One strategy that ladders in a downtrend and grids in a range

Why this is hard on a bot platform

Because a DCA bot and a grid bot are two different products. They have separate configuration screens, separate capital allocations, and no awareness of each other. A DCA bot cannot become a grid bot any more than a spreadsheet can become a slideshow. So the platform's honest answer to this strategy is: run both, and turn one off yourself at the right moment.

The switch fires when you are looking,

not when the condition is true. If the regime changes at 4am, the wrong bot runs until you wake up.

The position is split in two.

The DCA bot's inventory is not the grid bot's inventory. When you switch, you either liquidate and re-enter — paying spread and tax for a bookkeeping problem — or you run a grid on a pair where another bot is also holding size it does not know about.

Your risk is unmeasured.

Neither bot knows the other's exposure, so neither can enforce a limit across both.

The interesting part is untestable.

The switching logic is in your head. It cannot be backtested, reviewed, or improved, because it was never written down.

That last one matters most. The regime call is the highest-value decision in the whole strategy, and it is the one part the platform does not hold.

The graph

One graph, one position, a switch in the middle.

The regime node is the strategy.

It reads a trend filter and a volatility measure and decides which branch is live. Because it is a node, its inputs are yours to choose — EMA separation, ATR relative to its own average, price holding inside the 20-bar range for N bars, or all three.

One position, across both branches.

The inventory the ladder built is the inventory the grid works. Nothing is liquidated to change behaviour, because nothing was ever two separate bots.

The switch is hysteretic, on purpose.

A naive threshold flickers — one candle over the line and the strategy flips, then flips back, churning fees. The switch node takes a confirmation period: the condition has to hold for N bars before the branch changes. That is a field on a node, not a thing you hope about.

One risk guard for everything.

Both branches route through it. Total exposure, daily loss, maximum position — enforced once, in one place, whichever behaviour is running, as in one risk limit across everything.

What you see while it runs

The regime node shows its current state and the values feeding it. When the graph switches from laddering to gridding, you see which input crossed, on which candle, and how long it had to hold before the switch accepted it — the glass box argument applied to the one decision that matters most here.

If it switches when you think it should not have, you have something to look at and something to change. When the switch lives in your head, a bad call is just a bad day you cannot review.

Variations on the same shape

Three regimes instead of two.

Trending up, trending down, ranging — a third branch that works breakouts.

Volatility-only switching.

Skip the trend filter; use ATR alone. Grid while volatility is low, ladder when it expands.

Switch the parameters, not the behaviour.

Same grid on both branches, wider spacing in high volatility. A smaller change, same structure.

Session-aware.

Different behaviour during Asia hours than during the US open — the regime node reads a clock as easily as a price.

Keep reading