Daily Commentary - Posted on Tuesday, June 23, 2009, 5:41 PM GMT +1


Jun Tuesday 23

Cluster of 90% Down Days – Bullish or Bearish ?

On Monday’s session market breadth on the NYSE and the NASDAQ was heavily lopsided on the downside with NYSE Advancing Issues/Declining Issues at 0.13, and NYSE Advancing Volume/Declining Volume at 0.07 (NYSE TRIN at 1.77), and NASDAQ Advancing Issues/Declining Issues at 0.20 and NASDAQ Advancing Volume/Declining Volume at 0.11 (NASDAQ TRIN at 1.74). This was the second 90% down day (volume in declining stocks accounted for at least 90% of NYSE Total Volume) and third 80% dwon day during the last 6 trading days.

I thought it would be interesting to check if -and to what extend- cluster of 90% down days might provide a tradable edge on any side -long or short- of the market.

I therefore checked for the following setups (Setup S2 was triggered on Monday’s close):

  • One 90% Down Day (volume in declining stocks accounted for at least 90% of NYSE Total Volume) during the last 10 sessions (Setup S1),
  • Two 90% Down Days during the last 10 sessions (Setup S2),
  • Three 90% Down Days during the last 10 sessions (Setup S3),
  • Four or more 90% Down Days during the last 10 sessions (Setup S4), and
  • Four or more 75% Down Days during the last 10 sessions (Setup S5).

Table I shows the ES (E-MINI S&P 500) performance (since 01/03/2000) over the course of the then following 10 sessions immediately following those sessions where setups S1 to S5 listed above had been triggered.


It is interesting to note that the ES (S&P 500 E-MINI) shows a below-average performance over the course of the then following 10 sessions immediately following the first or second 90% down day (due to a below-average probability for a higher close 10 sessions later and a below-average profit factor of 0.76 and 0.62 respectively compared to an at-any-time profit factor of 0.91 concerning the ES‘ performance of the course of any 10 sessions), while the outlooks gets significantly positive from a third 90% down day onward, although sample sizes concerning setups S3 and S4 are way too low to read any statistically relevant into it.


Bottom line:

Due to the fact that we’ve just experienced the second down day during the last 10 sessions on Monday’s session, it seems more than likely that any rebound may be temporary only and sellers will probably drive the market lower over the course of the following 10 sessions (even if we’ll not get a third 90% down day). So for the time being any tradable edge beyond a potential short-term/intraday edge on the long side (like on today’s/Tuesday’s session on any significant weakness before, on or shortly after the open) will probably be provided on the short side of the market.

Successful trading,


P.s.: I’ll regularly make some intraday updates as well using Twitter. If you’re interested in, please have a look at the blog during the trading session as well or subscribe directly to Twitter (recommended).

Disclaimer: Long DAX (German Aktien Index) at time of writing.

Add to Technorati Favorites

Comments (3)


  1. Aidyn says:

    Aidyn again.
    I find that results for 90% clusters differ depending on whether they happen at the highs or lows of the move. My results differ greatly depending on whether 2 or more 90% down days occur when a 50 day high happened within the last 10 days or conversely, whether a 50 day low happened within the last 10 days.

    • Aidyn,

      thanks for the comment and your kind words.

      But my answer is always the same: I show unbiased/unfiltered data only and (regularly) do not add a second (leave alone any additional) condition based on x-day SMAs/EMAs and/or x-day high/low. The problem would than be that it is always possible (if not probable) that if you don’t add a 50-day high but a 10/20/100-day high/low instead (there is no ‘magic’ behind a 50-day high), the results may differ again (with respect to a 50-day high).

      So -at least concerning my methodology of trading (and analysis of historical data)-, I always keep it as simple as possible, show unbiased data only also in order to keep the sample size statistically relevant.


Leave a Reply

Your email address will not be published. Required fields are marked *