Hi

I have a wonder or a thought I want to share with you guys. It is about the difference between a strategy based on "real foundations" which works and a strategy based on "faith foundations" which also works. I expect the first to not be biased and the second one to be biased.
I am thinking about that since it is difficult to know the difference between "adjusting the code to get a PF>1 in the backtest" and "improving a strategy because an ineficiency is being used
It is about knowing if the foundations to change the code are real or faith.

And I think the difference could be that real foundations are based on market itself and faith foundations on external things.

Example:
Market itself foundations -> There is a correlation between the price movement and the first trading day of the month.
Market external foundation -> There is a correlation between the raining days and the price movement.

So the first would be more like an ineficienty which can be used and the second would produce a biased strategy (if it can be named like this)

Another way to consider if the strategy wont pass a WRC is the degrees of freedom that the strategy is based on. Or in other words: How many hypotesis or supositions do I need to make so that the strategy fits the price behaviour?. In that way external foundation needs more hypothesis like: In raining days people is more depress so it influence how they trade in a way that fits the correlation.

Another way could be the Occam's razor.

I wonder if our thinking process could be checked too like: If someone created a strategy which needs 5 or more hypothesis to get a PF>1 in a backtest then it is probably biased. Like
Case1: A (Eternal:weather ) => ... => D (Price movement) => Trade
Case2: A (Internal:priceUp) => D (Price movement) => Trade
In case1 we need like "bridge thoughts" to fit the data like:
A (Eternal:weather ) => B => C => D (Price movement) => Trade

I wonder if this can be tested by using external data like:
case1: Temperature GoesUp => Buy
case2: Moon Increase => Buy
case3: Price Up => Buy
and so on for many cases. Well the WRC itself tests strategies with different grades of complexity. I just wonder if there is something in the structure of the code itelf which can tell us when there is too much B=>C=>D=>E=>Trade over here.

Or maybe it is more about how many changes and hyphotesis have been added after the initial strategy is tested. If you begin with 5 hyphotesis its ok. The problem is when you start to develop "bridges" to get a pf>1 in the backtest and if there is a limit in the amount of "bridges" that someone can create.

Some ideas?