Grid EA Risk Settings Explained: Control Your Worst Day

A grid EA’s settings panel looks like a menu of preferences. It is actually a set of engineering tolerances, and the difference matters because preferences can be chosen by taste while tolerances must be chosen by arithmetic. Somewhere in that panel, usually spread across reassuringly technical names, are the seven decisions that jointly determine the worst day your account can have. This guide walks through each one, states what it really controls, and shows you how to set it from the only document that should govern it: your written loss limit, applied to the worst basket your settings legally permit.

If you run, or are tempted to run, any grid system — including the famous ones with long public records of both smooth gains and deep drawdowns — this is the page that decides which of those two stories your account gets to tell. The strategy’s entries are the vendor’s department. The tolerances are yours.

Setting one: grid spacing, the tempo of danger

Spacing is the distance in points between successive positions as price moves against the basket. It controls how quickly a move fills your grid, and therefore how much of your exposure is committed before the market has proven anything. Tight spacing feels productive: baskets complete quickly, profits recycle often, and the equity curve purrs. It also means an ordinary trend day can open most of your permitted positions before lunch. Wide spacing is slower and duller, and it reserves your capital for moves that have demonstrated real extension. A practical anchor is the instrument’s own volatility: spacing expressed relative to the average true range on your timeframe scales the grid to the market’s breathing, wider for gold and the volatile crosses, narrower for the sleepy majors. Whatever anchor you use, spacing should be set by the question “over what distance do I want my full basket deployed?”, answered from the instrument’s history of one-way runs, not from the profits of a ranging fortnight.

Setting two: lot progression, the exponential in the room

Progression is how position size changes from level to level: flat, additive, or multiplicative. This is the single most consequential number on the panel. Flat progression grows exposure in a straight line you can compute in your head. A multiplier — the martingale family — grows it geometrically, and geometric growth has a character that linear intuition chronically underestimates: the last two levels of a multiplied grid routinely outweigh all the earlier levels combined, so the basket’s fate, your margin, and your account’s survival concentrate into the deepest, most adverse positions, entered at the worst moment, at the largest size. Multiplied grids rescue themselves on smaller retracements, which is the entire seduction, and the price is that the rescue arithmetic and the ruin arithmetic are the same arithmetic. If you run any progression above flat, every other setting in this guide becomes load-bearing, and the paper stress test stops being advice and becomes the price of admission.

Setting three: maximum positions, the cap that defines the tail

The maximum number of grid levels is where the strategy’s loss stops growing by design, and it must be a number you chose. An uncapped or generously capped grid has, in effect, chosen your broker’s margin call as its risk manager, and brokers perform that role without sentimentality. Set the cap together with spacing and progression as one equation: cap × deployed distance × sizes = worst basket. Note what happens at the cap, because EAs differ: some stop adding and wait, some close the worst position, some close everything. Waiting is a bet that the range returns before margin runs out; the others convert the tail into a defined loss. None is universally right, but you must know which one you own, in writing, before the day it triggers. Compare this with Stop Loss Take Profit MT4 Indicator | Free Risk Control before you decide.

Setting four: the basket exit, where profits and rescues are decided

Baskets exit by rules: a combined profit target, an average-price retracement, a time limit, or partial closes. The exit rule shapes your whole result distribution. A small basket target recycles capital quickly and keeps exposure young, at the cost of many tiny wins that must outrun the rare disaster. A patient exit holds baskets longer, earns more per rescue, and spends more time in the danger zone with margin committed. Time-based exits deserve more respect than they get: a basket that has been wrong for three weeks has provided evidence about the regime, and a rule that cuts it accepts information the price-based rules are designed to ignore. Whichever exits you run, rehearse them against the worst historical stretch your pair offers, because the exit rule is where “temporary drawdown” either becomes a managed cost or an open-ended one.

Setting five: the equity stop, your signature on the whole design

Somewhere in the settings there must be a line that ends the experiment: a maximum floating loss, in money or in percent of equity, at which the EA closes the basket or stops opening new positions. If the product offers one, set it from your loss budget — the amount you decided, when calm, represents this strategy’s total tuition. If the product does not offer one, you are the equity stop, and you should write down the number and the action now, because the person reading the plan during a drawdown will be frightened, tired and extremely creative with exceptions. An equity stop converts a grid’s open-ended tail into a defined one. It will occasionally fire just before a recovery, and that frustration is the premium; pay it gladly, because the same rule is what makes the system’s worst case a number instead of a narrative.

Setting six: filters that keep grids out of their killing fields

Grids die in trends and in news, and the filter settings are how a configuration declines those invitations. A trend filter — even a crude one based on a higher-timeframe moving average or ADX — that suspends new baskets when the market is marching will cost some ranging profits and save some accounts; that is a trade worth pricing deliberately rather than discovering. A news filter or a simple time block around major releases addresses the minutes when one-way moves are manufactured. A spread ceiling matters more than it looks: grids add positions at the worst moments by design, and adding them at crisis spreads overpays for every level. Session rules play the same role for instruments like gold, whose character changes violently by the hour. Filters are frequently dismissed as performance-killers by backtests run on calm data. Run the test on the pair’s worst quarter instead and see what they earn.

Setting seven: the account itself — leverage, margin and stop-out

The final settings are not in the EA at all; they are in the account you point it at. Leverage decides how much basket your balance can carry before margin complains. Your broker’s stop-out level decides where forced liquidation begins, and a grid designed around surviving to a 50 percent stop-out lives a different life on an account that liquidates at 80 or at 20. Margin mode and hedging permissions decide whether the basket can even be constructed as designed. Treat the account specification as part of the configuration, record it on the same settings sheet, and re-run your basket equation whenever the broker changes any of it, because brokers do, usually by email, in language nobody reads, a week before it matters.

The account mode underneath: hedging versus netting

One specification decides whether your grid can even be built as designed, and it sits in the account, not the EA: whether the account is a hedging account or a netting account. On a hedging account, each order becomes its own position with its own ticket, so a grid’s eight levels exist as eight separate positions that the EA can count, adjust and close individually, in any order. On a netting account — the standard arrangement on much of MT5 — every order on the same symbol merges into a single position whose size and average price update with each fill. The grid does not disappear on a netting account; it dissolves into one position with a moving average entry, and every EA behaviour written for distinct positions quietly changes meaning. Per-level take profits have nothing to attach to. Closing “the oldest position” closes a slice of an average. Partial basket exits become arithmetic on a blended price that no level ever actually traded at. Worst of all, the change is silent: the EA runs, the chart populates, and the divergence between the designed strategy and the executed one only becomes visible in a drawdown, when the basket management you rehearsed does not occur. You can also check The Best ATR Multiplier for Your Stop Loss: A Working Guide for a second view.

The practical rule is plain. Establish which mode your account uses before the EA is attached — it is visible in the terminal and in your account specification — and use the mode the EA’s documentation assumes. If the documentation does not say, that silence is itself a finding about the product’s maturity, and a demo on the actual account type stops being a courtesy and becomes the test. Some brokers offer hedging and netting variants of otherwise identical accounts, and switching modes after positions exist is rarely possible, so this decision belongs at account opening, alongside the leverage and stop-out choices in setting seven. A grid is a structure of separate promises; the account mode decides whether the broker keeps them separate or averages them into one.

The two-pair temptation: correlation and the doubled grid

Sooner or later a grid owner has the same bright idea: the system works on one pair, so run it on two, and let the baskets diversify each other. The reasoning borrows the language of portfolios and ignores their arithmetic. Two grids diversify only if their bad periods arrive at different times, and pairs chosen by familiarity — EURUSD beside GBPUSD, the commodity currencies beside each other, anything beside gold in a dollar storm — share the driver that produces grid emergencies. A determined one-way move in the dollar does not visit one pair and spare the other; it marches across the whole family in the same hours, filling both grids level by level, in the same direction, against the same margin pool. The result is not two independent risks but one risk with two charts, plus a subtler failure: margin is shared, so the pair whose basket is merely deep can be liquidated by the pair whose basket is desperate, and the stop-out sequence will not consult your preference about which system deserved to survive.

If a second instrument is genuinely wanted, the burden of proof is specific. Show, from history, that the two instruments’ worst one-way runs in the grid’s trading direction did not coincide — not that their daily correlation prints a comforting number, because correlation measured across calm months is precisely the statistic that deserts you in the month that matters. Then run the basket equation jointly, as setting three demands for a single pair: both baskets at their caps, simultaneously, against one balance and one stop-out level. If the combined worst case exceeds the budget, the second grid is not diversification; it is leverage obtained by duplication, and it should be sized — or declined — as such. Owners who proceed anyway should at least run the two systems on separate sub-accounts or with hard per-system equity stops, so that one basket’s emergency cannot spend the other basket’s survival money without a rule being visibly broken.

Recovery mode: the settings that promise to earn the loss back

Many grid products carry a recovery feature under various names: after a basket closes at a loss, or after a drawdown threshold, the EA increases size — sometimes briefly, sometimes until the loss is recovered. Read this feature with complete clarity about what it is. It is a decision, made in advance by the vendor and accepted by you, to respond to the strategy’s worst outcomes by enlarging the next bet. The mathematics of recovery modes is the mathematics of every doubling scheme wearing a seatbelt: it converts many small, survivable losses into a high probability of recovery and a small probability of a loss taken at maximum size, at the moment the account is least able to bear it. The seatbelts vary — caps on the recovery multiplier, limits on the number of recovery cycles, a hard stop after which the feature disengages — and a recovery mode without all three is not a feature but a fuse already lit. A closely related guide is Neptune EA Review: Performance, Settings & Results.

Two disciplines govern any use of it. First, the recovery sizing must enter the worst-basket computation from the start: the largest basket the EA can build includes recovery-enlarged lots, so a configuration whose ordinary basket fits the budget and whose recovery basket does not is a configuration that does not fit the budget. Second, decide in advance what a completed recovery means. A system that claws back a loss and then continues at ordinary size has spent risk to stand still, and the temptation to leave recovery engaged “while it is working” transforms a bounded repair mechanism into the strategy’s permanent character. If you cannot state, in writing, the conditions under which recovery mode turns off, the honest setting is the one where it was never turned on, and no month’s results, however flattering, should talk you out of that honesty afterwards.

When to retire a configuration, not just stop it

Grid owners stop systems for two bad reasons — after one frightening week, or never — and the settings conversation is incomplete without the third option: retiring a configuration on evidence, by criteria fixed before the evidence arrives. A configuration deserves retirement when the market it was rented from has demonstrably changed character — when the pair’s behaviour across several quarters no longer shows the ranging personality the spacing and caps were engineered around, and the change is visible in the pair’s own statistics rather than in your frustration. It deserves retirement when the broker’s terms change beneath it: stop-out levels, margin requirements, swap, or execution quality altered enough that the basket equation you signed no longer describes the account you have. And it deserves retirement when the equity stop has fired repeatedly — not because stops firing is failure, but because repeated firing is the market reporting, at tuition prices, that the configuration’s tail arrives more often than its design assumed.

Retirement is not deletion. The settings sheet, the stress test, and the journal of what the configuration did are kept, because they are the data from which the next configuration — wider, smaller, filtered, or on a different instrument entirely — is designed rather than guessed. What retirement forbids is the two common resurrections: re-enabling the same settings after a calm fortnight because the account “can take it now”, and resurrecting them at larger size to recover what they cost. A configuration retired on evidence and revived on mood has not been given a second chance. It has been given the same chance, and the market keeps its receipts.

A worked basket, level by level

Abstractions survive contact with a grid better than with most strategies, so here is one basket built in full. The numbers are an illustration, constructed for the arithmetic: a major pair priced near 1.08, spacing of twenty-five pips, eight permitted levels, a starting lot of 0.01, and a standard contract where one pip is worth ten dollars per lot. First, run it flat. Eight positions of 0.01 lots total 0.08 lots. Let price fall two hundred pips from the first entry, filling every level on the way. The individual losses, at distances of 200, 175, 150 and so on down to 25 pips, sum to nine hundred pip-units at 0.01 lots — a floating loss of about ninety dollars. The basket’s average entry sits eighty-eight pips below the first entry, so price must retrace roughly one hundred and thirteen pips from the low for the basket to break even before costs.

Now change exactly one setting: a lot multiplier of 1.3 between levels. The lots become 0.01, 0.013, 0.017, 0.022, 0.029, 0.037, 0.048 and 0.063 — notice the last two levels outweigh the first five combined — totalling about 0.24 lots, three times the flat basket’s exposure. The same two-hundred-pip fall now carries a floating loss of roughly one hundred and ninety dollars, and the margin committed at 1:100 leverage is about two hundred and sixty dollars. On a one-thousand-dollar account, that single ordinary trend — not a crisis, not a news spike, just two hundred pips in one direction — has produced a nineteen percent floating drawdown with the grid fully deployed and no levels left. In exchange, the multiplied basket’s average entry sits about one hundred and nineteen pips below the first entry, so it breaks even after a retrace of only about eighty-one pips, against the flat basket’s one hundred and thirteen. There, in one worked example, is the entire seduction and the entire danger of progression: thirty-two pips of rescue purchased with a doubled loss, tripled exposure, and an account a fifth underwater. Neither configuration is “right” in the abstract. But only one of them was chosen by arithmetic the owner had seen in advance, and that — not the outcome of any single basket — is the difference this whole guide has been pointing at. For more context, see Best Currency Strength Meter Indicator for MT4/MT5 Free.

The rent a long basket pays: swap, spread and time

Basket arithmetic usually stops at floating profit and loss, but a basket held for weeks pays rent while it waits for its retracement. Each open level that crosses the daily rollover incurs swap, and the direction grids trade — against the move, patiently — often places the basket on the swap-paying side of the pair, so the carrying cost accumulates on every level, every night, at crisis-era sizes. A basket whose floating loss looks static on the chart is in fact deteriorating nightly by the swap line, and the retracement it needs grows correspondingly: break-even is a moving target receding at the rate of the rent. Spread is the second quiet charge. A grid adds its deepest, largest levels at the worst moments by design — in fast markets, at widened spreads — so the average entry quality of the basket’s most consequential positions is systematically worse than the backtest’s fills, which were typically computed at polite, average spreads. None of this appears in the settings panel as a single number, which is exactly why it must be estimated on paper: take the worked basket above, hold it for twenty trading days at your broker’s actual swap for that pair and direction, and add the result to its cost of rescue. Configurations that only survive if baskets resolve quickly are telling you something about themselves; a time limit on basket age, unfashionable as it is, exists precisely to cap the rent.

Touching it by hand: rules for intervening in a live basket

Every grid owner eventually faces the screen, the deep basket, and the urge to help. The plan written in calm weather must govern this moment, because the improvised interventions available are few and each is dangerous in a characteristic way. Adding manual size to “help the average” is the progression fallacy performed by hand, at the worst emotional moment, usually without recomputing the basket’s worst state. Closing the deepest level to relieve margin feels like surgery and functions like amputation of the rescue: the loss is banked at maximum size while the remaining basket’s average price retreats further from reach. Closing the whole basket at the point of maximum pain converts a designed drawdown into a realised one at the exact price the design existed to survive. This guide’s position is not that intervention is never right — an equity stop reached, a broker term changed, a configuration retired on evidence are all interventions, and honourable ones. It is that the legitimate interventions are the ones specified in advance, triggered by numbers rather than by the feeling of watching. Write the short list now: the floating loss at which you close everything, the margin level at which you stop the EA from opening further levels, and the news events during which the system is simply off. Three lines, decided by the calm version of you, are the entire difference between managing a grid and being managed by one. Rehearse them on demo at least once — deliberately let a test basket reach the stop line and execute the written response — because a rule performed once in rehearsal is a rule that exists, while a rule only ever read is a hope with formatting. The demo rehearsal also exposes the practical frictions the plan must survive in reality: the equity stop’s exact trigger behaviour in your EA, the time it takes to close a full basket in a fast market, and whether your broker’s execution at that moment resembles the calm-market fills the rehearsal began with. Discovering those frictions at demo prices is, like everything else in this guide, the cheapest version of the lesson on offer.

The paper stress test that ties the seven together

Before demo, before live, take your full configuration and one ugly hypothetical: your pair trends, without a meaningful retracement, for twice the distance of its worst recent one-way run. Compute, level by level, the positions opened, the total lots, the margin consumed at your leverage, and the floating loss at the far end. Compare that loss with your written budget and your broker’s stop-out. Three outcomes are possible: the configuration survives inside budget, which earns it a demo; it survives only via the equity stop, which tells you the stop is doing the real risk management and must be honoured absolutely; or it does not survive at all, in which case the settings — not the market, not luck — are rejected on paper, at the lowest price this lesson is ever offered. The stress-test worksheet we attach to our own grid material exists to make this computation a form rather than a project.

Grid trading is neither magic nor madness; it is a way of renting ranging markets with a known, occasionally severe, tail cost. The seven settings are where you negotiate that cost in advance, in numbers, while negotiation is free. Set them from your loss limit, write them down, and let a good month tempt you into changing exactly none of them.