Post New

Never heard of this before but looks interesting: https://ritholtz.com/wp-content/uploads/2023/01/BA...
Has anyone backtestes this before 1945?

<< Never heard of this before but looks interesting: https://ritholtz.com/wp-content/uploads/2023/01/BA...
Has anyone backtested this before 1945?>>

Me neither.. Excellent find, and excellent question. If genuine, and he did not data-mine the parameters after the 1970s discovery, then the 1980 onwards post discovery data looks very effective.

It is always important to understand the details about if/how the parameters are adjusted over time. (That is why MI on this board, with everything timestamped and unable to be editing, is useful to track whether parameters are changed - I'm also impressed that the community is carrying forward screens developed as far back as 2000 such as Peg26, plowRS etc.)

- Manlobbi

"If genuine, and he did not data-mine the parameters after the 1970s discovery"

That is open to debate. The 2 people who can answer this question about long term effectiveness are Robbie and Zee.

There are many variations of Breadth Thrust - this falls into this class.

Essentially the BAM indicator is a Pooled Ratio of Advances/Total >67% cumulated over 10 days. Question is it can be done on Issues and/or Volume. Which one?

The classical Zweig Breadth Thrust is long term backtested ..... It has iirc never failed to mark the initiation of a Bull ... while on the other hand it DOESNT purport to being 100% accurate - like this one. 100% accuracy tends to be a bit fallacious - and brings on the Data Tuning questions.

NOTE: This one doesn't trigger in the 2002-03 Bear end.... which makes this a bit suspect.

The classical ZBT did not trigger on this rally - because it has an Oversold requirement (<40) and then rise to 60 over 10day period.
BUT - if you relax it a bit ...ie not require <40 but require the 20 point rise

it did trigger on 1/12/2023 - from a low of about 43 on 12/28 to high of 67 on 1/12.

Concept is the same : You HOLD 1-3 months out and Indices are higher

Hope it helps!

Best

The difficulty with Advance/Decline studies is different sources use a different number of stocks even though it is fo the same exchange. This was discussed with NH/NL calculations. WSJ has different numbers than Barchart that is different from other sources.

Using GTR1 and the following steps:

excd.a = 1
styp.a = 10!11!18!48
dspo(1) >= 1
trp(1,1)<0 (<0for decline, >0 for advance)

and using the count feature there have been 102 times when the sum of the last 10 day advances divided by the sum of the last 10 day declines exceed 2. Seventy occurred from 19441231 onwards.

Returns 19260114 to Present A/D >2

1 mth 3 mth 6 mth 12 mth
Median 1.04 1.07 1.13 1.15
Average 1.05 1.07 1.11 1.18
Min 0.91 0.78 0.71 0.77
Max 1.46 1.55 1.38 1.94



1926014 to 19441231 A/D>2

1 mth 3 mth 6 mth 12 mth
Median 1.05 1.02 1.06 1.02
Average 1.08 1.04 1.03 1.12


19450101 to Present A/D>2

1 mth 3 mth 6 mth 12 mth
Median 1.04 1.09 1.13 1.19
Average 1.04 1.08 1.15 1.21


Returns 19260114 to present
ignoring A/D ratio
1 mth 3 mth 6 mth 12 mth
Median 1.01 1.03 1.07 1.13
Average 1.01 1.03 1.06 1.12

Post New

Mechanical Investing FAQ

---

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