A grid system opens additional positions at fixed intervals as the price moves against the first one. A martingale system opens additional positions of increasing size after each loss, so that one winning trade recovers all of the previous losses plus a small profit. Both refuse to accept a loss on any single trade, and both produce a record that looks excellent until it doesn't.
How a grid works
Buy at 1.0900. Price falls to 1.0850; buy again. Falls to 1.0800; buy again. If price now rises to 1.0870, the basket is closed at a small overall profit. The trades never lose individually, because none is closed until the whole basket is in profit. If price keeps falling, the basket keeps growing, the floating loss grows with it, and the margin runs out at the point of maximum loss.
How a martingale works
Risk one unit. Lose. Risk two units. Lose. Risk four, then eight, then sixteen. The first win recovers everything. Six consecutive losses require the seventh bet to be 64 units, and eight require 256. Every strategy has runs of eight losses. The martingale's record shows years of small, steady gains, then the run that no account can fund.
Why the record looks good
Both methods produce a rising balance curve, a win rate above 90 per cent, and no losing months, because losses are held open rather than taken. The equity curve tells the truth: it shows the floating loss growing during every adverse move. A record that shows balance only, or shows equity with long periods well below balance, is showing one of these methods.
The names they hide behind
Recovery, smart recovery, drawdown recovery, averaging, cost averaging, dollar-cost averaging, basket trading, basket close, hedging (when it means holding a losing trade and opening the opposite one), zone recovery, and "adding to a position". An EA that does any of these is doing one of these. The absence of a stop loss on each trade is the confirming sign.
A grid EA runs for twenty months with 1,140 trades, a 97 per cent win rate and a balance curve rising 60 per cent. In month twenty-one the euro falls 700 pips in three weeks. The grid opens fourteen positions on the way down, the floating loss reaches 85 per cent of the account, and the broker's stop out closes everything at the low. The account is down 88 per cent. The twenty months were real, and they were never worth what the curve said.
The question to ask of any EA is whether every trade has its own stop loss and is closed at it. If yes, losses are taken as they occur and the record means what it shows. If no, the record is a loan against a loss that has not yet been called in. There is no third case, and there is no setting on a grid or martingale EA that makes it safe, because the method is the problem.