Types of forex Expert Advisors: the nine EA categories and how to tell them apart from the numbers

Why the category matters more than the sales page

Every forex Expert Advisor on sale describes itself as intelligent, adaptive and low-risk. Almost none of them describe what they actually do. Yet the type of automated trading strategy decides nearly everything you care about: how the equity curve will look, how it loses, whether a 90% win rate is a feature or a warning, and whether a good year can be erased in an afternoon.

After running and watching a lot of MT4 and MT5 trading bots, I have found that nine categories cover practically every EA you will meet. They are not equally dangerous. Three of them account for most of the blown accounts I have seen, and all three produce the prettiest curves right up to the end. Each type leaves fingerprints in the trade statistics, and on a live-synced or broker-verified account you can read those fingerprints yourself, whatever the vendor calls the product.

The nine types in one table

Here is the taxonomy, with the single fastest tell for each. The linked articles go deep on each one.

• Trend following: win rate 30–45%, winners two to four times the size of losers, holds for days, long loss streaks. Safe if you can stomach being wrong most of the time.
• Mean reversion: win rate 60–80%, losers bigger than winners, holds hours, shallow frequent drawdowns. Safe only when stops are real.
• Grid: many same-symbol positions closed in the same minute, smooth balance curve, ugly equity curve. Dangerous.
• Martingale: lot size jumps after every loss, win rate above 85%, one loss wipes months. The most dangerous type, and the most common in "free EA" downloads.
• Scalping: average hold under an hour, tiny average win, dozens of trades a week, lives or dies on spread. Honest but fragile.
• Breakout: trades cluster at session opens, win rate around 40%, fat right tail. Honest, streaky.
• News trading: a handful of trades a month at exact minutes, enormous slippage sensitivity. Demo results mean nothing.
• Hedging: simultaneous buys and sells on one symbol, swap bleed, losses parked rather than taken. Usually a grid in disguise.
• Arbitrage and latency: near-perfect win rate, holds seconds, broker terminates the account. Not a strategy you can keep.

How the ConfirmedTrades review classifies an account

The AI strategy analysis on every account page runs a set of deterministic checks before any narrative is written. Structural patterns are checked first: if lot sizes rise by more than 40% straight after losses in a meaningful share of cases, or the win rate is at least 65% with an average loss at least twice the average win, the type is Martingale. If 35% or more of trades belong to baskets of three or more same-symbol trades closed inside the same minute, the type is Grid. Only after those two does holding time decide: under an hour average is Scalping, over 48 hours is Trend / Swing, up to 24 hours is Day trading, and the gap between is Mixed. Fewer than 15 trades gets Insufficient data.

Notice what that ordering means. A grid EA that holds positions for days is still classed as Grid, because the basket structure is what will hurt you, not the timeframe. Breakout, news, hedging and arbitrage are not separate labels in the review; they show up as Day trading or Scalping with telltale patterns in the hour distribution, open positions and execution numbers, which the individual articles explain.

Reading the type from six numbers

Open the Trade statistics section and pull out win rate, profit factor, average win, average loss, average trade length and trades per week. Those six numbers nearly always identify the category before you read anything else.

A 38% win rate with an average win of $180 against an average loss of $65 and a 3-day average hold is a trend follower. A 74% win rate with $45 wins and $120 losses held for 5 hours is mean reversion. An 88% win rate with $12 wins, a $310 average loss and lot sizes that vary tenfold is martingale, full stop. A 92% win rate with 30-second holds and a profit factor near 1.2 is either a scalper on a generous broker or a latency bot about to lose its account.

Then cross-check with the Activity distributions section. The holding-time histogram uses eight buckets, from under a minute through 1–5 minutes, 5–15, 15–60, 1–4 hours, 4–24 hours, 1–7 days and over a week. A scalper lives in the first four buckets; a swing EA lives in the last two; a grid often has a strange two-humped shape, with quick basket closes and a long tail of positions that sat underwater for a week. The weekday and hour charts show whether an EA only trades a session open (breakout), a few exact minutes a month (news) or all day (scalping, grid).

Which types are dangerous, and why it is always the same reason

Grid, martingale and most hedging EAs share one mechanism: they convert a small, frequent gain into an occasional catastrophic loss by refusing to take losses. The balance curve climbs in a staircase because only winners get closed. The losers sit open, growing, until the market turns and the basket closes for a small profit, or it does not and the account is margin-called. A vendor can honestly show you a year of green months from a strategy that is going to lose 100%.

The mathematics is unforgiving. Take a martingale with 1.8× lot progression starting at 0.1 lots. After six consecutive losses the seventh position is 0.1 × 1.8⁶, about 3.4 lots, 34 times the first. Total open exposure across the sequence is roughly 7.5 lots. If the win rate of each individual entry is 70%, six straight losses happen once in about 1,370 sequences. That sounds rare until you realise the EA runs a new sequence every time it wins, so over 2,000 trades you should expect to see it once or twice. One is enough.

ConfirmedTrades exposes this in three places no screenshot can fake. The growth chart folds open equity into the index, so a drowning basket shows as a dip before any trade closes. Glance shows Max DD (equity) next to Max DD (balance) from a ledger that survives deleting and reconnecting the account. And the review raises high-severity flags for martingale progression, for an average loss over 2.5× the average win, and for stop-losses on fewer than 15% of trades. Maximum drawdown explained covers the balance-versus-equity point.

Why the risky types dominate free downloads

Search for a free grid EA or a free martingale EA and you will find hundreds. Search for a free trend-following forex robot with a verified three-year record and you will find almost none. A martingale or grid can be written in an afternoon and will show a 95% win rate on any demo within a week, which makes it perfect for lead generation: give it away, collect email addresses, upsell the "pro" version or an affiliate broker account. A trend follower loses most of its trades and needs months to prove itself, so it makes a terrible free sample. Free is not the problem; the problem is that the types cheapest to make look good are the ones that hurt the most.

What the Score does with each type

The Score does not know the category, but it reacts to the shape. Profitability (25 points) comes from profit factor, which martingale and grid systems often score well on until the blow-up. Risk control (25 points) is mostly maximum drawdown plus chance of ruin and recovery factor, which is where basket systems lose points once the equity ledger records their first deep dip. Consistency (20 points) rewards profitable months and punishes symbol concentration, which hurts single-pair scalpers and gold grids. Longevity (15 points) needs about six months and 200 trades, so a scalper gets its trade count quickly while a swing EA has to wait. Trust is the last 15.

So a trend EA with a 22% drawdown can sit at a B while a 97%-win-rate martingale sits at an A+ for a year, then drops to an E in a week. Read the breakdown, and read the category.

A type-first checklist

• Read the strategy type in the AI review, then the six numbers in Trade statistics. Do they agree with what the vendor says the EA does?
• Compare Max DD (equity) with Max DD (balance) in Glance. A big gap means losses are being held, not taken.
• Check the Open positions section. Several same-symbol positions with mounting floating loss is a basket, whatever the product name.
• On deeper plans, look at Risk Lab's chance of ruin ladder and Performance depth's deepest drawdowns table with their durations.
• Use By magic number when several EAs share an account, because one safe strategy can hide a dangerous one next to it.
• Then work through how to vet an EA and spotting fake EA results.