MidasIn short: Grid trading is only as safe as its sizing: it earns small gains while price stays in a range and has no built-in cap on loss when price trends out of it.
This applies to grid expert advisors on MT4 and MT5 and to the grid bots and martingale or DCA-style bots offered by exchanges and platforms such as Binance, Bybit, OKX, 3Commas, Veles and Pionex. Setting names differ; the arithmetic is the same. The examples use made-up round inputs. They are not a forecast and not advice.
On this page (8)
Is grid trading safe?
Not automatically: a grid has no loss limit built in, so its safety is whatever the order size, order count, multiplier, step and leverage make it. A grid keeps buy and sell orders at fixed price steps and banks one step each time price swings back. Inside a range that adds up to many small gains. When price leaves the range and keeps going, the grid keeps adding orders into the move, so the position is largest exactly when the loss is largest. A stop-loss, if the platform offers one, is an add-on that books the loss instead of removing it.
When does a grid work, and when does it blow up?
A grid works while price oscillates inside the range it covers, and blows up when price trends through the range without pulling back, especially with leverage or growing order sizes.
| What price does | What a long grid does | Result |
|---|---|---|
| Swings inside the range | Buys dips, sells bounces | Every completed swing banks one step × order size |
| Rises through the top | Everything bought is already sold | Gains stop; nothing is lost, but the rise is missed |
| Falls through the bottom without a pullback | Every level fills, the position is full | Loss grows with each step: in a straight line with equal orders, faster with a multiplier |
| Gaps over several levels | Orders fill together at worse prices | Loss larger than the plain calculation |
For equal orders the arithmetic is short. Gain per completed swing = step × order size. With n levels filled and price X below the first one, loss = order size × (n × X − step × n × (n − 1) / 2).
Example: price 100, ten levels at 100, 98 … 82 (step 2), 0.1 coin per level. One swing gains 2 × 0.1 = $0.20. A fall to 80 (X = 20) loses 0.1 × (10 × 20 − 2 × 45) = 0.1 × 110 = $11.00. That is 11.00 / 0.20 = 55 completed swings spent on one 20% slide.
What decides the risk of a grid?
Five settings decide it: order size, number of orders, multiplier, step and leverage; together they fix the total position, the range covered and the margin needed.
| Setting | What it sets | Formula |
|---|---|---|
| Order size (money per order) | Scale of everything: loss is proportional to it | first order q |
| Number of orders N | How many levels can fill | total = q × (m^N − 1) / (m − 1) |
| Multiplier m | How much bigger each next order is (1 = equal, 2 = doubling) | same total formula |
| Step s | Gap between levels, so the range covered | (N − 1) × s |
| Leverage L | How much of the position's value must sit as margin | margin = position value / L |
Leverage does not change what a given position loses per dollar of price move. It changes how much of your deposit stands behind that position, and so how early the exchange or broker closes it. The multiplier is the setting that surprises people: with m = 2 the total is q × (2^N − 1), so ten orders are 1,023 times the first order, not 10 times. More in grid step, multiplier and order count.
Worked example: same ten levels, three multipliers
On the same ten levels, the same first order and the same fall, the multiplier alone decides whether the loss is $11, $64 or $407. Inputs: price 100, levels 100 … 82, first order 0.1 coin, deposit $1,000, cross margin, fees and maintenance margin ignored.
| Multiplier | Total coins (= $ lost per further $1 down) | Cost of all 10 orders | Loss at 80 | Loss at 70 | On a $1,000 deposit |
|---|---|---|---|---|---|
| ×1.0 | 1.0 | $91 | $11 | $21 | Paid in full at 1×; worst case −$91 |
| ×1.5 | 11.33 | $971 | $64 | $177 | Paid in full at 1×; worst case −$971 |
| ×2.0 | 102.3 | $8,591 | $407 | $1,430 | Needs about 10.8× leverage to place all 10 orders (the margin equals the deposit at 8.6×, before the $202.6 floating loss); deposit gone at 74.2 (−25.8%) |
The idea of a grid is the same in all three rows. What changes is how much position the ladder piles up and how much leverage it takes to carry it. The ×1.5 row is slow, but a fall to zero would still take nearly the whole deposit. The full ×2 arithmetic is in how much deposit a grid bot needs.
A checklist before you start any grid
- Total volume. Compute q × (m^N − 1) / (m − 1). If you cannot state this number, you cannot state the risk.
- Worst case in money. Multiply the total position by a fall you consider possible, not the fall you expect.
- Funding. The money or margin for every order must fit in the deposit with room to spare, or the bot stops adding orders or is closed out before the last level.
- Range against history. Compare (N − 1) × s with moves the instrument has actually made.
- Stop. Is there a stop-loss or a maximum-loss rule? Without one the plan is "price comes back".
- Leverage. Lower leverage leaves more of the deposit behind the same position.
- Test. Put your numbers into the calculators below before any money is used.
What can go wrong
- The example is kinder than a market. Fees, funding, spread, slippage, gaps and maintenance margin are not in it, so a real liquidation comes earlier.
- No probability is stated. The table says what a given fall costs, not how likely it is.
- No setting removes the risk. A smaller step, more orders or lower leverage change how far price must go, not whether it can.
This is arithmetic of the settings, not a forecast. A grid with leverage carries risk: with too small a deposit a stop-out is possible. Size your deposit with a safety margin. Nothing here is individual investment advice.
Check your own numbers
- Grid bot risk calculator for exchange and platform grid bots.
- DCA bot calculator (safety orders) for base and safety orders.
- Gold grid calculator for MT5 gold grid EAs with lots, cent accounts and stop-out.
- Midas Grid emulator replays a gold grid EA's settings day by day on real XAUUSD ticks since January 2025. It does not test crypto bots.
Related: grid bot vs DCA bot, martingale blow-up arithmetic, margin call vs stop-out.
FAQ
Is grid trading safe for beginners?
No setting makes a grid beginner-safe. Equal orders, few levels and low leverage make the loss per step small, but a grid has no loss limit of its own, and a long fall still fills every level. Start by computing the total position and the worst case in money.
Can a grid bot lose all my money?
Yes. With leverage, a long grid is closed out when the floating loss uses up the margin. Without leverage, the position can still lose the full cost of its orders if price goes to zero. In the example, the ×1.0 ladder can lose at most $91 and the ×2.0 ladder reaches a $1,000 loss at a price of 74.2.
Does a grid bot work in a trending market?
It earns while price swings and keeps losing when price trends against it. In the example, a 20% slide costs $11.00 against $0.20 per completed swing, which is 55 swings. A trend in the grid's favour, such as a rise for a long grid, ends the gains once the range is passed.
Is a grid safer than a martingale or DCA bot?
Per setting, usually yes. Equal orders add position in a straight line, while a multiplier above 1 multiplies it with every order. A grid with a multiplier above 1 follows, in effect, a martingale size rule, whatever the product calls it. See grid bot vs DCA bot.
What is the safest grid setting?
There is none; there are criteria. A small total position relative to the deposit, a multiplier of 1, low leverage, a range that covers moves seen in history and a stated stop make the loss smaller and slower, but never zero.
Does a stop-loss make grid trading safe?
It caps the loss at the stop level and ends the grid. The loss is booked, not avoided, and a gap through the stop can fill worse than the stop price.