Glossary

Terms used throughout the platform and these docs, defined precisely.

---

Term	Definition
Asset class	A category of instrument, equity_cash, crypto_spot, equity_option, index_future, crypto_perpetual and others. It says what kind of thing you are trading, never where: an Indian and a US cash equity are both equity_cash and reach different brokers. The market is the venue. What is reachable differs from what is defined, and what is live-tradeable is narrower still, see Instruments.
Backtest	A simulation of a strategy's rules applied to real historical price data to estimate past performance.
Capital allocated	The amount of capital assigned to a specific strategy. Enforced as a ceiling - the strategy cannot trade more than this.
Capital used	Capital currently locked in open positions for a given strategy.
Badge	An automatic performance tag on a Marketplace listing card (e.g. Elite Sharpe, Lowest Risk). Each strategy shows its single highest-priority earned badge. Some are cohort-relative, some clear a fixed threshold.
Collection (watchlist)	A named folder for saved Marketplace strategies. Each saved strategy belongs to one collection at a time.
Candle close	The moment a candle's time period ends and its final OHLCV values are set. Entry and exit conditions are evaluated at candle close.
Dynamic strategy	A strategy that evaluates and trades a universe of assets rather than a single instrument. The AI ranks all assets in the universe and trades the top-ranked ones.
Egress IP	A static outbound IP address Stretus can assign your account, added to an exchange API-key allowlist so live orders aren't blocked by IP restrictions.
Entitlement	A specific capability your Stretus plan grants - a feature or usage limit. Locked features prompt a plan upgrade.
Entry condition	An indicator-based rule that must be true for a position to open. Multiple entry conditions are combined with AND logic.
Exit condition	An indicator-based rule that triggers position closure. Multiple exit conditions use OR logic - any one true causes exit.
Featured / Verified	A Marketplace listing curated by Stretus admins into featured/leaderboard placement. Curation is platform-decided, not self-serve; "Verified" filters to these listings.
Grade (A, B+, B, C+, C, D)	Six grades from a 0-100 score summing five components: return (30), Sharpe (25), drawdown (25), consistency (15) and sample size (5). A threshold score over the tested window, not a detection of edge, a forecast, or a recommendation. Zero trades grades D immediately. See Strategy grades.
Health score	A 0-100 composite for a published listing: 0.35 times risk control, plus 0.25 times return quality, plus 0.20 times consistency, plus 0.10 times trade density, plus 0.10 times profile discipline. Zero means no live activity. Rendered as Excellent (≥85), Healthy (≥70), or At Risk (anything below 70, including 0), there is no "Unrated" label. Formula and weights: How listings are ranked.
Instrument	The specific tradable item - e.g. the ETH/USDT spot pair on Binance. Instruments belong to an asset class and trade on a venue.
Leaderboard	A ranking of published strategies and traders by real performance (return, Sharpe, drawdown), filterable by asset class, risk, and time window.
Live	Execution mode where real orders are placed at a connected broker or exchange. Real capital is at risk.
Margin used %	capital_used / capital_allocated × 100. At 100%, the strategy has no free capital and cannot open new positions.
Multi-target take-profit (TP1/TP2)	A staged exit that closes part of a position at a first profit target and more at later targets, banking gains while letting a portion run.
Ownership NFT	An optional on-chain record proving authorship of a strategy. Concerns ownership provenance only - not capital, custody, or profit rights. Availability is gated.
Max drawdown	The largest peak-to-trough decline in a strategy's equity curve during a given period. Expressed as a negative percentage.
Paper	Execution mode where trades are simulated by the in-platform engine. No real orders are placed. No real capital is at risk.
Per-trade risk %	The maximum percentage of allocated capital to risk on a single trade. Used with stop-loss % to determine position size.
Profit factor	Total gross profit divided by total gross loss. A profit factor > 1 means the strategy made more than it lost.
Research Analyst (RA)	A person registered under the SEBI (Research Analysts) Regulations, 2014, verified against documents on this platform, permitted to publish non-crypto strategies. Not an Investment Adviser, a different registration under different regulations with different permitted activities. No Stretus role holds Investment Adviser capacity. See Research Analyst programme.
Remote (blackbox) strategy	A strategy whose signals are generated outside Stretus and sent in over an authenticated endpoint (API key + HMAC), then executed through Stretus's risk and capital controls.
Sharpe ratio	Annualised excess return divided by annualised volatility. Measures risk-adjusted performance. Higher is better; negative means underperforming cash.
Stop-loss	A price level at which an open position is automatically closed to limit losses. Stretus evaluates stops on every price tick.
Strategy	A complete set of rules defining what to trade, when to enter, when to exit, how to size positions, and when to trade.
Trailing take-profit	A take-profit that moves with the price. It arms at a threshold gain, then trails below the peak price. Position closes when price falls to the trail level.
Venue	The market an instrument trades on, NSE, NASDAQ, LSE, ASX, Binance Spot. A venue determines the currency, the session, the calendar and which broker can reach it. Identified by a short code, never by its MIC: NASDAQ, not XNAS. See Markets & brokers.
Quote unit	The unit a venue publishes prices in, which is not always the currency it settles in. LSE quotes pence (GBX) and settles pounds (GBP). See Currencies & the GBX case.
Watchlist	A private list of Marketplace strategies you're tracking without subscribing. The advisor is not notified.
Win rate	The percentage of completed trades that closed in profit. Meaningful only in context of the average win/loss ratio.

---

AI & strategy compilation​

Term	Definition
Router	The component that reads a message and picks exactly one tool per turn from a catalog of 18, filtered to the tools valid in the current conversation phase. An unrecognised phase returns the full catalog as a fail-safe.
Extractor	A second, concurrent read of the whole message against a wide schema of strategy fields, independent of which tool the router picked. It exists because one tool call could only ever carry a fraction of what a message states.
Intent	The resolved meaning of your request: instrument, timeframe, objective, direction, rules, risk controls and gates, each as a slot carrying a value and its provenance.
Provenance	Where a value came from: stated (you said it), defaulted (the platform supplied it), derived (inherited from earlier in the conversation), or unsupported (you asked and the platform cannot express it).
Compiler	Turns an Intent into a formal strategy object, or into typed gaps naming what it could not read and why.
Gap	One span the compiler could not read, carrying the exact words and a reason.
UNKNOWN_PHRASE	No pattern matched. A model may still read it, so it is retried narrowly.
MISSING_THRESHOLD	A real comparison with no number, "near the 20 EMA", "strong volume". Structure understood, value absent.
AMBIGUOUS_SCOPE	"unless", "except", "other than", "rather than", words that re-scope everything around them. Never retried, because the failure mode of guessing is an inverted rule that looks plausible. You are asked.
NOT_REPRESENTABLE	No field on the strategy specification can hold it. Refused, never silently approximated.
Repair pass	The bounded second attempt at spans the first pass could not close. Exactly two passes, never three; only the failed spans are re-read, and every patch is parsed against the engine grammar before insertion.
Disposition	What happens about a validation finding: REPAIR (the right value is known, apply and note it), DEGRADE (drop an optional element and note it), ASK (only you know), BLOCK (no runnable strategy exists). An unregistered finding falls back to BLOCK.
Fail-closed	A control that blocks when it cannot evaluate. The strategy validator hard-errors if the engine is unreachable; the derivatives kill switch blocks entries if the halt state is unreadable.
Fail-open	A control that permits when it cannot evaluate. The execution risk-config parser yields no enforced rules on a malformed payload, so a bad payload can never block a strategy trying to exit.
Safety boundary	The dedicated refusal path for requests asking the platform to recommend a trade. Returns a composed template, not a generated reply, so it cannot be talked around.
Intent cache	Compile results cached on what an intent means rather than on the prompt text, so infinitely many phrasings collapse onto finitely many intents.

Backtesting & metrics​

Term	Definition
Cost model version	A counter stamped on every stored backtest result (currently 4) so results computed under different cost assumptions can be told apart rather than silently compared.
Slippage (bps)	Modelled market impact, spread plus execution imperfection. Default 5 bps.
STT	Securities Transaction Tax. Indian equity: 0% buy / 0.025% sell intraday, 0.1% both legs on delivery. F&O: sell-side only, however long held. Zero on crypto and US equity.
Statutory charges	Transaction taxes and regulatory fees applied by the simulator. For Indian F&O: STT, stamp duty, exchange transaction charges, IPFT, the SEBI turnover fee, and 18% GST on the fee components.
Gross vs net return	Return before and after modelled costs. The gap is your cost drag, and it is the most reliable single diagnostic for an over-trading strategy.
Profit factor	Total gross profit ÷ total gross loss. Above 1 means the strategy made more than it lost.
Sample size component	5 of the grade's 100 points, from trade count. What stops a three-trade backtest scoring an A, though only just, so read the trade count yourself.
Recovery time	Days from a drawdown's trough back to the prior peak. Absent, not zero, when the peak was never regained within the window.
VaR (95%)	The 5th percentile of the daily-return distribution. Not a maximum loss, 1 day in 20 is worse by definition.
Expected shortfall (95%)	The mean of returns worse than VaR(95). Describes what happens inside the tail rather than where it starts.
Calmar ratio	Annual return ÷ max drawdown. Returns 0 when there was no drawdown.
Risk profile match	A computed label describing what a strategy demands, its drawdown depth, recovery time and trade frequency. Not a suitability assessment, and not a statement about you.
Pass / fail	A boolean separate from the grade, against objective-aware thresholds: 40% minimum win rate for both objectives, plus a 1.2 profit-factor floor for positional only.
Intra-candle path	The order in which a candle visited its high and low, which the simulator cannot know. When a stop and a target share one candle, the stop is assumed to fill first, deliberately pessimistic.

Risk & execution​

Term	Definition
Order gate	One of nine validators evaluated on every order before it reaches a broker. The first violation stops the order. Seven are disabled until you configure a value, zero means off.
max_order_value	Ceiling on a single order's notional, in quote currency.
per_trade_risk	Caps stop-distance × quantity against a percent of allocated capital. Skips when the entry carries no usable stop price.
max_open_positions	Concurrent open entries, FIFO buy lots plus fully-unfilled working buy entries. Working orders occupy a slot.
max_trades_per_day	New entries per UTC day. The counter rolls at 00:00 UTC, which is 05:30 IST.
max_consecutive_losses	Halts new entries after N losing round-trips. Resets on a win, or clears via cooldown.
daily_loss_limit	Halts entries once today's realised loss breaches a percent of allocation.
mark_to_market_loss	Halts entries once realised + unrealised loss is down by an absolute amount. Skips when a market order carries no usable mark.
max_capital_allocation	Ceiling on total deployed capital.
available_capital	Rejects an order the strategy cannot fund from free capital. Cannot be disabled.
Cooldown	Bars to sit out after a consecutive-loss halt, at the strategy's own timeframe. Default 3 bars. Computed on every evaluation, never persisted.
Trading window	A time-of-day range for entries, interpreted in IST. Carries no day-of-week or holiday information.
Violation	A blocked order's record: the rule, a stable reason code, the current value and the limit.
Idempotency key	A stable identifier on every signal, unique per user, that prevents a duplicate order from a retry or restart. A signal without one is refused rather than sent.
Out of sync	A reconciliation state where the platform's view of an order or position differs from the broker's.
Kill switch	A halt on new derivative entries at global, tenant or strategy scope. Fail-closed.

Derivatives​

Term	Definition
Lot size	The exchange-set number of units in one derivative contract. Positions are whole lots, there is no fractional lot, so a lot's risk can exceed your per-trade budget.
Strike rule	How a strike is chosen: ABS:<price> (spot-independent), or ATM, ATM±n, OTM+n, ITM+n as ladder offsets. Offsets are oriented by option right. Resolution is fail-closed.
Expiry rule	nearest, weekly, next_weekly, monthly, next_monthly, or a date.
Structure	The named shape formed by a group's legs, from long_call through iron_condor. custom is the honest answer for a shape the classifier will not name, not an error.
Coordinated close	Closing a multi-leg group as a group. Prevents ending up with a naked short option by closing three legs of four.
Gross margin	Margin computed per leg with no spread or hedge benefit, so a bull call spread reserves as if the legs were unrelated. Conservative, and it bounds what an allocation can hold.
Assignment	A written option being exercised against you, creating an obligation in the underlying.
Premium vs underlying	An option's price versus its underlying's price. A percentage stop on a premium is a far larger move than the same percentage on the underlying.

Discovery & research​

Term	Definition
Setup Quality	A 0-100 score for how completely a detector's conditions are met: Setup Strength (35) + Liquidity (25) + Regime (20) + Timing (10) + Quality (10), minus penalties up to 30. Not a probability of profit.
Setup Grade	A band derived from Setup Quality: STRONG ≥80, GOOD ≥65, WATCH ≥50, else LOW. Not statistical confidence.
Trigger	The price at which a detector considers a setup activated.
Invalidation	The price at which a setup is void, a statement that the pattern no longer exists. Not a suggested stop-loss.
Trigger distance	How far current price sits from the trigger, as a percentage.
Why now	Plain-language evidence statements behind each scoring component, computed from the same numbers as the score.
Event risk	Scheduled events near a setup. Applies a scoring penalty and is listed separately.
Liquidity band	HIGH (≥18 points), MEDIUM (≥10), LOW, from the liquidity component.
Freshness	FRESH, DELAYED or STALE, with an age in seconds. Stale data both penalises the score and is labelled.
Snapshot	An addressable result set with an id, an as-of timestamp and a scoring version. Discovery screens read snapshots, not a tick feed.
Publication floor	Candidates scoring below 50 are never surfaced. You are seeing a filtered set by design.
Opportunity status	FORMING, READY, TRIGGERED, INVALIDATED, EXPIRED. Only READY and TRIGGERED are published.
Detector	A defined technical condition: breakout, momentum continuation, or volatility expansion. Each publishes its key and version.
Cluster density	The concentration of setup activity in one market slice. A measure of pattern frequency, not of opportunity.
Coverage	The proportion of a slice's instruments actually evaluated. Every aggregate is computed over instruments with bars, not the full count.
Analog event	A historical session matching a defined condition, with the next session's outcome. Seven event types, each precisely specified.
Similarity score	How closely a historical session resembles today, over features available for every session since 2010.
Outcome class	POSITIVE (≥ +0.5%), NEUTRAL, NEGATIVE (≤ -0.5%) for the next session's close-to-close move.
Opportunity watchlist	Saved discovery candidates, at /watchlist/opportunities. Distinct from the strategy watchlist.
Verification pack	A Trading Map feature exposing the score formula, every component with its formula, evidence mapped to components, penalties, and the raw inputs, so the arithmetic can be reproduced.
Rejection reason	Why a scanner candidate did not match: FAILED_FILTER, METRIC_UNAVAILABLE, or NO_BARS. Distinguishing these is the scanner's most useful output.

Community & publishing​

Term	Definition
Strategy author	A user who created a strategy. The term this documentation uses instead of "advisor".
Publisher	A user who published a strategy for others to deploy. Not an adviser, and owes deploying users no duty of care.
Community-published strategy	A strategy a user published. Deploying it clones its rules into your account. It is not copy trading, mirror trading, or a signal relay.
Open Strategy	A listing that discloses entry and exit conditions, indicator formulas and parameter values to subscribers, under a recorded licence. The opposite of a closed listing. Disclosed logic cannot be recalled from copies already made.
Closed listing	The default. Track record, risk profile, controls and instrument are public; the logic is private.
Publish consent record	The durable record written when Open Strategy is chosen: legal document version, a hash of the exact prompt text licensed, IP and user agent. Consent must be explicitly true, a missing value is a refusal.
Prompt hash	Pins which prompt text was licensed. The strategy prompt regenerates when configuration changes, so without it the record would only prove that some prompt was licensed.
Eligibility rule	One publish check: ownership, status gate, minimum days live (7), minimum live return (5% native / 10% remote), and the Research Analyst gate. Only ownership cannot be disabled.
Bypass switch	A master kill-switch that disables every configurable publish check for all tenants. Its documentation says the default should be false; the code default is true. Do not infer permission to publish from a successful publish.
Verified	Confirms identity and listing integrity. Not an assessment of merit or suitability.
Featured	Editorial placement by the platform. Not a judgement of quality.
Risk tier	Low / Medium / High, derived from a listing's live drawdown profile. A description of observed behaviour.
Cohort-relative	A ranking or badge computed against what is currently on screen, so it changes as you filter.
Minimum capital	The floor a subscriber must allocate, the larger of the publisher's figure and the platform's US$50 minimum. Enforced at live deployment.
Subscription price	A figure a publisher may set. It is displayed and never charged, no payment path exists.

Platform & data​

Term	Definition
Tenant	An organisation or deployment. Scopes venue activation, enabled broker adapters, asset-class access and AI models.
Adapter catalogue	Reference data describing each broker adapter's endpoints, field mappings and venue capabilities. Adding a broker is adding data, not code.
Tenant adapter	A per-tenant row carrying an enabled flag. A broker you cannot see is not enabled for your tenant, not absent from the platform.
Egress IP	A static outbound address the platform can assign your account, so a source-IP-restricted exchange API key still works.
Platform gate	A capability disabled globally for regulatory or safety reasons, live derivative entry, retail OTC forex, paid subscriptions. No plan upgrade reaches one.
Corporate action	An immutable, source-versioned event: split, cash dividend, merger, spin-off, symbol change, redemption, coupon. A revision is a new event linked to the one it supersedes.
Minimum bars	The history a metric needs to be meaningful. Below it the value is absent, not zero, a 60-day return on 30 bars is not a flat 0%.
Dynamic universe	A strategy that trades a selected set, reselected each bar by a ranking metric. Different from a multi-symbol backtest.
Membership mode	point_in_time (membership as it was on each date) or approximate (today's list applied across history, survivorship bias by construction).
Multi-symbol backtest	The same strategy run independently on several symbols, each with the full starting capital. Not a portfolio.
Higher-timeframe gate	A condition that must pass on its most recently closed higher-timeframe bar before a lower-timeframe entry can fill.
Knowledge-base signal	One of 31 named composite conditions with sensible defaults, usable directly, including India-specific ones like not_near_circuit and avoid_expiry_day.
GROUP:KEY	The address of a custom or alternative-data indicator, e.g. INDIA_EQUITY_FINANCIALS:CURRENT_RATIO. The builder never invents one.
Target Moves	Chart-guided strategy generation: you annotate historical moves and deterministic pattern matching, no model, produces testable hypotheses. The third Improve mode.
Telegram binding	An opt-in link between a Telegram chat and your account or a strategy, created through a single-use 15-minute deep-link token, revoked automatically if you block the bot.