Have been made aware of a grid where you start with a baseline value, a "zero point", and only takes buys if market moves x pips above, and only take sells below the zero-point.
this way, you are only taking one position (not 2 as per Shakones comments), and this position is in the direction of the move.
the downside is that at some point the move, whether up or down, will accumulate aggregate losing positions as the price starts to return to the zero-point. and calling off the losing positions can only come if the opposing positions on the other side of the zero-point are as strong. so here, you're still dependent on some form of median averaging.
since the problem of grid-trading hinges around margin, and the need for deep pockets to ride out the outlier trades, and hedge funds and professional money managers are the ones with deep pockets, and no hedge fund or pros uses grids, maybe that says it all as to their (lack of) utility.
the need to trade grids relies on the concept of return-to-median. the failure is due to the almost exponential accruing of liabilities when a trend occurs.
as part of my research, I came across Bird Watching In Lion Country, which seems to be a sort of "intelligent" grid.
it was certainly an interesting exercise to learn about grids.
EDIT: of all the variants I did analyse, across a number of forums, nothing seemed to work "properly". What I mean is, conceptually, the grid would "break" at some point, and successful grid-traders added some component of closing a grid when "x" amount made, of other reason to make profits. I couldnt find anything that was valid in the long term.
The one useful thing I learned was that counter_violents thread about not taking losses may have some value, as I tried a simple Stochastics buy/sell method, and if I left the losing trades open, I could nett them off at some point against a winning one to reduce margin reqs. However, commissions and overnight charges were not accounted for, so my conclusions are indeterminate.
NB: I was buying when stochs went over 80, and selling when stochs went below 20; not the other way around. I was effectively using stochs to anticipate breakouts, not reversions to mean. I used a time-frame of 15-mins and 30-mins, and used stochs settings where new trades were allowed to exist for days at a time, not intra-day.
(I also didnt accumulate positions at the same price; new signals had to be "better" by at least 50-100 pips, or the new position was ignored)
