it was really SPY that was funky, or if maybe my money flow index was at fault.  I looked at the money flow for my basket of 17 stocks that mirror the S&P 500 average from January, 2002 to August, 2005 (N = 913).  Interestingly, the correlation between money flow for the basket of stocks and daily SP change was .64--significantly positive as expected.  It does, indeed, seem as though SPY is marching to a different money flow drummer.  Shortly I'll look at QQQQ and see if a similar effect is present.

Meanwhile, a quartile split of the money flow data for the basket of stocks suggests favorable three-day returns when money flow over the past three days has been weak, but subnormal returns when three-day money has been strong.  Specifically, we've seen an average three-day change of +.03% since January, 2002 (N = 913; 483 up, 430 down).  When money flow for the basket has been in its strongest quartile, the average three-day price change has been -.06% (119 up, 109 down).  But when money flow has been in its weakest quartile, the average three-day price change has been .31% (136 up, 92 down).  Clearly, I need to explore this further to see the degree to which money flow predicts short-term prices above and beyond the influence of mere price change.    

Tuesday's market weakened into midday before staging a rally and closing near its day's average price of ES 1220.  Our average prices over the last five sessions have been 1223.5, 1220.5, 1224, 1224.5, and 1220, clearly keeping us in the neutral trending mode.  A plurality of issues traded in short-term downtrends and short-term new lows dominated.  The adjusted TICK was +5; the Institutional Composite was -26.  Demand was 58; Supply was 69.  New 20 and 65 day highs expanded to 465 and 295; new 20 and 65 day lows also grew to 787 and 280.  Interestingly, although we touched multiday lows on the ES, the number of stocks registering new lows has not expanded either in the broad market or even among my basket of large cap issues (which, per the analysis of money flow above, has been quite weak).  We continue in an intermediate-term consolidation range and need to see expanded new lows to sustain a downtrend.

August 23, 2005

My money flow index (cumulative sum of volume times price change for each transaction) for the SPY ETF is pretty funky; its correlation with open-to-close ES price change is near zero.  Because money flow is very sensitive to large-sized transactions, it suggests that the large players in SPY are doing something different from the whales in the ES.  Perhaps they are hedging stock and stock sector positions, spreading against SP and ES, etc.  Whatever, it got me looking at the predictive aspect of SPY money flow.  Interestingly, since April, 2003 (N = 599), extreme negative values of three-day SPY money flow *and* extreme positive values have been associated with superior SPY performance over the next three days.  For the sample overall, the average three-day change is +.17%.  When my index of SPY money flow over three days is under -900,000,000 (N = 15), the average three-day change is .53% (11 up, 4 down).  When the index is over +600,000,000 (N = 30), the average three-day change is .56% (20 up, 1 unchanged, 9 down).  This may be relevant, because any negative money flow tomorrow would give us one of those occasions under -900,000,000 over a three-day period.  I'll be researching this variable further in coming days.

Monday's market opened strong, but reversed in sympathy with bond yields (see the 8/20 blog).  We closed below the day's average price of ES 1224.5, continuing the neutral trending range.  Although a majority of stocks traded in short-term uptrends, this figure began to falter very early in the session.  We also saw an increase in short-term new lows in the afternoon.  The adjusted TICK was +154; the Institutional Composite was +78.  So far, bounces in the TICK and the Composite are producing little price change, which is worrisome for bulls.  We also see tepid momentum in the Overbought/Oversold index.  Demand was strong at 97; Supply was 37.  New 20 and 65 day highs rose to 445 and 286; new 20 and 65 day lows were 717 and 248.  We continue in an intermediate-term consolidation range.  A break below ES 1218 that expands new lows would set us on a bearish short-term path; we need to surmount Monday's highs with expanding new highs to return to a bullish mode.

August 22, 2005

[pic]

The above chart shows the Dow Industrials from January, 1950 to August, 2005, adjusted for the CPI rate of inflation.  We clearly see a long-term market uptrend, but we also see a lengthy period (1966-1982) of negative inflation-adjusted returns.  An even further look back shows a long-term uptrend from the early 1900s through the late 1920s and then a period of negative inflation-adjusted returns from the late twenties to the late forties.  During the secular bull periods from 1950 to 1966 and 1982 to 2000, we never had a five year stretch of negative returns--which is what we're seeing right now.  This raises an important question for investors as to whether or not 2000 represented a secular market peak, with the current bull market representing a lower high in a longer-term formation of lower peaks and valleys.  In the 1970s we had Vietnam; now we have Iraq.  In the seventies and early eighties we had mounting budget deficits and rising rates; now we have growing deficits and recent Fed tightening and yield curve flattening.  In the 1970s we had oil shocks...well, you get the picture.  I'll have an article on this topic shortly.

Friday's market held above its prior day's lows, but closed below its day's average price of ES 1224, sending us into a neutral short-term trending mode.  The proportion of stocks trading in uptrends vs. downtrends was relatively even, and we saw few new intraday highs and lows among the large caps.  New lows have dominated among the large caps in recent sessions, and money flow has been particularly weak.  The adjusted TICK finished at +58; the Institutional Composite ended at -439--again reflecting the heavier selling among large caps.  Demand rose to 66; Supply was 35.  New 20 and 65 day highs were 352 and 209; new 20 and 65 day lows dropped to 727 and 256.  We are in a three-day trading range; a break from this range that expands new highs/lows will likely set the next short-term trend.

August 21, 2005

Lots of gurus tell us when to trade.  This week's article, the first in a three-part series, focuses on when to not trade.  

In a recent discussion on the Trade2Win site, I mention Snag-It and Camtasia as worthwhile trading resources.  The first is primarily a screen capture tool, but also can take simple videos of your desktop.  The second is a much more sophisticated desktop video program that allows you to review and archive your trading days.  No reason a trading journal can't be multimedia!

I continue to find great trading-related news links on The Kirk Report.  Sort of a Drudge Report for traders, and I mean that in a positive way.

I'm doing a question-and-answer series for the Trading Markets site and another one for the Investing Online Newsletter based in Australia.  If you have any questions you'd like me to address, send them to my Q&A address:  sonderstel@.  

FYI, sonderstel is short for the word sonderstellung.  Thanks to the many new and longtime visitors to the Weblog who have given me an exceptional and privileged position in their trading lives.

August 20, 2005

[pic]

[pic]

[pic]

[pic]

 

Correlated markets from 8/8 to the present:  Falling 10-year rates, strengthening dollar, weakening stocks.  My sole trade on Friday was buying the ES at 8:53 AM (CT) and selling at 9:07 AM (CT) for a little over a two-point profit.  Lucky timing?  Or was it seeing that bills and the dollar were struggling to catch a bid shortly after 9:00 AM?  Awareness of market regimes is crucial to short-term timing.  If you're following your stock or index only, you're missing a lot of what guides the decision-making of institutional players--and that affects near-term supply and demand.  I hope to tackle this and related topics in my forthcoming book...

August 19, 2005

[pic]

The above chart tells an interesting story.  We've had 15 days of heavy selling in the large cap stocks, as measured by the Institutional Composite.  This has also corresponded to a 15-day period of heavy selling in the broad market, as assessed by our adjusted NYSE TICK.  Looking back over the past two years (N = 496), there have only been 27 days with 15-day readings as weak as the one we're currently seeing for the Institutional Composite (-2552).  Interestingly, when the Composite has been < -2000 for a 15-day period (N = 74), there has been no edge over the next five days (.23% vs. .22% for the entire sample).  When the Composite over this period has been < -2000 *and* the adjusted TICK has been < -1500 (N = 35), the average gain over the next five days has been .91% (24 up, 11 down).  When the Composite has been < -2000 and the adjusted TICK has been > -1500 (N = 39), the next five days have fallen by an average of -.39% (13 up, 26 down).  Intense selling in the large caps is thus much more bullish in the short run when accompanied by broad market selling, as is currently the case.  

I notice that at least some of my articles are being picked up on the Yahoo! Finance site; welcome to Yahoo! readers.  Also check out my recent article on the Trade2Win site.  A new piece should be coming shortly on the Trading Markets site, focusing on psychological exercises that are helpful during periods when you need to stop trading because of frustration, fear, overexcitement, or other emotional factors.

Thursday's market once again finished below its day's average price of ES 1220.5, continuing the short-term downtrend.  A plurality of issues traded in short-term downtrends, and, while new short-term lows dominated new highs, they were more muted in the large caps than we've seen recently.  The adjusted TICK was +47; the Institutional Composite was -217 (see above).  Demand was 32; Supply was 73.  New 20 and 65 day highs dipped to 330 and 208; new 20 and 65 day lows rose to 1194 and 315.  We continue to grind lower; as long as we see lower prices, expanding 20-day lows, a negative Demand/Supply balance, and dropping average prices, selling bounces will continue to make us money.  

August 18, 2005

An interesting note from Austin Passamonte of :  If you calculate pivot points for the past day, the next day's market touches that pivot point roughly 75% of the time.  This fits with the theme of mean reversion in the markets.  To test this notion, I went back to January, 2004 (N = 404) and examined how often we touched the previous day's volume weighted average price (VWAP).  Starting at 9:30 AM EST and ending with the day's close of trading at 4:15 PM EST, we touch the prior day's VWAP on 266 out of 404 occasions.  Interestingly, the odds are higher of touching the VWAP if the first 45 minutes of trading are lower (156/213) than higher or unchanged (109/191).  Given the market's bullish bias over this period, the finding makes sense.  Here's another tidbit: If we start the market "day" at 4:00 PM EST (thus including overnight action), the market hits its previous day's VWAP more often: 297/404 times.  Given the recent strength of the overnight session (see the blog from 8/16) compared to midday, this also makes sense, as the market reverses prior weakness to return to its mean value.  A trading system idea, fading range extremes and targeting average prices within those ranges, lurks in these data.  Hats off to Mr. Passamonte, who BTW, also writes for Trading Markets.

Wednesday's market attempted to rally off the prior day's weakness, only to fall back in the afternoon, closing below its day's average value of ES 1223.5 and continuing the short-term downtrend.  A small plurality of issues traded in short-term downtrends and new short-term lows were decidedly muted compared to Tuesday.  The adjusted TICK ended at +15; the Institutional Composite finished at -55--its fourth consecutive day of net selling.  Demand was 45; Supply was 76.  New 20 and 65 day highs dipped to 398 and 248; new 20 and 65 day lows expanded to 1087 and 305.  Given the short-term bearish bias, selling bounces that cannot surmount the day's average price remains the operative strategy for the AM.

August 17, 2005

We sometimes forget that stock indexes have a denominator; in the U.S., it's the dollar.  Below is a chart of the NYSE Composite denominated in Swiss Francs.  Until we see problems with the numerator or denominator, it's tough to ignore the fact that we're sitting near multiyear highs despite sky-high oil prices, concerns about housing bubbles, deficits, rising interest rates, and Iraq.  And, while on the topic of interest rates, check out this interesting chart from Decision Point--an excellent site.  A rising rate environment is historically supportive of the dollar.

[pic]

Tuesday's market opened weak, as a steep plurality of issues traded in short-term downtrends early on and never recovered.  We broke through the lows of the intermediate-term trading range, closing well below the day's average price of ES 1227.5.  We created a short-term downtrend in the process, with short-term new lows swamping new highs.  The adjusted TICK was a weak -586; the Institutional Composite was also weak at -371.  Demand fell to 32; Supply was 107.  New 20 and 65 day highs dropped to 408 and 265; new 20 and 65 day lows expanded to 961 and 297.  The latter is the greatest number of 65 day lows in over six weeks.  With the bearish short-term trend in place, selling bounces that remain below the day's average price is the preferred early strategy.

August 16, 2005

Interesting piece of research from an article that will appear shortly on the Trading Markets site: Since July, 2003 (N = 533), the Dow Jones Industrial Average has gained over 1500 points.  Notice how those are distributed as a function of hour of day:

|Dow's Total Change |Change During First Hour |Change During the Midday |Change During the Last Hour |

|1559.36 |1966.54 |-956.76 |549.58 |

Some interesting trading ideas can be derived from the above, such as buying and holding midday weakness.  Holding midday short positions overnight has not worked well in this market.

Monday's market continued to show price firmness in the face of selling pressure.  We closed above the day's average price of ES 1234, continuing the intermediate-term neutral trend.  A growing proportion of issues traded in short-term uptrends, and new short-term highs outnumbered new lows on afternoon strength.  The adjusted TICK was +190; the Institutional Composite was -8.  Demand was 75; Supply was 38.  New 20 and 65 day highs were little changed at 582 and 405; new 20 and 65 day lows were a bit lighter at 693 and 233.  We continue in a consolidation range; yesterday's note re: breakout criteria and an upside bias remains pertinent, although I would not be buying moves to the top of the range that do not expand new highs.

August 15, 2005

Update:  Here's a tidbit.  When the adjusted TICK over the past 10 sessions has averaged less than -200 (as it has at present), the market 10 days later has been up 55 times, down 33, for an average gain of .64% (N = 88).  For the remainder of the sample (N = 414), the market over the next 10 days was up 251 times, down 163, for an average gain of .40%.  It's worth noting that we've only had 53 days (out of the last 502) with a weaker 10 day adjusted TICK reading, and yet the decline in the S&P over the last 10 days has been quite modest.  As long as we remain above the downside parameters mentioned in today's blog, I look for higher prices.

Note:  After a pretty thorough hard disk crash that wiped out many of my files, I've managed to restore most of my database and reinstall my software.  Not fun.  I should be up to speed with updated modeling in the next day or two.

Friday's market continued its intraday roller coaster, closing below its day's average price of ES 1232 and continuing its neutral trending range.  A plurality of issues traded in short-term downtrends, but this proportion firmed through the day.  New short-term lows also dominated among the large caps.  The adjusted TICK finished at -228; the Institutional Composite was weaker at -409, indicating greater selling among the large caps.  We're seeing a loss of momentum both in the TICK and in the Institutional Composite.  Demand was 32; Supply was 74.  New 20 and 65 day highs were 581 and 414; new 20 and 65 day lows were 894 and 247.  We need to see an expansion of new 20 day highs beyond the 787 from 8/10 to return to a short-term uptrend.  An expansion of new 20 day lows beyond the 1224 seen on 8/8 would re-establish the short-term downtrend.  Absent those events, we continue in an intermediate-term consolidation range, with average prices from the last six days at 1230.5, 1229, 1232.5, 1238, 1235, and 1232.

August 14, 2005

Here is the third article in my series about market modeling.  The first article asserts that changing markets making ongoing trading success particularly difficult to achieve.  The second article introduces modeling as a means for analyzing and understanding changing markets.  The final article in the series examines how modeling can serve as a foundation for implicit learning and the acquisition of tacit knowledge, and actually help discretionary traders adapt to  stationarity challenges by building their aptitudes and helping their search for an edge.  

August 13, 2005

My new trading system came about as a result of investigating outcomes following X period highs and lows in the ES.  The system provides three entry signals:

|[p|Buy X period highs when criteria of price movement and participation are met; |

|ic| |

|] | |

|[p|Sell X period highs when criteria of price movement and participation are not met; |

|ic| |

|] | |

|[p|Buy X period lows when criteria of price movement and participation are met. |

|ic| |

|] | |

Price movement refers to pattern of past price change, including momentum and acceleration.  Participation refers to the extent to which components of the market display price movement equal to or exceeding that of the ES.  The system holds for the next X periods, until a different system signal is generated, or exits MOC--whichever comes first.  

Although the system is constructed for intraday trading, its logic appears to be valid over longer time frames.  For example, from January, 2004 through July, 2005 (N = 397), the average 10 day price change has been .25%.  The average 10 day price change following an up day in SPY has been only .10% (N = 223; 126 up, 97 down), while the average 10 day price change following a down day has been .43% (N = 170; 101 up, 69 down).

If, however, we look at down days when the SPY volume has exceeded its 40 day average (N = 114), the market has been up 10 days later by an average of .65% (70 up, 44 down).  When SPY has been down on volume below its 40 day average (N = 56), the market has been down 10 days later by -.02% (31 up, 25 down).

If we focus on those occasions when SPY has been down on volume exceeding 150% its 40 day average, the next 10 days have been up by .86% (N = 21; 15 up, 6 down).  It appears that buying when SPY traders are panicking has been a worthwhile strategy; buying when they've already been buying has not made money on average.  This logic has made money on an intraday basis the last couple of days.

August 12, 2005

Note: A few computer problems remain, but those should be corrected by Monday.  If the modeling results highlighted the past couple of days pan out, we should have seen a short-term market bottom on Thursday.  

Thursday's market showed resilience in the face of selling squalls, ending above the day's average price of ES 1235.  We have re-entered a neutral trending range, with Demand at 66 and Supply at 42.  The adjusted TICK gave its first positive reading in seven sessions, ending at +313.  The Institutional Composite ended at +384.  New 20 and 65 day highs were 755 and 567; new 20 and 65 day lows were 760 and 220.  A plurality of issues traded in short-term uptrends (note how this measure nicely anticipated the market bottom) and short-term new highs dominated in the afternoon.

I've posted a couple of articles on "When to Stop Trading" on the Trading Markets site.

August 11, 2005

Model:  I continue to show neutral results in the short run (1-2 days out), a positive bias 4-5 days out.  After five-day declines similar to the one we've just had (N = 164), the market has been up five days later 104 times, down 60, for an average price change of .50%.  

Note: I'll post the model results before Thursday's open.  I've been hard at work on a trading system idea that, at this juncture, is looking promising.  The basic premise, derived from the recent articles on the Modeler, is that distinctive patterns show up in the market at particular times of day.  Because of non-stationarity within the day, it may well be a mistake to look for the same setups in the morning, midday, and afternoon.  More this weekend.

Wednesday's market was very strong in the morning, only to give back its gains in the afternoon, before staging a small recovery at day's end.  The result had us closing below the day's average price of ES 1237.5, reinstating the bearish short-term trend.  A decreasing proportion of issues traded in uptrends through the day and short-term lows swamped new highs in the afternoon.  (Note how the waning of short-term highs in the morning nicely set up the afternoon weakness).  The adjusted TICK closed at -129; the Institutional Composite finished at -72.  Our daily TICK readings have now been negative for six consecutive sessions.  Demand finished at 80; Supply at 56.  New 20 and 65 day highs rose to 787 and 584; new 20 and 65 day lows finished at 900 and 261.  Momentum is weak in this market, and the intermediate-term trend remains down.  Selling bounces that remain below the day's average price is the preferred AM strategy.

August 10, 2005

Note: I continue to see the bifurcation in modeling, where price-based predictors suggest subnormal returns over the next several days, but models utilizing the Weblog measures find no significant short-term edge, but favorable returns over the next week.  As with my analysis yesterday morning, we find twelve occasions of similar four-day weakness in the NYSE TICK since July, 2003: five days later, the market was up 10 times, down 2 for an average gain of .74%.   

Tuesday opened stronger, moved higher still after the Fed announcement, and then pulled back before staging a late attempt at comeback.  The resulting action had us closing near the day's average price of ES 1232.5, returning us to a neutral short-term trend.  Although a majority of issues traded in short-term uptrends, this proportion waned through the day, as the broad market lagged the large caps.  Among the large caps, new short-term highs dominated, as we moved above Monday's highs.  The adjusted TICK was -121 and the Institutional Composite ended at -113: weaker numbers than we would normally expect on an up day.  Demand rose to 61; Supply was 49.  New 20 and 65 day highs dipped to 521 and 387; new 20 and 65 day lows also fell to 936 and 238.  We continue in an intermediate-term downtrend and need to see a break above Tuesday's highs with an expansion of new highs to alter that situation.

August 9, 2005

Additional Note:  Since July, 2003, we have only had 12 occasions in which there has been a similar pattern of three consecutive days of heavy selling in both the adjusted TICK and the Institutional Composite.  Expectations for the next day were mixed, but were decidedly bullish five days out.  Ten occasions were up, two down, for an average gain of 1.31%.

Note: Monday's market was weaker than the small drop in the ES would suggest.  As a result, the Modeler continues to crank out forecasts that indicate subnormal returns in the short run (1-2 days out) and mixed findings thereafter.  It's an interesting situation because the models that use traditional price-based predictors are more bullish than the models that incorporate the supplemental predictors from the Weblog measures.  This is exactly the situation I faced prior to Monday's market, when multiple models gave mixed results.  My sense is that the supplemental predictors, which capture the breadth as well as extent of market movement, add value to the forecasts.

After an up open on Monday, the market sold off, taking us once again below the day's average price of ES 1229 and keeping us in a short-term downtrend.  Again, a plurality of issues traded in short-term downtrends, and, while new lows continued to dominate among large caps, they receded in number.  The adjusted TICK was weak for the third consecutive day at -735; the TICK oscillator is clearly oversold.  The Institutional Composite also had its third straight day of selling; its oscillator is also oversold.  Demand was 36; Supply was 79.  New 20 and 65 day highs rose to 559 and 408; new 20 and 65 day lows also rose to 1224 and 272.  We are now in an intermediate-term downtrend; selling bounces below the day's average price remains the preferred strategy in early trade.

Check out this chart of the S&P vs. the Dow Jones 20 Bond Index since 2003.  It pretty well tells the story of what has supported this bull market--and what could eventually be its undoing.

August 8, 2005

Note:  Running multiple models generated by the Modeler yields mixed results--not as robust as the forecasts of weakness last week.  If there's an underlying theme, it's that of near-term weakness (1 day out) and later strength (4-5 days out).  Not something I'd hang my hat on, but suggestive.

Friday's market continued its weakness, closing below its day's average price of ES 1230.5 and continuing the downward short-term trend.  A plurality of issues traded in short-term downtrends, and new short-term lows dominated among the large caps.  The adjusted TICK was a very weak -1031; the Institutional Composite was also dominated by selling at -424.  (See the Aug. 7th blog for more re: that selling; note also how the rolling over of the TICK and Institutional Composite lines nicely anticipated this latest market weakness).  Demand was 21; Supply was 159.  This means that more than seven times as many stocks displayed significant downside momentum as upside.  New 20 and 65 day highs dropped to 455 and 302, and new 20 and 65 day lows expanded significantly to 1120 and 213.  The expansion of new lows in the large caps and in the broad market, combined with the very weak TICK, suggests that this is a broad-based decline.  We are close to a negative point in momentum normally associated with short-term market lows, but we're not yet there on an intermediate-term basis.  Given the intermediate-term breakdown, selling bounces that remain below the day's average price remains the reigning strategy.

August 7, 2005 

Note:  July's blog entries are now archived above.

Here is a blog site that I highly recommend: The Kirk Report.  Charles Kirk does an excellent job of collecting interesting information about stocks, the economy, and trading, with personal touches added in.  Quite good.  Another blog worth a look is John Rutledge's site.  He offers economic perspectives, insights about sectors, and straightforward opinion.  In my August 6th entry, I mention other resources worth a look.

The Micropsychology Modeler:  How can statistical modeling provide an edge in the market?  How can it help traders adapt to changing market conditions?  The first article in this three-part series looked at the challenge of ever-changing market cycles.  The new, second installment takes a look at the process of modeling and how it can be integrated with discretionary trading.  Next week, the third and final segment in the series will explore the use of modeling as a market learning tool.

We've had two days of broad selling, retracing two week's worth of gains during that time.  I notice that the adjusted TICK on Friday was -1031.  Since July, 2003 (N = 518), we've only had 22 days with the TICK below -1000.  One day later, the market was up 14 times, down 6, and unchanged twice for an average gain of .13%.  This compares with an average one-day gain during the entire period of .04%.  Four days out, the market was up 14 times, down 8, for an average gain of .25% (compared to .17% for the entire sample).

Further, we've only had 13 days during that time in which the two-day adjusted TICK has been below -800.  One day later the market was up 8 times, down 4, and unchanged once for an average gain of .17.  Four days later, the market was up 10 times, down 3, for an average gain of .80%.  

When, however, we look at occasions where the two-day TICK is below -800 *and* the most recent day's TICK is below -1000 (N = 9), we see an average daily change of -.01% (5 up, 3 down, 1 unchanged) but an average four-day change of .76% (7 up, 2 down).

I've run a number of analyses across the database and the conclusions are similar: the day following a broad selling day--particularly one that follows a broad selling day--shows mixed-to-subnormal results, but we see favorable expectations 3-4 days out.  In general, it appears that buying weakness 1-2 days out might be a viable swing-trading strategy.  Note, however, that these results are based on a very small sample.  More modeling results on the larger database Monday AM.

August 6, 2005

So far the results of my modeling research have been promising.  I have developed a second Modeler as a check on the findings of the first.  The original modeler utilizes predictors largely derived from price and volume over multiple time frames, including price change, acceleration, volatility, trend, and volume change.  The new modeler incorporates predictors from the Weblog measures, including the adjusted TICK, Institutional Composite, Demand/Supply Index, and intermediate-term New Highs/Lows.  Both Modelers are designed to identify directional tendencies in the ES up to five days out.  My article detailing the methodology and rationale will be posted this weekend.

I like the direction being taken by the new Investing Online Newsletter, which is offered at the Sharetrading Education site.  Based in Australia, John Atkinson and his colleagues are taking a global view of investing as well as trading.  I'll be writing a weekly Q&A column for the newsletter based on readers' questions.  My hunch is that if you email John and mention my site, he'll email you a complimentary issue of the letter.  

While I'm at it, I'll also mention a few other resources.  Daryl Guppy's site appears to have particularly sophisticated technical analysis tools and education, with a variety of instructional materials available.  He also produces an informative newsletter and has developed several proprietary indicators.  I've exchanged a few emails with Ian, the proprietor of Shrewd Investing based out of Scotland.  There are quite a few links and research tools on the site; it's easy to get caught up in all the offerings.  I'm also doing a column for Trading Markets; the newest installment on "When to Not Trade" was posted Friday evening.  The editor, Eddie Kwong, does an excellent job of updating site content regularly.  I continue to benefit from the excellent research of Jason Goepfert on the Sentimentrader site.  He did a nice job of anticipating the recent market weakness.  My next column for the Trade2Win site will appear around mid-month; there are some good discussion threads on the site, many of which have a global focus.  I also continue to enjoy the highlights from Victor Niederhoffer and Laurel Kenner's Spec List, as posted on the Daily Speculations site.  You can find some worthwhile observations on markets and life on the site; new content is posted regularly.

August 5, 2005

Note: The Modeler picked up on the weakness we saw on Thursday.  Even given that decline, it continues to suggest subnormal returns through three days out.  Knowing this, and gauging the market trend, can prepare the trader for opportunities on the sell side, as was the case Thursday.

The market sold off steadily on Thursday, closing well below the day's average price of ES 1240 and initiating a short-term downtrend.  A plurality of issues traded in short-term downtrends and, among the large caps, new short-term lows dominated.  The adjusted TICK continued its weakness at -687; the Institutional Composite reflected selling in the large caps at -292.  Demand dropped to 25; Supply soared to 136.  New 20 and 65 day highs dropped to 806 and 569; new 20 and 65 day lows rose to 579 and 141.  Although we remain in the intermediate-term consolidation range, the weakness in momentum noted in previous blog entries is now showing up as outright selling, especially in the small and mid caps that had been market leaders during the rally.  Selling bounces that remain below the day's average price is the initial strategy for Friday AM.

August 4, 2005

Note: The next iteration of the Micropsychology Modeler is complete; an article will give the details this weekend.  It forecasts subnormal performance for the ES over the next 3-4 days.

The market recovered from a modest bout of selling to close near its highs, above its day's average price of ES 1245.  This returns us to a neutral trending mode, as the last five days have had average prices of 1244, 1242, 1239, 1244, and 1245.  Despite the late strength, a majority of issues traded in short-term downtrends and only about a quarter of the number of stocks in my basket made short-term highs.  The adjusted TICK was -68; note how buying momentum in the TICK has waned.  The Institutional Composite was +256, as large caps led the S&P higher.  The tepid momentum in the broad market is evident in the Demand/Supply figures, with Demand at 41 and Supply at 69.  New 20 and 65 day highs dipped to 1326 and 1033; new 20 and 65 day lows were 383 and 122.  We continue to see higher prices and mediocre volatility; until we see an expansion of new lows, it is premature to sell this market.

August 3, 2005

Note: I'm currently working on procedures for updating the predictions of market models.  It's a very interesting question as to how to weight old vs. new information.  For example, if a model forecasts a move three days out, but a large move occurs the next day, a re-running of the model may produce very changed expectations for the next two days.  My current work suggests that the most recent market events must be weighted the most and that frequent model revision is critical.  More on this later.

Tuesday's market broke above the ES 1142 level mentioned yesterday, closing above its day's average price of ES 1244 and initiating a short-term uptrend.  We saw a moderate expansion of short-term highs, and a plurality of stocks traded in short-term uptrends (although this did not expand during afternoon strength).  The adjusted TICK was +182; the Institutional Composite was +58, as recent buying has been less robust than it was in early-mid July.  Demand rose to 67; Supply dipped to 44.  New 20 and 65 day highs rose to 1554 and 1235; new 20 and 65 day lows were 378 and 124.  We remain in an intermediate-term consolidation, but the short-term uptrend will dictate buying dips that remain above the day's average price.

August 2, 2005

Note: Nice setup for a short trade late in the day Monday, as we saw higher prices in the S&P, but fewer issues participating in the rise.  On good breakout moves, we generally see market sectors moving in unison; when the big caps move alone, the moves are often retraced.  On a different note, I'm pleased to be contributing columns to the Investing Online Newsletter, published in Australia.  My first column for Trading Markets appeared on Monday.  I'll post here when my next Trade2Win column appears.  

The market traded in a narrow range Monday, closing just below the day's average price of ES 1239 and continuing the neutral short-term trend.  The number of stocks making new short-term highs was muted, and the proportion of issues trading in short-term uptrends waned through the day.  The adjusted TICK was -110, its fifth negative reading in the past six sessions.  The Institutional Composite was -295; it had been quite strong through the recent rally but now is negative for four of the past six sessions.  Demand was 53; Supply was 57.  New 20 and 65 day highs dipped to 1409 and 1083; new 20 and 65 day lows rose slightly to 377 and 119.  We continue in consolidation mode; a break above 1142 with an expansion of new highs or a break below 1135 with an expansion of new lows would likely put us in a short-term directional mode.

August 1, 2005

Note: I continue to work on the Modeler.  As it looks now, I will have two separate modeling platforms: one that looks 1-5 days out and another that creates forecasts up to 40 days out.  Most recently I've been testing the addition of a new predictor variable that captures whether a given price change has occurred in the context of rising or declining market strength.  This contextual variable seems to make quite a difference in initial tests, which will likely necessitate a reworking of the three variable model I had recently created.  That model, FWIW, shows no distinctive edge in Monday's trade, but bearish expectations early this week into the Wednesday/Thursday period.  

Friday's market retraced much of the recent gains, as we closed below the day's average price of ES 1242 and returned to the neutral trending mode that has characterized much of the trade over the past weeks.  We saw an expansion of short-term lows and a plurality of issues trading in short-term downtrends (although this plurality, interestingly, did not expand late in the session when the market made daily lows--a pattern that often sets up early strength the next day).  Selling was broad: the adjusted TICK was -464, and the Institutional Composite was -599.  Demand was 42; Supply was 73.  New 20 and 65 day highs were 1501 and 1166; new 20 and 65 day lows actually dipped to 354 and 104.  

Overall, we've seen broad market strength in the recent high-low numbers and the Cumulative Demand/Supply Index.  While momentum has clearly waned in the large cap measures of Highs and Lows and Money Flow and in the broader Overbought-Oversold and Efficiency measures, nothing so far has taken us out of the low volatility, intermediate-term consolidation range that has thus far conformed to historical precedent and given us grudgingly higher prices over time.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download