Single Position vs Grid: Why Fixed Risk Beats Averaging Down

18 十一月 2026, 06:12
Xiang Jun Jiang
0
2
Most losing expert advisors I tested over the last two years fail for the same
reason, and it is not the entry signal. It is how they add to a losing position.

I ran the same gold (XAUUSD) data through two families of systems:

- grid / martingale systems that add positions as price moves against them
- single-position systems that open one trade, apply a hard stop, and wait

The difference is not subtle. One group survives two years. The other destroys
the account even when the trade count and the "win rate" look impressive.
# What the numbers look like
==============================

In one test, a martingale grid EA produced 13,208 orders over 23 months. It
started with 0.01 lots and ended with positions above 0.4 lots. The balance went
from 1,000 to 27 — a 97% loss. The same EA with the multiplier fixed at 1.0 and
the level count capped lost only 18%. Nothing about the entry changed. Only the
position sizing.

Another system, a 500-level grid, reported a very high win rate for months and
then lost 83% of the balance in the same period. When the whole strategy depends
on price returning, a single trend month removes the account.

By contrast, single-position systems with a fixed 0.01 lot and a hard stop can
be evaluated honestly: win rate, average win, average loss, drawdown. A system
with a 55% win rate, average win of 20 USD and average loss of 14 USD is a
business. A grid with a 90% win rate and an unbounded tail is not.
# Why grid systems look good in backtests
==============================

1. MetaTrader test data has no real spread widening and no slippage. Grid EAs
   rely on both being absent.
2. The martingale sequence only fails in the tail. In a two-year sample you may
   see the failure only once — or not at all if the period is a ranging market.
3. Win rate is a comfortable metric precisely because it hides the losing tail.
# The simple rule I use now
==============================

- one position at a time
- fixed lot (or risk-based lot with a hard cap)
- a stop-loss on every trade, defined before entry
- a daily loss limit as a percentage of the day-start balance
- a drawdown guard that closes everything when the account is out of budget

These rules do not make a bad strategy good. They make the account survive long
enough to find out whether the strategy has an edge.
# Practical checklist before you run any EA
==============================

1. Backtest with fixed lot. If the EA cannot be configured to a fixed lot, treat
   it as a grid and size it as such.
2. Check the maximum lot used by the tester, not the initial lot.
3. Look at maximum simultaneous positions, not just trade count.
4. Compute the cost of the spread: 30 points on gold at 0.01 lot is 0.30 USD per
   trade. Multiply by the number of trades per month.
5. If the drawdown comes mostly from added positions, stop.

If you want to enforce the same rules on an EA you already own, a position
manager can do it: suggested lot by risk, automatic stop-loss attachment,
break-even, trailing stop, daily loss limit and an equity drawdown guard.

Risk first. Edge second.