QuoteDisliked1) Let's say that the goal is to either keep the balance at break even or increase the balance. Therefore, a drop in the account balance is not allowed.
2) Place a random trade (time zero; t0) in the random price series (price zero; p0)
As the price series develop (p+n); What happens to you entry at t+n ?
What needs to occur so that 1) is never violated ?
Thus you have to accept 1) will be violated.
Let's forget the exact Monty Hall problem which requires that you move the market to force it to giveback some clues from its reaction. Let's try and see if there is a way to use conditional probabilites to acquire a more educated guess about the future direction is order to play some sort of marty (even though I'd prefer a snowball scheme). Of course if the price movement is a random walk no information can be acquired. We need a better model of the market movement than the RW. We must therefore accept that the past price action contains information about the future. The model needn't be exact but we need to know how much wrong it can be.
For instance here I posted that market seems to exhibit short term momentum: https://www.forexfactory.com/showthread.php?p=8464204#post8464204
Long ago I posted results showing that the trends in market data are more stable that the trends in a randomly generated market --using bootstrap with the same tick return--. I don't remember the thread.
From this model, either the data (current price) is within the model tolerence and you can average down or the hypothesis is invalidated and you flip with a higher size to recover. Your nemesis is a flat trend with small swings that will force you to flip all the time. => you need a model of volatility (GARCH?) to adapt the grid size.
All in all the job is to identify the trend, the current momentum and the volatility is near real-time. I call this the state of the market. You also have to accept that this state will only vary slowly to have the time to profit .
No greed. No fear. Just maths.