Glossary

Algorithmic trading is the use of pre-programmed rules to enter and exit positions in financial markets. The rules can be as simple as a moving-average crossover or as complex as a machine-learning model. The goal is to remove emotion from execution and trade systematically — the same setup, same size, same risk every time. Backtesting is how an algorithmic trader tests a rule against historical data before risking real capital.

See alsoWhat is backtesting?·Strategy library

The ask is the lowest price a seller is currently willing to accept in the order book. When you place a market buy order, you typically fill at the ask. The difference between the bid (best buyer's price) and ask (best seller's price) is the spread — your immediate cost of entering and exiting a position. On highly liquid markets the spread is a fraction of a tick; on illiquid markets it can be wide enough to make short-term strategies unprofitable after costs.

ADX is a non-directional momentum strength indicator developed by J. Welles Wilder in 1978. It measures how strongly a market is trending — without telling you which direction. Values range 0–100. Below 20 typically indicates a weak trend or sideways market; above 25 suggests a meaningful trend; above 40 is strong. ADX is most useful as a filter — many strategies that work well in trends fail in sideways markets, and gating entries on ADX ≥ 25 removes a large fraction of those losing trades.

See alsoADX indicator reference

ATR measures market volatility — specifically the average size of recent bars including overnight gaps. Developed by J. Welles Wilder in 1978, it computes True Range as the maximum of (high − low), |high − previous close|, and |low − previous close|, then smooths over N bars (default 14). ATR is most often used to size stop-losses, take-profits, and trailing stops dynamically — a 2× ATR stop adapts to current volatility instead of using a fixed dollar amount.

See alsoATR indicator reference·How ATR drives slippage models

A backtest is a simulation that runs a trading strategy against historical price data to estimate how it would have performed. A solid backtest applies realistic costs (commission and slippage), uses point-in-time data without look-ahead bias, and is validated across multiple market regimes. It cannot guarantee future returns, but it filters out a large fraction of bad ideas at near-zero cost — most strategies that look great on paper fall apart in backtests.

See alsoWhat is backtesting?·Run a backtest

The bid is the highest price a buyer is currently willing to pay in the order book. When you place a market sell order, you typically fill at the bid. Bid prices are always below ask prices in a healthy market — the difference is the spread. Watching the bid and ask ladder (the order book) reveals where larger participants are willing to transact, which can hint at near-term support and resistance levels.

Bollinger Bands plot a moving average flanked by two bands placed N standard deviations above and below — typically 20-period SMA with 2σ bands. The bands widen when volatility expands and contract during quiet periods. Touches of the upper band suggest extreme upside momentum (potentially overbought); touches of the lower band suggest the opposite. Mean-reversion strategies fade these touches; breakout strategies trade in the direction of band breaks.

See alsoBollinger Bands reference·Bollinger Band Bounce strategy

A breakout occurs when price moves decisively beyond a defined level — a recent high, a chart pattern boundary, or a moving average. Breakout strategies enter in the direction of the break, betting that the new level signals continuation. The Donchian Channel breakout, made famous by the Turtle Traders in the 1980s, is the canonical example. Breakouts work best in trending markets and fail badly in choppy ranges where every breakout reverses.

See alsoDonchian Breakout strategy·Donchian Channel reference

Calmar ratio is annualized return divided by maximum drawdown. It directly answers the trader's most uncomfortable question: how much pain do I trade for how much gain? A Calmar above 0.5 is workable for most retail traders; above 1.0 is excellent; above 2.0 is rare and often a sign of survivorship bias or short test periods. Unlike Sharpe ratio, Calmar focuses on the worst-case loss rather than overall volatility, making it the more honest metric for strategies you actually plan to live through.

See alsoBacktest metrics that matter

A candlestick (or candle) is a single chart bar showing four price points across a time interval: open, high, low, and close. The body shows the open-to-close range; the wick (or shadow) shows the full high-to-low range. Bullish candles are typically green or hollow (close > open); bearish are red or filled (close < open). Patterns formed by consecutive candles are the foundation of price-action and pattern-based strategies.

See alsoConsecutive Candles strategy

Commission is the explicit fee charged by your broker or exchange for executing a trade. It can be a flat amount per side (typical for futures: $3.71 per round-trip leg) or a percentage of notional value (typical for crypto perps: 0.045% per side). Commission is small per trade but compounds — a strategy that wins by 5 basis points per trade is not a strategy after commissions. Always model commission realistically in backtests.

See alsoHow costs destroy strategies

Contract size is the underlying asset quantity represented by one contract or unit. For futures, it's the number of units of the commodity (e.g. one E-mini S&P contract represents $50 × index value). For crypto perps, it's typically 1 unit of the base asset (e.g. 1 BTC). Contract size determines how much price movement translates into P&L. Knowing it precisely is essential for position-sizing — a 1-tick move on a 10,000-bushel corn contract is meaningfully different from a 1-tick move on a 100-share stock.

A crossover is the moment one indicator line crosses through another. The most common example: a fast moving average crossing above (golden cross) or below (death cross) a slow moving average. Crossovers are the trigger for many trend-following strategies because they mark visible shifts in short-term vs longer-term momentum. The trade-off: crossovers always lag the actual price turn, so they trade smoother trends well and choppy markets poorly.

See alsoEMA Crossover strategy·MACD Crossover strategy

Divergence occurs when an oscillator (like RSI or MACD) moves in the opposite direction to price — for example, price makes a new high while the oscillator makes a lower high. The interpretation: momentum is fading even as price continues, suggesting an upcoming reversal. Bullish divergence is the symmetric setup at lows. Divergence-based strategies require careful swing detection to avoid look-ahead bias and tend to work best at late-cycle exhaustion phases rather than mid-trend.

See alsoMACD Divergence strategy·RSI Divergence strategy

The Donchian Channel plots the highest high and lowest low over the last N bars (default 20), with a midline in between. It defines an objective N-period range — useful for breakout strategies that trade beyond the channel boundaries. Made famous by the Turtle Traders in the 1980s, Donchian breakouts work best in trending commodity-style markets where new N-bar extremes signal genuine momentum continuation rather than noise.

See alsoDonchian Channel reference·Donchian Breakout strategy

Drawdown is the percentage decline from an account's all-time-high equity to its current value. The maximum drawdown over a backtest is the worst peak-to-trough decline — the deepest pain you would have endured had you traded the strategy from the start. Even if the strategy recovered, a 40% drawdown means at some point your account was 40% below its high. Most live traders abandon systems at less than half that drawdown — backtests assume infinite patience; humans don't have it.

See alsoReading equity curves·Why max drawdown matters

The equity curve is the headline output of every backtest. It plots cumulative profit or loss over time, starting from your account size and ending wherever the simulation finished. One glance gives you trajectory, drawdown, and smoothness — qualities that matter more than the final P&L. A healthy equity curve makes many small new highs separated by modest pullbacks; an overfitted one makes most of its gains in one concentrated period.

See alsoHow to read an equity curve

Expectancy is the average expected profit (or loss) per trade, calculated as (win rate × average win) − (loss rate × average loss). A positive expectancy means the strategy makes money on average; negative means it bleeds. Expectancy ties together win rate and the asymmetry of wins versus losses — a strategy with 40% win rate can still be profitable if winners are twice as large as losers. Use expectancy to evaluate strategies with very different win rates on equal footing.

The exponential moving average (EMA) is a weighted average of recent prices that gives more weight to recent bars. Compared to a simple moving average (SMA), the EMA reacts faster to price changes — useful for catching trend turns earlier, at the cost of being slightly noisier. EMAs are the building blocks of many strategies (EMA Crossover, MACD) and are also used as trend filters for entries on other strategies.

See alsoEMA reference·EMA Crossover strategy

A forward test runs a strategy on out-of-sample data the strategy was never tuned on. It's the closest you can get to live performance without risking real money. A backtest tells you whether a strategy survived hindsight; a forward test tells you whether it generalizes. If your strategy looks great in backtest but fails forward, you've likely overfit. Walk-forward analysis is a structured form of forward testing across rolling windows.

See alsoWalk-forward analysis

Funding rate is a periodic payment exchanged between long and short perpetual futures holders to keep the perp price anchored to spot. When the perp trades above spot (longs dominate), longs pay shorts; when below (shorts dominate), shorts pay longs. Funding is typically every 8 hours on most exchanges and can be 0.01–0.1% per period. Sustained funding cost is one of the most overlooked drag factors on crypto perp strategies — a 0.05% per 8h funding rate equals roughly 55% per year on holding cost.

Hold period is the number of bars (or amount of time) a strategy stays in a position after entry. Fixed hold periods exit on a fixed bar count regardless of subsequent price action — simple and disciplined, but blind to opportunity. Adaptive hold periods exit on conditions like indicator reversals, opposite candles, or moving-average breaks. Different strategies suit different hold philosophies: trend-following typically wants to ride moves, mean-reversion wants to exit before the fade reverses.

Ichimoku Kinko Hyo ("one-glance equilibrium chart") is a five-component trend system developed by Goichi Hosoda in pre-war Japan. The components — Tenkan, Kijun, Senkou A, Senkou B, and Chikou — provide trend direction, support and resistance, and entry timing in one view. The "cloud" formed between Senkou A and Senkou B is the most distinctive visual: price above the cloud is bullish, below is bearish. The system rewards trends, frustrates ranges, and was calibrated for daily Japanese equities.

See alsoIchimoku reference·Ichimoku Cloud strategy

An IOC order executes whatever portion can be filled immediately at the limit price and cancels the rest. Useful when you want to take liquidity at a specific price without leaving a resting order in the book. Crypto perpetual exchanges like Hyperliquid and dYdX route market orders as IOC under the hood with a wide slippage tolerance. The risk: in thin order books, your IOC may only partially fill, leaving the strategy unexpectedly under-sized.

In-sample data is the slice of history a strategy was trained, optimized, or tuned on. Performance on in-sample data is always inflated because parameters were chosen specifically to fit that period. The corollary: any strategy looks great on the data it was optimized on. The whole point of out-of-sample testing and walk-forward analysis is to expose how much of in-sample performance was real edge versus overfit.

See alsoWalk-forward analysis·Common backtest mistakes

Leverage is the ratio of position size to margin posted. 5× leverage means a $1,000 margin controls a $5,000 position. Leverage amplifies both gains and losses proportionally — and a 20% adverse move at 5× leverage is enough to liquidate the position. Crypto perp exchanges typically allow up to 50–100× leverage; most professional traders use 2–5×. High leverage rarely means high return — it usually means high probability of liquidation before the strategy's edge plays out.

Liquidation occurs when an exchange forcibly closes a leveraged position because the trader's equity has fallen below maintenance margin. The exchange takes over the position, dumps it into the order book, and the trader loses most or all of their margin. Liquidation cascades — large liquidations push price further, triggering more liquidations — are a common cause of violent crypto wicks. The further a trade is from its liquidation price, the more breathing room a strategy has to survive normal volatility.

Liquidity is the ease of buying or selling at the current market price without significantly moving it. Highly liquid markets (e.g. EURUSD, BTC perps on Binance) have tight spreads and deep order books — your trade fills near the displayed price. Illiquid markets show wide spreads, thin books, and large fills that walk through multiple price levels. Strategy edge can disappear instantly if you trade in volume the market cannot support.

A long position profits when price rises. "Going long" means buying with the expectation of selling higher later. The maximum loss on a long is the entry price (price can fall to zero); the maximum profit is theoretically unlimited. Long is the natural side for buy-and-hold investors and most retail traders. In algorithmic trading, the same strategy logic typically generates long signals in uptrends and short signals (the opposite) in downtrends.

Look-ahead bias is using information at time T that you would not actually have known until T+1 (or later). The classic case: computing today's signal from today's closing price and assuming you could trade on the close. In reality, the bar had to finish before you knew the signal — by then the close was gone. Look-ahead bias makes a backtest look better than reality. Fix it by lagging signals by one bar and using only confirmed swings or completed indicators.

See alsoCommon backtest mistakes

A market order executes immediately at the best available price — bid for sells, ask for buys. It guarantees execution but not price. In thin or fast-moving markets, the actual fill price can differ significantly from the price you saw when placing the order — this is slippage. Market orders are ideal when getting in or out matters more than the exact price; limit orders are better when price discipline matters more than execution certainty.

Maximum drawdown (max DD) is the largest peak-to-trough percentage decline in equity over a backtest. A strategy with 30% max DD means at some point your account was 30% below its all-time high. Max DD is arguably the single most important number in a backtest — bigger than total return, bigger than Sharpe — because it represents the worst-case psychological pain you would have to endure to keep the strategy on. Most retail traders abandon systems before hitting their backtested max DD.

See alsoBacktest metrics that matter

Mean reversion is the trading thesis that prices oscillate around an average and tend to revert toward it after extreme deviations. Mean-reversion strategies fade extreme moves: buy oversold dips, sell overbought rallies. They work in range-bound markets and fail catastrophically in strong trends — "price rides the upper band" is the classic failure mode. Bollinger Band Bounce, Stochastic Oscillator, and VWAP Mean Reversion are all built on this thesis.

See alsoBollinger Band Bounce·Stochastic Oscillator strategy·VWAP Mean Reversion

Momentum is the tendency of recently strong assets (or moves) to keep going. Momentum strategies enter in the direction of recent strength — buying assets that have been rising, shorting those that have been falling. Empirically robust across centuries of equity data and most asset classes. Momentum and trend-following are closely related — trend-following typically uses smoothed indicators (moving averages); pure momentum looks at raw return rankings over fixed windows.

Monte Carlo simulation in backtesting shuffles the order of historical trades (or resamples them with replacement) to generate thousands of alternate equity paths. The distribution shows how dependent the original equity curve was on the lucky sequence of trades. Pessimistic outcomes (P5, P10) reveal worst plausible drawdowns; optimistic outcomes (P95) reveal what could have gone right. A strategy whose median Monte Carlo path is much worse than its actual backtest is fragile to trade-order luck.

See alsoReading equity curves

MACD, developed by Gerald Appel in the late 1970s, captures both trend and momentum from the difference between two EMAs (typically 12 and 26 periods) plus a signal line that is itself an EMA of the MACD line (typically 9 periods). The difference between MACD and signal lines is plotted as the histogram. Used either as a crossover trigger (MACD crosses signal) or as a divergence detector (MACD swings disagree with price swings).

See alsoMACD reference·MACD Crossover strategy

Net profit and loss is gross trading profit minus all costs — commissions, slippage, funding (for perps), and fees. The single dollar number at the bottom of every backtest. Net P&L on its own is a marketing number; it tells you nothing about volatility, drawdown, or trade count. Always interpret it together with Sharpe ratio, max drawdown, and per-trade P&L. A $100,000 backtest profit from one giant winning trade and 100 small losers is not a tradeable edge.

OBV is a cumulative volume-flow indicator developed by Joseph Granville in 1963. It adds the day's volume to a running total when price closes up and subtracts when price closes down. Rising OBV alongside rising price confirms the trend; OBV diverging from price suggests volume isn't supporting the move and a reversal may be near. Used most often as a confirmation indicator rather than a primary signal generator.

See alsoOBV reference

Out-of-sample (OOS) data is the slice of history reserved during strategy development and never used for parameter tuning. Performance on OOS data is the closest backtest proxy for live trading. A strategy with a 2.0 in-sample Sharpe and a 0.3 out-of-sample Sharpe has degraded 85% — almost all of the apparent edge was overfitting. Walk-forward analysis automates the OOS evaluation process across multiple rolling windows.

See alsoWalk-forward analysis

Overfitting is tuning a strategy's parameters until the backtest looks perfect on a specific historical period — but the parameters describe noise, not signal. The classic symptom: in-sample performance is excellent and out-of-sample collapses. The cure is discipline: minimize the number of free parameters, validate with walk-forward analysis across multiple windows, and resist the urge to optimize until everything looks good. A strategy that wins on perfect parameters in-sample almost always loses out-of-sample.

See alsoCommon backtest mistakes·Walk-forward analysis

Paper trading is executing a strategy in real time without real capital — usually through a simulated brokerage account. It tests live signal generation, entry/exit timing, and emotional discipline without financial risk. Paper trading exposes friction that backtests can't simulate: bad fills, exchange downtime, your own hesitation. A strategy that survives 1–3 months of paper trading earns the right to be sized small with real capital next.

Perpetual futures (perps) are derivatives contracts that track an underlying asset's price but never expire. To keep the perp tethered to spot, exchanges use a funding rate paid periodically between longs and shorts. Perps dominate crypto trading volume because they offer leverage, work 24/7, and have no settlement date to manage. Hyperliquid, Binance, dYdX, Bybit, OKX are the major perp venues. Standard perp leverage on retail platforms is up to 50–100× — though prudent strategies rarely use more than 5×.

Position size is the dollar (or contract) amount allocated to a single trade. Sizing is arguably the most important variable in trading — far more impactful than entry signal quality. Common approaches: fixed dollar per trade, fixed percentage of equity, ATR-scaled (tight stops in calm markets get bigger size), or Kelly-based. A bad strategy with disciplined sizing can survive; a good strategy with reckless sizing can blow up on a single bad trade.

Profit factor is gross profit divided by gross loss. A profit factor of 1.5 means the strategy makes $1.50 for every $1.00 it loses. Below 1.2 means a small change in market structure could flip the strategy negative; above 1.5 is healthy; above 2.0 is unusual and worth a re-check for overfitting or look-ahead bias. Profit factor is more robust than win rate alone because it accounts for the size asymmetry between winners and losers.

See alsoBacktest metrics that matter

RSI is a bounded momentum oscillator (0–100) developed by J. Welles Wilder in 1978. It compares the magnitude of recent gains to recent losses over a lookback window (default 14 bars). Conventionally, RSI > 70 is "overbought" and < 30 is "oversold" — though these levels are heuristics, not laws. RSI is most useful as a filter, a divergence detector, or a centerline crossover signal (rising through 50 = strengthening uptrend).

See alsoRSI reference·RSI Divergence strategy

Resistance is a price level or zone at which selling pressure has historically overcome buying pressure, capping further upside. It can be a prior high, a round number, or a moving average — anywhere market participants remember being burned and place sell orders. Strong resistance is repeatedly tested and respected; broken resistance often becomes future support. Strategies use resistance levels for stop placement, breakout targets, or as no-trade zones.

Risk-reward ratio is the size of your stop-loss compared to your take-profit, expressed as 1:N. A 1:2 risk-reward means you risk $1 to make $2. Strategies with risk-reward >= 1:1 can be profitable even with sub-50% win rates — at 1:2, a 40% win rate still produces positive expectancy. Lower risk-reward (e.g. 1:0.5) requires very high win rates to overcome the asymmetry, which is rare in real trading.

Sharpe ratio is annualized excess return divided by annualized volatility. It measures how much return you earned per unit of risk taken. Above 1.0 is decent for retail strategies; above 2.0 is genuinely good; above 3.0 is rare and usually indicates a short test period, overfit parameters, or unmodeled costs. Sharpe is the most-cited single number in performance measurement, but it treats upside and downside volatility equally — Sortino ratio refines this by only penalizing downside.

See alsoBacktest metrics that matter

A short position profits when price falls. To go short, you borrow the asset (or use a derivative like a futures contract) and sell it now, planning to buy it back cheaper later. The maximum profit is the entry price (price can fall to zero); the maximum loss is theoretically unlimited (price can rise indefinitely). Short positions can also pay or receive carry costs — borrow rates for stocks, funding rates for crypto perps.

The simple moving average is the unweighted average of the last N price values. The 50-day and 200-day SMAs of stock prices are the two most-watched moving averages in the world — when the 50-day crosses above the 200-day (a "golden cross"), it's a widely-cited bullish signal. SMAs lag price more than EMAs do (they weight all bars equally), making them smoother but slower to react.

See alsoSMA reference

Slippage is the difference between the price your strategy expected to fill and the actual fill price. On a market buy, slippage is positive (you paid more than the displayed price); on a market sell, you received less. Slippage is driven by spread, order book depth, and order size. Realistic slippage modelling is non-negotiable in backtests — strategies with thin per-trade edges look profitable on zero-slippage backtests and bleed in production. Common models: fixed pips, percentage of price, or ATR-multiplied.

See alsoHow slippage destroys strategies

Sortino ratio is a refinement of Sharpe ratio that penalizes only downside volatility, not total volatility. The intuition: traders don't mind upside swings, only downside ones. Sortino is calculated as annualized excess return divided by annualized downside deviation. A strategy with steady gains punctuated by occasional sharp wins (e.g. trend-following) often shows a low Sharpe but a high Sortino — Sortino is the more honest metric for asymmetric strategies.

Spread is the difference between the best ask and the best bid in the order book. It is your immediate cost of round-trip entry and exit. On highly liquid markets the spread is one tick or less; on illiquid markets it can be wide enough to make short-term strategies unprofitable. Spread typically widens during high volatility and during low-liquidity hours (overnight, weekends), so a strategy's true cost depends on when it trades.

The Stochastic Oscillator measures where the current close sits within the high–low range of the last N bars (default 14). The %K line tracks raw value, the %D line is a smoothed version. Stochastic is bounded 0–100, with 80+ traditionally overbought and 20- oversold. Most-used as a mean-reversion timing signal: %K crossing %D inside an extreme zone (above 80 for shorts, below 20 for longs) is a classic reversal trigger.

See alsoStochastic reference·Stochastic strategy

A stop-loss is an order that closes a position automatically if price moves a defined distance against the entry. It defines the maximum loss per trade. Stops can be fixed (a specific price), percentage-based (X% from entry), or volatility-aware (multiples of ATR). The right stop distance balances giving the trade room to breathe against capping downside — too tight and normal noise stops you out; too wide and the worst case is unbearable.

Supertrend is a single-line trend-following indicator that combines ATR-derived bands with directional flip logic. It plots one line that flips between an upper band (during downtrends) and a lower band (during uptrends) based on whether price closes through the opposite band. The line itself acts as a built-in trailing stop. Compared to EMA-based trend strategies, Supertrend gives unambiguous direction signals — either you're above the line or below it.

See alsoSupertrend reference·Supertrend strategy

Support is a price level or zone at which buying pressure has historically overcome selling pressure, halting further downside. Like resistance, support can be a prior low, a round number, or a moving average. Strong support is repeatedly tested and respected; broken support often becomes future resistance. Strategies use support levels for stop placement, mean-reversion targets, or as confirmation of trend continuation.

Survivorship bias is testing a strategy on a universe of assets that excludes those that went bankrupt, got delisted, or failed. The strategy looks great because the dataset already filtered out the losers. The classic example: backtesting on the current S&P 500 components ignores the dozens of companies removed for poor performance over the test period. Use point-in-time universes to avoid this — what was actually tradeable on each historical date.

See alsoCommon backtest mistakes

A take-profit is an order that closes a position automatically once price reaches a defined gain. It locks in upside without requiring you to watch the screen. Take-profits can be fixed-price, percentage-based, or volatility-aware (multiples of ATR). The trade-off: tight take-profits cap upside on big winners; wide ones can give back gains during reversals. Many strategies skip take-profit altogether and rely on time-based or trailing-stop exits to ride trends.

Tick size is the minimum price increment at which a market trades. For US equities it's $0.01 for most stocks; for E-mini S&P futures it's 0.25 index points worth $12.50; for BTC perps it's typically $0.10 or $0.50 depending on price. Tick size sets the granularity of stops, take-profits, and limit orders. It also defines the smallest possible spread — a tighter tick size means tighter spreads and finer-grained price control.

A trailing stop is a stop-loss that moves with the trade in the favourable direction, locking in profit as the position runs. For a long, the trailing stop only ratchets up — never down. Common implementations: fixed-distance trail (e.g. 2% behind highest seen price), ATR-multiplied trail (volatility-aware), or chandelier exits (anchored to recent swing). Trailing stops are essential to trend-following strategies that want to ride extended moves but still cap reversals.

Trend following is the trading thesis that markets in motion tend to stay in motion. Trend strategies enter in the direction of recent strength — buying assets that have been rising, shorting those that have been falling — and let winners run with trailing stops or extended hold periods. Trend following has been profitable across centuries of futures and equity data, but suffers long flat stretches in choppy markets. EMA Crossover, Supertrend, and Donchian Breakout are all trend-following strategies.

See alsoEMA Crossover·Supertrend·Donchian Breakout

Volume is the number of units (shares, contracts, coins) traded during a given time interval. High volume signals strong participation and conviction; low volume suggests indifference or thin liquidity. Volume is most useful as confirmation — a breakout on rising volume is more credible than one on declining volume; a divergence between price and volume often precedes reversals. Many strategies use volume relative to its moving average (e.g. "volume > 1× SMA") as an entry filter.

See alsoVolume SMA reference·OBV reference

VWAP is the average price weighted by volume traded at each price level over a session. Unlike a moving average, VWAP gives more influence to bars that traded heavily. Institutional traders use VWAP as an execution benchmark — they want to fill orders at or better than VWAP. Retail strategies use VWAP as an intraday mean-reversion target (price tends to revert toward VWAP) or as a trend-direction filter (longs above VWAP, shorts below).

See alsoVWAP reference·VWAP Mean Reversion strategy

Walk-forward analysis splits historical data into rolling pairs of in-sample (parameter tuning) and out-of-sample (evaluation) windows. The strategy is optimized on each in-sample window, then tested on the immediately following out-of-sample window — and the parameters never see the OOS data before being scored. Walk-forward is the gold-standard test for whether a strategy's edge is real or overfit. Pass rates above 60% across multiple OOS windows are interesting; below 40% is junk.

See alsoWalk-forward analysis explained

Win rate is the percentage of trades that closed profitably. It's the most-cited statistic in trading marketing and the most over-interpreted. A high win rate (e.g. 80%) means little if the average loser is much larger than the average winner. A low win rate (e.g. 35%) is fine if the average winner is 3× the average loser. Always interpret win rate alongside risk-reward ratio and expectancy — never on its own.

See alsoBacktest metrics that matter