ES -- DX/CL -- isee -- cboe put/call -- specialist/public short ratio -- trinq -- trin -- aaii bull ratio -- abx -- cmbx -- cdx -- vxo p&f -- SPX volatility curve -- VIX:VXO skew -- commodity screen -- cot -- conference board

Thursday, January 10, 2008


betting on something more

yesterday i noted the huge fraction of the s&p hitting new 250-day lows on tuesday. with new intraday lows hit yesterday (wednesday) the fraction was driven to a new peak -- 34.8%.

i ran the data back to the earliest meaningful point -- the analysis i use examines only the current composition of the index, so as one goes back their participation in the s&p 500 declines. in 1984, about 250 of the current 500 are the components from which i'm calculating high-low data. so that is a caveat, but i suspect those 250 are a reasonable proxy for the entire s&p at that time. i can go back no further than that because of a january 1, 1984 shakeup in the composition of the s&p.

it is stunning, however, to consider -- each of the points upon which so large a fraction of the s&p components were hitting new lows constitutes a major -- not minor, but major and multiyear -- launching point for the index.

then take a look at the aaii bull ratio four-week moving average. at 34.5%, it is at its lowest reading since 2005 -- and really was only exceeded significantly from here last in 1990. the bearish component is at a stunning 59%.

so what am i saying? well, if you don't mind some turbulence going forward, this looks a magnificent buy point with massive risk-reward advantage.

as a trader who does mind some turbulence, we might examine the past cases a little more closely.

the 2002 low was first put in in july, then retested in october with a further abortive test in march 2003. from the initial new low spike, there was no real downside, but considerable volatility with two 15%+ rallies and failures between tests.

the 1998 low hit at the end of august, bounced, then retested in early october before launching. a couple percentage points of intraday drawdown on the retest, but nothing material. volatility again, though, with a 10% rally and failure sandwiched in between.

the 1990 low hit in august, bounced, tested -- and failed by a couple of percent -- in late september, then successfully retested in early october. it marked reduced new yearly lows at all three lows. the drawdown from the initial new low spike ended up being less than 3%. quite volatile, however, with a number of 5% two-day rallies and subsequent three-session failures.

the 1987 low was put in in mid-october, bounced 15% in three days, tested the closing low in late october, rallied another 12% or so in five days, and then retested into december before recovering. no downside at all from the intraday low on the initial spike, but massive volatility.

the 1984 low (shown here zoomed in a bit and with arrows for clarity) struck in late may, bounced 5%, tested again in mid-june, bounced 5% again, then fell back to test again in late july. again there was very little downside risk from even the first low. clearly there are data-size limitations here, however.

the key theme is clearly that not until the divergence of new lows became clear did the subsequent longer-term rally hold -- but also, that there was already very little material risk of deeper price lows from the time of the initial new low spike.

one might well believe that something different is afoot. if this is, contra the BCA, the end of the debt supercycle... well, i don't have data here for the 1929-33 period, but i'm sure it was unbelievable by these standards. i also can't divine much due to data limitations from the 1966-1982 period -- which is a crying shame, because that may be the most relevant recent precedent.

so i'm left to go forward on what i have information about. and from that perspective, the following plan of self-instruction looks sensible:

  1. treat current long holdings as a first-bounce trade with 5-10% upside from s&p 1380 -- a range of 1450 to 1520. watch 20-day highs and lows -- and close the long with extreme prejudice, particularly on any untoward expansion of new lows, even a single suspicious day. accept that each and every 5% move is unobtainable by your methodology.
  2. short any first-bounce reversal anticipated by 20-day new lows with the expectation of a retest.
  3. buy very aggressively on price retests of the intraday lows. often, 20-day hi-lo data will anticipate bounce timing, but downside risk should be quite limited.
  4. be unreasonable about cutting the long trade until either the second test is in, or a very plainly successful first test is in.
  5. be open to the possibility of much higher highs, in spite of whatever you think is happening. trade the charts, not the rationale.

now, however, there is also a devil's advocacy point -- what if it isn't a multiyear low?

september 21, 2001 saw a new annual low spike of similar proportion to what we're discussing here. no retest followed, though the internals gave some indication that they could in late october following the initial three-week 15% bounce. instead the market rallied, topping at 23% off the low in december and installing a triple top at that level into march. the rally then collapsed into the july 2002 low.

on february 24 2000 a new annual low spike emerged. it too never really received a proper price test -- though there was an ominous shakeout march 7 that would likely have sent me short by the rules above -- before exploding into march 14 and putting in the (until recently) all time high some 15% higher than the low four weeks on.

the index stayed in a range thereafter, retesting the 1340-1360 range on the downside and 1520-1540 on the upside (sound familiar?) and forming a complex triple top into september.

both these cases were characterized by an improper test of the price low, but their fault resulted not in sudden lower lows -- indeed, in both cases a hefty 15% bounce materialized -- but simply an impermanence of the bottom several months later. as a trader, the signal will be long past by then and all this of little concern. but it does especially point out the need to be nimble, especially in going short on anticipation of the first retest of 1380 -- watch those 20-days and fear them!

but it's worth noting, i think, that the aaii sentiment data was not nearly so negative at these junctures as it is now. this not only has some major technical qualities of a nascent durable low, but also the sentiment component. as such, i'd expect at least one retest over the coming three months -- and then a long, strong rally -- one that flies in the face of my every current expectation....


This page is powered by Blogger. Isn't yours?