Subject: Re: A mechanical strategy
Mark/For anyone that's interested ....My thoughts FWIW

(1) I dont think I am fully prepared to classify the backtest link as "complete garbage" like Ray, at least upfront. A lot of times we get enamored with Stock/Indices data having longer and longer lookback thinking it actually adds Robustness --- that's a bit of a fallacious concept.

Because it depends on the objective of the test - a timing system involving avoidance of bears - might just survive scrutiny if it only covers 2000+ - because if we go back to 1928 ( most long term data sources - GTR1 is 1926 IIRC) - it simply ADDS 2 REAL Bears -- the Depression ( 20 year bear) and Stagflation (1970s - 10 year bear)

So in essence this guys backtest covers 2 bears and 2 bulls - so in some sense one would be initially tempted to call it "balanced"

(2) Except is it ? ie in the context of the strategy pursued.

(3) Full Disclosure : My lens and general ideas are influenced by a test I ran many years back - for a family 529 strategy on the SP500 and NDX100( mostly academic - most 529s only have the Vanguard 500 or Extended option). The basic rules tested were VERY SIMILAR
(a) Money invested in monthly equal amounts - goes to cash by default
(b) "buy the dip" thresholding - all accumulated funds go into the index on weekly threshold drop
(c) Investigation of a threshold/BCC - based only on the index price to exit market
(d) Investigation of 10 year fixed return ( tells you when I was doing it :) ) comparison against DCA and impact of starts

Learnings :
a. Thresholds cant be too low or too high ( like IIRC 7%+ didnt have any returns other than T-Bills for early 1950s starts - ie the 50s were literally the golden non-volatile period for the US stock market.)
b. Starts matter ( from the previous statement - a VERY VERY FAMOUS investor got his start there --- and the rest they say is History!)
c. You are roughly looking at 5-10% for your ideal BTD threshold
d. Blind thresholding ie akin to a profit latch on the way up - Decimates returns.
e. You need some sort of a BCC - in this case I came up with a variant of BCCII ie Slope and BCCIII ie DBE to book your profits and exit.

With these - other than those 4-5 year ( 1951 to 1955 I think) starts you beat DCA ( Not B&H - you cant) - this is DVA vs DCA.

(4) Which brings us back to Ray's contention of the backtest validity - and the notion of adequacy/availability of funds.

First the funds question - I wrote it down ( ie the rules from that link) on paper - algebraically. Its a Geometric progression with corpus allocated 1/2 and 1/2 at the beginning. Being a progressive series - it will NOT run out.

Except its proportional to the degree of the Monthly drop and/or gain and its product - ie Hyperbolic function - which is an arch visually or an upside down one.

(5) This means the initial few hits matter the most- majority of the money will get either Invested or Booked into profits at the 2 ends ( the hyperbola reaches max/min at 0.5)

So you dont really accumulate on the way down - because money becomes scarcer and scarcer. This isnt' meant for a deep bear - ie value buying.

MOST IMPORTANTLY: And this is where at least from an index standpoint ( or for a stock try AAPL) - the longer duration matters. If you extend the stock market data to pre 1900s ( there are a FEW sources) - other than the Great Depression - you can smooth it to an ever Exponential increasing curve.So you end up selling most of it too quickly also.

Net net : The geometric ( or really any block type) selling on the way up is a "Dud" strategy - you really need a range bound index/stock for this to work in that scenario. And if you LUCKY enough to pick Apple during the tech bear - you really shouldn't deploy this AIM thing - whenever AAPL drops precipitously - its associated with some sort of a systemic risk - else it wont give you the dip entry ( For a stock I am guessing it will be 10%+)


THINGS TO CONSIDER/Thoughts

1. This is more suited for Index ETFs than anything else
2. Analyze a drop metric ( use GTR1 if you can or Yahoo! and spreadsheet) - Std Deviation, DDD3 or if you have TA software - ATR ( Average True Range) - over a very long term like 1950 to 2000s -dont include till today - that's the Post or Out of Time validation. The period you want should be like 3 months at least - this determines the frequency of trading the dip. You can do 1 month also if you want to exactly mimic AIM.

Pick a value just under - this will trigger the buy. Higher the value - higher the chance of sitting on cash. While lower the value - higher the chance of depleting all dry powder in the first 1-2 lots and then taking it in the chin in a deep bear

3. Figure out an appropriate BCC rule ( you can just use BCC itself - just because I was constrained by spreadsheet doesnt mean that's the correct way to go)

The system should look like this
a. Start - Just like AIM
b. Buy the dip rules - like AIM with the analyzed threshold
c. Exit all longs on BCC - take portfolio to cash
d. Enter back longs on BCC bullish - all stored cash into index ( Not AIM - you need to go all in)
e. Deploy something like 1.5 or ideally 2x of dip threshold or calc the Upside deviation or something, to book some profits to raise cash for next dip.
f. The above condition exists ONLY post a BCC buy and you are trickling out profits.

DISCLAIMER: Like everything it AINT PERFECT. For example I can readily come up with 2 counter examples of V shaped recoveries ( COVID and Fall 2018) - which can trigger the BCC buy ie sell and then buy - being net negative ie loser and then trap you in a real bear.

But intuitively these should help to keep you on the better side ( I didnt say righter :) if that's even a word)

Hope this helps!
Best