Glossary

Delegating trade decisions to explicit rules executed by software. The edge comes from the rules and their validation, not from the automation itself — a bad strategy loses money automatically.

A volatility indicator measuring the average range of recent bars. EAs commonly use ATR to scale stop losses and position size so risk stays constant as volatility changes.

Running a strategy against historical price data to see how it would have performed. A backtest is a hypothesis check, not a promise — results depend heavily on data quality and assumptions about fills.

Moving the stop loss to the entry price once a trade is sufficiently in profit, so the trade 'can't lose'. It feels safe but often converts winners into scratches by exiting on normal pullbacks.

The round-trip time between your terminal sending an order and the broker confirming the fill. Above a few hundred milliseconds, scalping strategies quietly lose their edge to slippage.

Compound annual growth rate — the smoothed yearly return that turns a starting equity into the ending one. CAGR without drawdown context is marketing; the two numbers must be read together.

Automatically mirroring another trader's positions in your own account. The copier inherits the signal's drawdowns plus extra slippage and latency — and often joins at the equity peak.

How strongly two instruments move together. Running 'diversified' EAs on highly correlated pairs is secretly one big position — drawdowns arrive on all of them at once.

An alternative trading platform whose automation layer (cBots) is written in C#. Popular with developers who want modern tooling and cleaner APIs than MQL, at the cost of a smaller ecosystem.

Demo accounts fill instantly at quoted prices with no liquidity constraints; live accounts slip, requote, and partially fill. The divergence between the two is the most common reason 'proven' EAs disappoint.

The decline from an account equity peak to the following trough. Drawdown — not profit — is what decides whether you can psychologically and mathematically keep running a system.

The schedule of macro data releases (NFP, CPI, rate decisions) that move markets. EAs use it to pause trading around high-impact events whose slippage and spread spikes break normal assumptions.

The plot of account equity over time. Its shape matters more than its endpoint: smooth curves often hide tail risk, while honest systems show visible, survivable drawdowns.

An account-level circuit breaker that halts all trading when equity falls by a set amount or percentage. It protects against the failure mode per-trade stops can't catch: a strategy that has stopped working entirely.

The average amount a system wins or loses per trade: (win rate × average win) − (loss rate × average loss). It is the single number that tells you whether a strategy has an edge at all.

A program that automates trading decisions on MetaTrader platforms, written in MQL4 or MQL5. EAs range from simple alert scripts to fully autonomous systems with their own money management.

Running a strategy on live market data (usually on a demo account) after development is finished. Because no historical fitting is possible, forward tests are far harder to game than backtests.

An optimization method that 'evolves' parameter sets through selection, crossover, and mutation instead of testing every combination. It finds good regions faster, and overfits those regions just as efficiently.

Placing layered buy/sell orders at fixed intervals to harvest range-bound chop. Grids print steady small profits in sideways markets and catastrophic losses when a strong trend finally appears.

Holding offsetting positions to reduce net exposure. In retail FX it often just doubles spread costs while locking in a loss — genuine hedging reduces a risk you actually have, not one you imagine.

Borrowed exposure that lets a small deposit control a large position. Leverage multiplies both gains and losses; at 1:500 a 0.2% adverse move can erase a fully margined account.

How much size the market can absorb without moving price. Liquidity is why demo fills are fiction: real orders move thin markets, and news events drain liquidity exactly when strategies need it most.

The standardized trade volume unit: a standard lot is 100,000 units of base currency, with mini (0.1) and micro (0.01) lots for smaller sizing. Correct lot calculation from stop distance is the core of position sizing.

A broker's demand to deposit more funds when account equity falls below the required margin. If unmet, positions are force-closed at the worst possible prices — the mechanical end of overleveraged accounts.

Doubling (or multiplying) position size after each loss to recover with one win. The equity curve looks smooth for months, then a single losing streak wipes the account — the risk is hidden, not removed.

The largest peak-to-trough equity decline in a test or track record. A single backtest's max drawdown is usually optimistic — Monte Carlo resampling shows the realistic worst cases.

The 2005-era platform that still hosts a huge share of retail forex EAs. Single-threaded tester, M1-interpolated backtests, and an enormous legacy library of indicators and bots.

MetaQuotes' newer multi-asset platform with a faster multi-currency strategy tester, real tick support, and MQL5. Migration from MT4 is one-way for code but not for habits.

The MT4/MT5 report metric estimating how much of a backtest used real versus interpolated price points. '90% modeling quality' means the tester guessed intra-bar movement — a materially different test from real ticks.

Resampling a strategy's trade list in random orders (or with random slippage) many times to build a distribution of outcomes. It reveals the realistic range of drawdowns a single backtest path hides.

The C-like language for writing EAs and indicators on MetaTrader 4. Mature with enormous existing codebases, but its order model and tooling are dated compared to MQL5.

The object-oriented language for MetaTrader 5, with a faster multi-currency tester and a position/deal order model. Most new EA development — and most AI code generators — now target MQL5.

EA logic that blocks new entries (or closes positions) around scheduled high-impact news. Filtering too little exposes you to spike slippage; filtering too much removes profitable volatility.

Systematically testing parameter combinations to find the best-performing set. Optimization is a useful search tool and the fastest way to overfit — the two are separated only by out-of-sample discipline.

Tuning a strategy's parameters until it matches historical noise instead of a real market edge. Overfit systems show beautiful backtests and then fail immediately in live trading.

Simulating trades without real money to validate a process. It removes financial risk but not execution assumptions — and it removes the psychological pressure that changes real decision-making.

An order placed away from current price (limit or stop) that triggers when the market reaches it. Pending orders change fill behavior versus market orders — usually less slippage, but no guarantee of a fill at all.

The standard minimum price move in forex — the fourth decimal for most pairs (0.0001), the second for JPY pairs. Spreads, stops, and targets are all conventionally measured in pips.

Deciding how large each trade should be, usually as a fixed fraction of equity scaled by stop distance. Sizing — not entries — is what determines whether a losing streak is survivable.

Gross profit divided by gross loss. A profit factor of 1.5 means the system earned $1.50 for every $1 lost — but on small trade samples it is easily inflated by luck.

A company that lets traders trade its capital after passing an evaluation with strict daily-loss and max-drawdown rules. For EA users, the challenge rules — not strategy quality — are usually what cause failure.

Net profit divided by max drawdown. It answers 'how much pain per unit of gain' — a recovery factor of 5 means the system earned five times its worst drawdown.

The probability that a strategy loses so much of the account that trading must stop. Even positive-expectancy systems carry meaningful ruin risk when position sizing is too aggressive.

The planned loss versus planned gain on a trade. Combined with win rate it determines expectancy — a 1:3 risk-reward system only needs to win about 30% of the time to break even.

Taking very short trades for a few pips each, relying on high frequency and tight spreads. Scalping is the style most sensitive to spread, slippage, and broker latency.

Restricting trading to specific hours — typically the London or New York sessions — where a strategy's assumptions about liquidity and volatility actually hold.

Average excess return per unit of total volatility. Useful for comparing smoothness of returns, but it penalizes upside volatility and can be inflated by strategies with hidden tail risk.

A trader or system that publishes entries and exits for followers to copy. Evaluate providers on verified track record length, max drawdown, and expectancy — never on a screenshot of one good month.

The difference between the price you requested and the price you actually got filled at. It grows during news, low liquidity, and with slow brokers, and it silently eats scalper profits.

Like the Sharpe ratio but only penalizes downside volatility. Often more informative for trading systems, where upside spikes are welcome and only losses count as risk.

The gap between the bid and ask price — the first cost every trade pays. Variable spreads widen around news and rollover, which can turn a profitable backtest into a losing live system.

A resting order that exits a trade at a predefined loss. Stops define risk per trade, but they are not guaranteed fills — in fast markets the exit price can be far worse than the stop level.

The overnight interest adjustment charged or paid for holding a position past the daily rollover. Long-term EAs must account for swaps; triple swap on Wednesdays can distort weekly results.

Holding positions for days to weeks to capture larger price swings. Swing systems tolerate wider spreads and slower execution, but must survive overnight gaps and swap costs.

A resting order that exits a trade at a predefined profit target. Fixed targets make backtests cleaner but cap winners; the choice between targets and trailing exits is a real strategy trade-off.

A record of every individual price quote, rather than one-minute bar summaries. Real tick data exposes spread behavior and intra-bar movement that interpolated data completely hides.

A stop loss that moves with price as a trade becomes profitable, locking in gains while leaving room to run. Poorly tuned trailing stops choke trades before the real move happens.

A rented always-on server used to run trading platforms 24/7 close to the broker's servers. For EAs, the metrics that matter are latency to the broker and uptime — not CPU marketing specs.

A testing method that repeatedly optimizes on one window of history, then validates on the next unseen window, stepping forward through time. It simulates how a strategy would have been re-optimized in real trading.

The percentage of trades that close in profit. Meaningless alone: a 90% win-rate system can lose money if the rare losses are huge, and a 35% win-rate system can thrive on large winners.