A working dictionary for everything in this guide. Each entry is 80–200 words and links to the canonical page where the concept is covered in depth. Use ⌘F / Ctrl+F to jump to a term.
A
Algorithmic trading
The use of pre-defined, mechanical rules to make trading decisions. The opposite of discretionary trading. A strategy that executes by code, by node graph, or by any rule set that doesn't rely on real-time human judgment is algorithmic. AlgoLift is a platform for designing and running algorithmic strategies without writing code. See introduction to trading.
ATR (Average True Range)
A volatility indicator that measures the average size of price ranges over a lookback period. Used to size stops and targets relative to current market volatility — a 1.5×ATR stop is wide in volatile sessions and tight in calm ones, automatically. See ATR and position sizing.
Ask
The lowest price any seller in the order book is willing to accept right now. When you submit a market order to buy, you pay the ask. The difference between ask and bid is the spread. See slippage and execution.
B
Backtest
A simulation of a trading strategy against historical data, used to estimate how the strategy would have performed in the past. A backtest cannot predict the future; it can only stress-test the past. See backtesting explained and the backtest engine.
Bar (or candle)
A summarized view of trading activity over a fixed time period (e.g. 1 minute, 5 minutes, 1 hour). Each bar has four price points: open, high, low, close (OHLC). Bar-based backtesters operate on these summaries; tick-based backtesters operate on the underlying prints. See tick vs bar data.
Bid
The highest price any buyer in the order book is willing to pay right now. When you submit a market order to sell, you receive the bid. See slippage and execution.
Breakout
A class of strategy that enters when price moves beyond a defined range or level — a 20-day high, the prior day's high, a session range. See breakout strategies.
Bootstrap (resampling)
A technique for estimating uncertainty by repeatedly resampling the data you have. In the Robustness tab, bootstrap takes your strategy's trade list and resamples it ten thousand times to produce alternate equity curves. Distributions across those resamples drive everything visible on the tab — percentile bands, P(profit), P(ruin), tail drawdowns. The platform defaults to stationary block bootstrap rather than iid because trade returns cluster in real data, and iid would smooth those clusters away. See bootstrap (resampling).
C
CAGR (Compound Annual Growth Rate)
The annualized return of a strategy, compounded. A 15% CAGR means the strategy grew the account at an effective 15%-per-year rate. Useful only when paired with the maximum drawdown that produced it. See trading metrics.
Calmar ratio
CAGR divided by maximum drawdown. One of the best one-glance quality metrics — it tells you how much return you got per unit of worst-case pain. Calmar above 1.0 is decent for a retail strategy; above 2.0 is rare. See trading metrics.
Cluster (node cluster)
A reusable group of nodes saved as a single block in the AlgoLift palette. Build a complex piece of logic once (a session filter, an ATR-based stop, a prop-firm kill switch), save it as a cluster, and drop it into every future strategy as one node. See Cluster and the Visual Builder.
Cooldown
A condition that prevents a strategy from re-firing an entry until a reset state is reached. The classic use: after entering on an EMA crossover, refuse to re-enter until the EMAs have crossed back. Without a cooldown, crossover strategies machine-gun losses in trending markets. See Cooldown and the your first strategy walkthrough.
Correlation
A statistical measure (between -1 and +1) of how two return streams move together. +1.0 means they move identically; 0 means independent; -1 means opposite. Two strategies with correlation above +0.7 are essentially the same trade. See strategy correlation and the portfolio builder.
Curve fitting (overfitting)
A failure mode where a strategy is tuned so precisely to historical data that it memorizes the past instead of finding a real edge. Curve-fit strategies look excellent in backtest and collapse in live trading. See overfitting in trading.
D
Data series
A specific stream of price data — a symbol, a timeframe, and a bar type (time, tick, volume, range). A single strategy graph can use multiple data series at once: a 60-minute trend filter on top of a 5-minute execution chart, or NQ data alongside ES data. See Visual Builder.
Discretionary trading
Trading decisions made manually, in the moment, based on judgment, intuition, and pattern recognition. The opposite of systematic / algorithmic trading. Discretionary traders can be excellent; their decisions just don't generalize into rules an engine can execute. See introduction to trading.
Divergence (forward-test)
The gap between a strategy's live forward-test performance and what the backtester would have produced over the same calendar window. A small divergence is normal. A large one means the backtest had hidden look-ahead bias or fill-model error. See forward testing.
Divergence (technical)
A pattern where price makes a new high (or low) but a momentum indicator like RSI does not. Bullish divergence: price makes a lower low, RSI makes a higher low. Often used as a contrarian entry signal. See RSI.
Drawdown
The decline in equity from a recent peak. A 12% drawdown means the account is currently 12% below its highest point. The largest drawdown over a period is the maximum drawdown. See risk management basics.
Deflated Sharpe Ratio (DSR)
A correction applied to a Sharpe ratio when it was picked as the best of N optimization trials. DSR adjusts for trial-count selection bias, return skewness, kurtosis, and sample size to produce a probability that the strategy's true Sharpe exceeds zero. A DSR of 50% is coin-flip; above 75% is meaningful; above 95% is strong. Bailey & López de Prado (2014). See Deflated Sharpe Ratio.
E
Edge
A statistical advantage that produces positive expected value over a long sample of trades. An edge can come from speed (HFT), information (insider, fundamental), structure (mean reversion in ranges, momentum in trends), or process (better risk management than peers). Without an edge, no amount of risk management produces profit. See introduction to trading.
Edge decay
The gradual deterioration of a strategy's profitability as the market conditions that supported it change. Edges in mainstream markets typically have lifespans measured in years, not decades. Detecting edge decay is what the rolling-Sharpe panel in Analytics is for.
EMA (Exponential Moving Average)
A moving average that gives more weight to recent prices. Faster to react than the SMA but with the same smoothing structure. The default trend indicator for most short-to-medium-term strategies. See EMA.
Expectancy
The average dollar profit (or loss) per trade, including costs. Calculated as (Win Rate × Avg Win) − (Loss Rate × Avg Loss). The most important per-trade metric — it tells you what to expect on the next trade in dollar terms. See trading metrics.
Equity curve
A line chart showing the cumulative balance of an account over time. Steep upward = profitable; downward = losing; sideways = no edge. The shape of the curve (smooth stair-stepping vs. one-big-trade carry) is often more informative than the headline numbers. See Analytics.
Edge Concentration
A Robustness Score component measuring how evenly profit is distributed across trades. Calculated as 1 minus the Gini coefficient of per-trade PnL. High score means many trades contribute to the equity curve; low score means a small lucky cluster carries the backtest. Concentrated edge is the most common pattern of in-sample overfitting on the platform. See Edge Concentration.
F
Fair Value Gap (FVG)
A three-bar price-action pattern where the wicks of bar 1 and bar 3 don't overlap, leaving a "gap" of price that wasn't traded. Often used as a level price returns to. See Fair Value Gap.
Fill
The completion of an order at a specific price. A market order fills immediately at whatever price the book offers. A limit order fills only when the market trades at (or through) the limit price and the volume ahead of it in the queue has cleared. See order types & fills.
Fixed fractional risk
A position-sizing rule where each trade risks a fixed percentage of current account equity, not a fixed dollar amount. Sometimes called "the 1% rule" because 1% is a common starting value. Self-adjusts position size up during winning streaks and down during losing streaks. See risk management basics.
Forward testing
Running a strategy on live data without real capital at risk. The bridge between backtesting and live deployment — catches look-ahead bias and fill-model errors before they cost real money. See forward testing.
G
Gap
A price discontinuity between two consecutive bars where no trades occurred in the intervening range. Common at session opens (overnight gaps) and around major news. Gaps complicate stop placement because the trigger price may never have actually traded.
I
Indicator
A computed value derived from price (and sometimes volume) that summarizes some aspect of market state — momentum, volatility, trend, volume flow. In AlgoLift, every indicator is a data node in the palette. See the indicator reference.
In-sample
The portion of historical data used to design and optimize a strategy. Strategies always look better in-sample than out-of-sample because they've been fit to it. The discipline of reserving an out-of-sample window is the single most important habit in honest backtesting. See overfitting in trading.
K
Kill switch
An automatic mechanism that flattens all positions and halts trading when a configured threshold (typically a daily loss limit or max drawdown) is breached. Built into AlgoLift's exported executable. See risk management basics.
Kelly Fraction (incl. Half-Kelly)
The position-size multiplier that maximizes long-run geometric return. Full Kelly is mathematically optimal but emotionally brutal — drawdowns at Kelly are roughly the size of average gains. Half-Kelly captures about 75% of Kelly's growth at half the variance and is the size most professional traders actually use. Above Kelly, geometric return decreases even though arithmetic return keeps rising, because variance penalty exceeds upside. See Kelly Fraction.
L
Latency
The delay between a market event and your system's reaction to it. For retail trading, latency is dominated by the broker round-trip (typically 50–250ms). For HFT, latency is measured in microseconds. See latency execution modeling.
Limit order
An order to buy or sell at a specific price or better. A buy limit at 4290 fills at 4290 or lower. Guarantees price, doesn't guarantee execution — the market can trade through your limit without filling you. See order types & fills.
Liquidity
The ease of buying or selling without moving the price. A liquid market has tight bid/ask spreads and deep order books at each price level. An illiquid market has wide spreads, thin books, and high slippage on any size. ES, MES, NQ during regular trading hours are highly liquid; back-month contracts at 3 AM are not.
Look-ahead bias
A backtest error where the strategy uses information that wouldn't have been available at the moment the trade was decided. The most common form: reading the closed-bar value of the current bar before the bar has closed. Look-ahead bias makes backtests look brilliant and live trading miserable. See bias & data integrity.
M
MAE (Maximum Adverse Excursion)
For each trade, the worst (most underwater) point the trade reached before closing. Used to evaluate whether stops are too tight or too loose. If most winning trades had MAE of -0.8R before reaching +2R, a -0.5R stop would have killed half your winners. See Analytics.
MFE (Maximum Favorable Excursion)
For each trade, the best (most profitable) point the trade reached before closing. Used to evaluate whether targets are too tight. If most wins had MFE of +3R but you took +1R, there's profit on the table.
Market order
An order to buy or sell at the best available price right now. Guarantees execution, doesn't guarantee price. Pays the spread by definition and may also pay slippage if size exceeds top-of-book.
Maximum drawdown
The largest peak-to-trough decline in account equity over a period, expressed as a percentage of the peak. The headline measure of worst-case pain. A 15% max DD means the account was at one point 15% below its high water mark. See risk management basics.
Mean reversion
A class of strategy that bets price will return toward an average after stretching away from it. Classic mean-reversion setups: RSI < 30 in a range-bound market, price 2+ standard deviations below a moving average. See mean reversion strategies.
Meta-labeling
An advanced machine-learning technique where an ML model filters the trades a base strategy would take, predicting which signals are likely to fail. Reduces false positives without changing the base strategy. See meta-labeling.
Momentum
A class of strategy that bets price will continue in its recent direction. Classic momentum setups: 20-day breakouts, MACD bullish crossovers, RSI > 70 in trending markets. See momentum strategies.
Monte Carlo simulation
A robustness test that randomly shuffles the order of a strategy's trades (or randomly skips a percentage of them) and recomputes the equity curve. Done 1,000+ times, the distribution of resulting equity curves tells you whether the original result came from edge or lucky sequencing. See Monte Carlo simulation.
N
Node
A unit on the AlgoLift canvas — an indicator, a logic gate, a math operation, an order action, etc. Strategies are assembled by connecting node outputs to node inputs. See the node reference.
Noise
Random price fluctuations with no predictive content. The opposite of signal. Strategies that fit to noise (instead of signal) are the textbook definition of overfitting.
O
OHLC
Open, High, Low, Close — the four price points that summarize a bar. The minimum data needed for bar-based analysis. Tick-based engines like AlgoLift's use the underlying prints instead of (or in addition to) OHLC. See tick vs bar data.
Out-of-sample
The portion of historical data deliberately not used during strategy design and optimization, reserved for validating the final strategy. Out-of-sample performance is the closest honest estimate of future live performance you have. See overfitting in trading and walk-forward optimization.
P
Parameter
A configurable input to a strategy — an indicator period, a threshold, a stop multiplier. Every parameter is a knob the optimizer can use to chase noise; the fewer parameters a strategy has, the harder it is to overfit. See parameter optimization.
Position sizing
The decision of how many contracts or shares to trade. Determined by your per-trade risk %, the stop distance, and account equity. The lever that most determines whether a strategy with an edge actually compounds or blows up. See position sizing.
Profit factor
Gross profit divided by gross loss. A profit factor of 1.0 is break-even; below 1.0 the strategy loses money. The rough floor for "real edge after costs" is around 1.4–1.5. Above 2.5 in a retail backtest is usually a flag for overfitting. See trading metrics.
Prop firm
A firm that funds traders to trade the firm's capital in exchange for a profit split. Typically requires passing an evaluation that has specific daily-loss and max-drawdown rules. AlgoLift includes prop-firm-specific simulation for FTMO, TopStep, Apex, and others.
Path Stability
A Robustness Score component (25% weight). Two signals multiplied: where your original backtest's total return lands in its resampled cloud (near the middle = a typical draw and scores well; the lucky top tail = penalized), and a scale-invariant width penalty that dampens the score only when the fan's downside is large relative to the typical gain. The single best detector of trade-order luck on the platform. See Path Stability.
Politis-Romano (stationary block)
The 1994 paper that introduced stationary block bootstrap — the resampling method the Robustness tab uses by default. Random-length blocks drawn from a geometric distribution preserve autocorrelation (streak structure) while maintaining the stationarity property needed for valid asymptotic convergence. The Politis-White (2004) auto-block-length rule is what picks the right block size per dataset. See Stationary Block Bootstrap.
Probability of Backtest Overfitting (PBO)
A direct probability that the in-sample champion of an optimization will underperform the median out-of-sample. Computed via combinatorial symmetric cross-validation across 16 time-folds and all 12,870 train/test split combinations. PBO above 50% means the optimizer is actively selecting configurations that look good in-sample because they look bad out-of-sample. Bailey, Borwein, López de Prado, Zhu (2017). See Probability of Backtest Overfitting.
Probability of Profit
The fraction of bootstrap-resampled alternate histories that end with positive total return. A path-level reliability metric — not the same as per-trade win rate. Above 70% is the practical deployability threshold for most strategies. Read alongside P(ruin) — the two together bracket the strategy's return-vs-survival profile. See Probability of Profit.
Probability of Ruin
The fraction of bootstrap-resampled paths where running account balance hits zero. Different from drawdown — drawdown is recoverable, ruin is terminal. Target below 1% for retail strategies; anything above 5% means the strategy is one bad cluster away from being unrecoverable at the current size. The right primary risk metric for leveraged strategies because it captures the absorbing-state asymmetry that drawdown averages erase. See Probability of Ruin.
Prop Fitness
A Robustness Score component (20% weight when prop firms are enabled) computed as the geometric mean of probability of passing the eval and probability of reaching funded status given a pass. The multiplicative structure catches strategies that breeze through the eval and bleed out under the funded-phase trailing drawdown rule. Drops to zero weight when no firm is selected; weight redistributes to Tail Discipline and Regime Breadth. See Prop Fitness.
Q
Queue position
For limit orders, your place in line at a price level. Other orders placed before yours fill first; you only fill once they've cleared. The platform's Simulate Fills mode models this accurately.
R
Regime
A persistent market state characterized by specific volatility, trend, and correlation behavior. Strategies often work in some regimes and fail in others. Detecting regime shifts is the subject of market regimes and regime detection.
R-multiple
A normalized way of expressing profit and loss as a multiple of the trade's initial risk. A trade risking $100 that makes $250 is a +2.5R win. Lets you compare trades of different sizes on equal footing.
RSI (Relative Strength Index)
A momentum oscillator measuring the speed and magnitude of recent price changes on a 0–100 scale. Common interpretation: above 70 is "overbought," below 30 is "oversold." Stays above 70 for weeks in strong trends, which is the most common RSI misinterpretation. See RSI.
Regime Breadth
A Robustness Score component (10% weight) measuring the minimum probability of positive Sharpe across detected market regimes. Scores the worst regime, not the average — anti-cherry-pick by design. A strategy that's brilliant in trends and unprofitable in chop scores low because the unprofitable regime defines the floor. When the parent backtest doesn't have enough regime variety to classify, the platform falls back to a 60/40 volatility split. See Regime Breadth.
Robustness Score
The 0–100 headline metric of the Robustness tab. A weighted composite of six components (Path Stability, Tail Discipline, Prop Fitness, Edge Concentration, Regime Breadth, Sample Size Confidence) capturing how well a strategy survives stress-testing. Below 40 is fragile; 40–70 is watch; above 70 is ship under tested assumptions. Designed so no single component can carry the score — saturating one dimension caps possible score around 65. See Robustness Score.
S
Sharpe ratio
Average return divided by standard deviation of returns, annualized. The most-quoted risk-adjusted return metric. Sharpe above 1.0 is decent; above 2.0 is rare in retail markets. Penalizes upside volatility equally with downside — see also Sortino ratio. See trading metrics.
Signal
In trading: a real, repeatable pattern in market behavior that can be exploited for edge. The opposite of noise. The point of strategy design is to capture signal without overfitting to noise. See overfitting.
Simulate Fills
A toggle in the AlgoLift backtest engine that enforces strict limit-order queue rules and walks the book on market orders. Off for fast iteration; on for any decision-grade test.
Sample Size Confidence
A Robustness Score component (5% weight) penalizing backtests with too few trades. Log-scaled: 0 at n=1, ~67 at 100 trades, saturates at 500. Below 100 trades the overall Robustness Score is also hard-capped at 70 regardless of other components — too few trades to justify ship-tier confidence. See Sample Size Confidence.
Slippage
The difference between the price you wanted and the price you got. Caused by your order size exceeding top-of-book liquidity, or by latency between trigger and order placement. See slippage and execution.
Slippage Drag
Total slippage cost as a fraction of gross PnL — the answer to "how much of my strategy's edge gets eaten by execution?" Reported in the Sizing panel at 1× and across the multiplier sweep. Grows nonlinearly under sqrt-law impact. At 1× a strategy might pay 5%; at 3× the same strategy pays 16% under orderbook fills. See Sqrt-law Market Impact.
Sqrt-law Market Impact
A model where per-contract slippage grows with the square root of order size, so total slippage scales as size to the 1.5 power. Doubling your position pays roughly 2.8× the slippage, not 2×. The Robustness Sizing panel applies this above 1× when the parent backtest used orderbook fills, producing a Kelly recommendation that accounts for the slippage tax. Almgren-Chriss (2000). See Sqrt-law Market Impact.
Stationary Block Bootstrap
The bootstrap variant the Robustness tab uses by default. Resamples in randomly-sized blocks (drawn from a geometric distribution) instead of one trade at a time, preserving the streakiness — clusters of wins, clusters of losses — that real trading produces. The Politis-White rule auto-selects the right block length from your data's autocorrelation. See Stationary Block Bootstrap.
Survival Probability
The conditional probability that a strategy ends profitable given its longest losing streak was at least N trades. Inverts the usual streak question — instead of "how likely is a long streak?" it asks "given that I'm in one, will I recover?" The streak length where the curve crosses 50% is the empirical bail-out point. See Survival Probability.
SMA (Simple Moving Average)
The arithmetic average of the last N prices. Slower to react than the EMA but with cleaner output. Common periods: 20, 50, 200. See SMA.
Sortino ratio
Same as Sharpe but only penalizes downside volatility. A more honest measure for asymmetric strategies that have fat upside tails. If Sortino is much higher than Sharpe, the strategy has good asymmetry.
Spread
The gap between the bid and the ask. The cheapest possible cost of any round-trip trade — paid whenever you cross the market with a market order. Tight in liquid markets (1 tick on ES, MES), wide in illiquid ones.
Stop loss
A protective order that closes a losing trade when price moves a defined distance against it. Limits the downside on any single trade. Implemented in AlgoLift via Set Stop Loss.
Strategy
A defined set of rules for entering, managing, and exiting trades. In AlgoLift, a strategy is a node graph in the Visual Builder.
Systematic trading
Trading governed entirely by pre-defined rules. The opposite of discretionary. Systematic strategies can be tested, audited, and scaled in ways that discretionary trading cannot. See introduction to trading.
T
Take profit (target)
An order that closes a winning trade when it reaches a defined profit level. The counterpart to a stop loss. Implemented via Set Target Profit.
Tick
A single price print — one transaction at one price for one size. The smallest unit of market data. Tick-level backtesting (which AlgoLift uses by default) processes every tick in order; bar-level backtesting compresses ticks into OHLC summaries. See tick vs bar data.
Tick size
The minimum price increment for an instrument. For MES, the tick size is 0.25 points ($1.25 per contract). For ES, also 0.25 points but $12.50 per contract.
Tail Discipline
A Robustness Score component (25% weight, tied for heaviest) capturing what happens when the strategy is unlucky. A composite of P(ruin), the 95th-percentile drawdown, and the probability of a drawdown exceeding 25% of starting capital. 100 when P(ruin) is below 1% and P(DD > 25%) is below 10%; 0 when P(ruin) exceeds 20%. The component that catches strategies which look great until they look catastrophic. See Tail Discipline.
Tail Drawdown
The 95th-percentile peak-to-trough loss across bootstrap-resampled paths. Half of all alternate histories had a worst drawdown smaller than the median; only 5% had one worse than the tail. The number to plan around emotionally — most traders quit during the gap between median and tail because they prepared only for the median. See Tail Drawdown.
Trailing drawdown
A prop firm rule where the maximum allowed drawdown follows your peak equity upward but never moves down. Three real-world variants: intraday-tracking (Apex), end-of-day (Topstep), and locks-at-start (MFFU). The variant matters — intraday is the most punishing because unrealized profits move the floor and give-back trips it even when realized PnL stays positive. The dominant prop firm failure mode for most strategies. See Trailing Drawdown.
Trailing stop
A stop loss that follows price in the direction of the trade, never moving against you. Locks in profit as the trade moves favorably. Common pattern: trail by N ATR.
Trend
A persistent directional move in price. Trends can last from minutes to years. Trend-following strategies bet the move continues; mean-reversion strategies bet it doesn't. See momentum strategies.
U
Ulcer Index
A composite drawdown metric that weights both the depth and the duration of underwater periods. Lower is better. Useful as a tiebreaker between strategies with similar Sharpe and Calmar.
Uncle point
The maximum drawdown at which you'll take a strategy offline and re-evaluate it. Typically 1.5–2× the strategy's historical worst drawdown. Hitting it doesn't mean the strategy is broken — it means the assumption "live ≈ backtest" has been falsified enough to justify a real audit. See risk management basics.
V
Volatility
The magnitude of price fluctuations over time, usually measured as the standard deviation of returns or via ATR. High volatility = bigger moves = wider stops needed, smaller positions. See risk management basics.
Volume
The number of contracts or shares traded over a period. Used to confirm price moves (high volume on a breakout suggests conviction) and to define alternative bar types (volume bars print every N contracts traded, regardless of time). See Volume.
VWAP (Volume-Weighted Average Price)
The average price of an instrument over a session, weighted by volume traded at each price. Often used as an institutional reference level — large orders typically aim to execute "at VWAP." See VWAP.
W
Walk-forward optimization
A robustness test that splits historical data into rolling in-sample / out-of-sample windows. The optimizer finds the best parameters on the in-sample window, then evaluates those parameters on the out-of-sample window. Repeated many times. If the strategy's best in-sample parameters keep working out-of-sample, the edge is plausibly real. See walk-forward optimization.
Warmup
The period at the start of a backtest where indicators are being computed but no trades are taken. A 200-period moving average needs 200 bars before its value is meaningful. AlgoLift handles warmup automatically. See backtesting explained.
Win rate
The percentage of trades that close profitably. Often over-emphasized — a 30% win-rate strategy with 4R wins beats a 70% win-rate strategy with 0.5R wins. See trading metrics.