Glossary

Canonical definitions of terms used throughout KEK documentation. Terms are defined in the context of KEK's architecture and strategy lifecycle.

A

Agent

A specialized AI component that performs a narrowly defined task (e.g., analysis, scoring, regime classification, strategy synthesis) within the KEK system.

Allocation

The amount of capital assigned to a strategy, portfolio, or position sizing model (simulated during validation; user-controlled in execution).

API

An interface that allows external systems or clients to programmatically interact with KEK services and data models.

Audit trail

A complete, time-ordered record of strategy changes, validation results, and lifecycle transitions used to support traceability and reproducibility.

Authorization

The explicit user approval required to execute an action that impacts live trading (e.g., wallet signing of a transaction).

B

Backtest

A historical simulation of a strategy against past market data to evaluate performance, risk, and failure modes.

Backtesting & optimization

The validation stage where strategies are tested historically and parameters are tuned to improve robustness and risk-adjusted behavior (not just returns).

Benchmark

A reference index or baseline used to contextualize strategy performance.

Bias

A systematic distortion in validation results (e.g., lookahead bias, survivorship bias) that can cause false confidence.

C

Candidate strategy

A structured strategy proposal generated by the intelligence layer prior to validation.

Churn

Excessive trading activity (high turnover) that may degrade real performance through fees, slippage, and spread costs.

CLOB (Central Limit Order Book)

An order-matching structure where trades occur through bid/ask orders aggregated in an orderbook.

Constraints

Hard limits applied during strategy generation and validation (e.g., max drawdown, leverage caps, position sizing bounds, asset filters).

Correlation risk

The risk that multiple positions behave similarly (highly correlated), effectively concentrating exposure even across different assets.

Custody boundary

The explicit separation between KEK's intelligence/validation systems and external execution infrastructure where user-authorized trades occur.

D

Data source of truth

The authoritative system record for strategy specifications, validation results, monitoring signals, and lifecycle state.

Drawdown

A decline from a portfolio or strategy's peak value to a subsequent trough, often measured as a percentage.

Drift

A measurable change in strategy behavior or performance over time due to regime changes, liquidity shifts, volatility changes, or execution constraints.

E

Edge

A repeatable advantage or mechanism that improves expected outcomes over time, after costs and risk are accounted for.

Event-driven backtesting

A backtesting method that simulates sequential market events and state changes (useful when fill logic and path dependency matters).

Execution (optional)

The stage where a user chooses to deploy a validated strategy for live trading through a non-custodial execution rail.

Execution metadata

Stored execution-context information (timestamps, venues, chain, order type, transaction references) used for monitoring and auditing.

Execution rail

The downstream infrastructure layer that routes and settles trades (external to KEK custody).

F

Fees

Transaction costs incurred by execution (trading fees, network fees, and implicit costs like spread and slippage).

Fill

The completion of an order (full or partial). Fill quality depends on liquidity, spread, volatility, and order type.

Funding rate

A periodic payment mechanism used in perpetual futures markets to keep contract price aligned with spot price.

G

Governance

A formal process for ecosystem-level decision making (proposals, voting, and coordination), separate from execution authority.

H

Hypothesis

A strategy concept treated as a testable claim rather than an assumed edge. Strategies must earn execution eligibility through validation.

I

In-sample

The portion of historical data used to tune parameters during optimization.

Indicator

A computed feature derived from market data (e.g., moving average, RSI, volatility bands) used in strategy rules.

K

KEK DEX

The execution interface that routes user-authorized trades through non-custodial infrastructure. KEK DEX is omnichain and built on Orderly Network's decentralized orderbook layer.

KEK Mix

The strategy generation and optimization engine that outputs machine-readable strategy specifications and supports validation workflows.

KEK Terminal

The primary interface for interacting with KEK's strategy lifecycle: research, strategy management, monitoring, and refinement tracking.

Knowledge network

A domain-specific intelligence layer that organizes market knowledge, signals, and outcomes to support agent reasoning and strategy improvement.

L

Latency

Delay between signal generation and execution (or simulated execution). Latency affects fill quality and slippage sensitivity.

Leverage

Borrowed exposure that amplifies returns and drawdowns. Leverage increases liquidation risk and execution fragility.

Lifecycle stage

The current position of a strategy within KEK's enforced pipeline (generated → backtested → paper traded → monitored → optionally executed).

Liquidation

A forced position close caused by insufficient margin in leveraged trading.

Liquidity

The capacity of the market to absorb orders with minimal price impact. Low liquidity increases slippage and spread sensitivity.

Lookahead bias

A backtesting error where future data is implicitly used to generate past signals, inflating results.

M

Margin

Collateral required to open and maintain leveraged positions.

Market impact

The adverse price movement caused by executing an order large relative to available market depth.

Market order

An order executed immediately at the best available prices (more slippage risk).

Matching engine

The infrastructure that matches buy and sell orders in an orderbook-based market.

MCP Agent Server

The orchestration layer coordinating specialized agents, tools, and shared context via MCP (Model Context Protocol). MCP is an open standard connecting AI applications to external tools and systems.

Meta-learning

A structured improvement loop that uses observed performance signals to refine strategy generation, selection, and validation over time.

Metrics (performance)

Quantitative measurements used to evaluate strategy quality (returns, drawdown, Sharpe, profit factor, etc.).

N

Non-custodial

A custody model where the platform does not control user funds or private keys; execution requires explicit user authorization.

Normalization

Transforming data into consistent formats/scales to enable stable features, indicators, and modeling.

O

Omnichain

An execution or settlement model spanning multiple chains while maintaining unified access and liquidity routing where supported.

Optimization

Systematic parameter exploration intended to improve robustness within constraints (not curve-fitting).

Orderbook

A real-time list of bids and asks representing market liquidity at different prices.

Orderly Network

External omnichain decentralized orderbook infrastructure providing high-performance, builder-facing perpetual futures trading rails.

Out-of-sample

The portion of historical data reserved to test generalization after in-sample optimization.

Overfitting

When a strategy becomes tuned to noise in historical data and fails to generalize out-of-sample.

P

Paper trading

A live simulation of strategy execution using real market conditions without deploying real capital.

Parameter

A tunable numeric or categorical strategy input (e.g., lookback length, threshold, risk cap).

Parameter sensitivity

How strongly performance changes when parameters are adjusted; high sensitivity usually implies fragility.

Perpetual futures

A derivative contract with no expiry that uses funding mechanisms to track spot price.

Portfolio

A set of positions and sizes evaluated as a combined risk system (correlation and concentration matter).

Position

An open long or short exposure in an asset, with size, entry basis, and risk constraints.

Position sizing

Rules that determine trade size based on risk limits, volatility, equity, or regime.

Profit factor

A ratio comparing gross profits to gross losses (a profitability efficiency metric).

R

Rebalancing

Adjusting portfolio weights or exposures back toward target allocations.

Regime

A classification of market conditions (e.g., trending, ranging, high volatility) used to condition strategy evaluation and behavior.

Risk-adjusted return

Return evaluated relative to risk taken (e.g., Sharpe, Sortino).

Risk controls

Validation-stage constraints and analysis designed to surface unacceptable risk before execution eligibility.

S

Sharpe ratio

A widely used risk-adjusted return metric comparing excess return to volatility.

Simulation

A controlled reproduction of trading behavior (historical or live-like) used to evaluate strategies without real capital.

Slippage

The difference between expected execution price and actual execution price, influenced by volatility and liquidity.

Smart contract

On-chain program that can execute transactions deterministically once authorized by a user signature.

Source of truth

The canonical store of strategy state and performance evidence used to coordinate monitoring and refinement.

Spread

The difference between the best bid and best ask; a core execution cost component.

Strategy

A defined set of rules and parameters specifying entry/exit logic, sizing behavior, risk constraints, and execution assumptions.

Strategy specification

The formal machine-readable definition of a strategy used across generation, validation, monitoring, and optional execution.

Stress testing

Testing strategy behavior under extreme or adverse assumptions (volatility spikes, liquidity shocks, regime breaks).

Survivorship bias

A backtesting bias where only assets that survived are included, inflating historical performance.

T

Take profit

A rule that closes a position at a defined profit threshold or signal condition.

Time-series database

A database optimized for storing and querying time-indexed metrics and events (used for performance tracking and monitoring).

Transaction (on-chain)

A user-signed action submitted to a blockchain network (trade placement, position management, settlement).

V

Validation

The enforced evaluation process (backtesting → paper trading → monitoring criteria) required before execution eligibility.

Validation gate

A pass/fail threshold or rule that determines whether a strategy advances to the next lifecycle stage.

Variant

A parameterized version of a strategy created to evaluate trade-offs in return, risk, trade frequency, and robustness.

VectorBT

A high-performance Python backtesting library using vectorized operations and Numba acceleration for efficient simulation and analysis.

Volatility

The magnitude of price changes over time; a driver of risk, slippage, and drawdown behavior.

W

Walk-forward analysis

A validation method that repeatedly optimizes on an in-sample window and tests on the following out-of-sample window to evaluate robustness.

Wallet signature

A cryptographic approval from the user authorizing a transaction. The core enforcement mechanism for non-custodial execution.

Win rate

The percentage of trades that close profitably (not sufficient alone as a quality metric).

Y

Yield / returns (non-guaranteed)

Any reference to outcome measures must be treated as non-guaranteed and conditional on risk, costs, and market regimes.