Post New

The 'Extremely Simple Bottom Detector' (ESBD) fired a Minor Bottom on the 3rd April in-line with the latest carnage as [%NYSE new highs]-[%NYSE new lows] were -17.41% of traded issues. While the ESBD has received some adjustment (in private by me) over the last year, a Minor Bottom would also have been detected by the previously public -16.67%/-33.33% version.

The updated / current version sets a Minor Bottom at the 'mean - 2SD' of the '[%NYSE new highs]-[%NYSE new lows]' expressed as a percentage of issues traded. A Major Bottom is set at 'mean - 4SD' & if required an Extreme Bottom (which really only occur during Bear Markets) is set at 'mean - 8SD'. In all cases the mean & SD are calculated since 20051027 (as that is how far back my detailed data series extends). The use of a Gaussian / Normal distribution is a simplification - typically the stock market has 'fatter' tails than the equivalent Normal distribution which will lead to a greater number of minor, major (& even extreme at 8 standard deviation) bottoms during times of stress. This will exhibit itself in more frequent fires of the detector and so is something I can probably live with.

The prior version used [%NYSE new highs] - [%NYSE new lows] < -16.67% = Minor Bottom & [%NYSE new highs] - [%NYSE new lows] < -33.33% = Major Bottom where -16.67% is basically 1/6 more of the NYSE issues traded achieved new lows than highs for a Minor Bottom & -33.33% is of course a 1/3 more new lows than highs (of total NYSE issues traded as obtained from the WSJ Market Data resource). In practice this means a Major Bottom is signaled when a little more than 1/3 of the NYSE issues traded are achieving new lows. This version had the downside of often failing to signal during corrections where panic didn't 'set-in' & so - in good times - went years without signalling a bottom. As a result the current version - with an Extreme Bottom added - is an attempt to add more texture to firing where it both fires more frequent 'Minor Bottoms' in corrections while firing relatively few 'Extreme Bottoms' in full-fat Bear Markets (9 'Extreme Bottoms' total in spread across 3 corrections or bears - late-2008, 2011 & 2020).

Historical chart is here: https://ibb.co/4RvdkMhY (this is up to the end of last week - will include another 'Minor Bottom' when updated).

Platy

In practice this means a Major Bottom is signaled when a little more than 1/3 of the NYSE issues traded are achieving new lows. This version had the downside of often failing to signal during corrections where panic didn't 'set-in' & so - in good times - went years without signaling a bottom.

Absence of panic does not appear to be a problem at the moment. Using Wall Street Journal Market Diary figures...

https://www.wsj.com/market-data/stocks/marketsdiar...

... today was a "major" bottom.

Issues traded 2,856
New Highs 15
New Lows 1042

(15-1042)/2856 = -35.95%

Can't say how this matches up to the SD-based model.

My apologies if I have done the math incorrectly.

Baltassar (who cannot calculate an "extreme" bottom, but expects one nonetheless)

I concur with Baltassar's calculation that the ESBD signaled a 'Major Bottom' Friday - those were the same numbers I calculated!

For the current version using the Mean & Standard Deviations the math works out as the following...

    %Hi-%Low Mean = 1.163% 20051027 through 20250328
    %Hi-%Low SD   = 6.685%

    MinorBot = 1.16% - (2*6.685%) = -12.21%
    MajorBot = 1.16% - (4*6.685%) = -25.58%
    ExtrmBot = 1.16% - (8*6.685%) = -52.51%

Of course a good approximate version would be just -12.5%, -25% & -50%. The exact numbers are probably not all that important - we'd all notice a %Hi-%Low = -49% as being extreme-ish in any case, so I'd expect anything in that ballpark would suffice. It's more just a somewhat formal way of defining/testing a thought process & frequency of firing.

(One other thing I should probably do, as the current formulation is a little dynamic (the mean & SD change slightly over time) is that a copy of past performance should be created for the values that were active then. Once I get a bit more time I'll do that also - it'll only be a little more s/sheet work to take an additional copy under those circumstances - s/sheets are great!)

Platy

And another 'Major Bottom' signaled Monday at %Highs - %Lows at -40.69%.

The commentary on that would almost exactly match Jims' latest - wait for an 'up' day on increased volume before buying. It's not uncommon in past bear-like markets for there to be occasional 'up' days before resuming the downward trend. In this instance it's hard to know given the man-made nature of the event (although it has been brewing for a while).

I agree Platy.

Past cases it was ingrained in the system - this time its been "injected" into the system.

Thus it needs a mechanism to pump it out - the current HOPE is , it will. If it starts to set in .....

Interestingly - after signaling a "MinorBot" on Tuesday 8th April, Wednesday 9th April was still a "MajorBot" as prior to the market gyrations upward the day had set a lot of New Lows (752) & very few (8) New Highs resulting in a [%New Highs]-[%New Lows]=-26.16% (although the -16.66%/-33.33% variant would have only signaled another "MinorBot"). And the New Lows were well in excess of the Declines - which does happen sometimes but is not unheard-of.

Be careful out there!

Similar on Nasdaq: I see a Minor Bottom on April 9th, and a major bottom on April 8.

Thanks for bringing up the New Lows being in excess of the Declines! I will watch that.

Another data point: The Nasdaq NH-NL is on a 35-day negative streak. That's unusual.

Post New

Mechanical Investing FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.