The Lore library
The concepts behind agentic trading.
16 working notes on market microstructure, risk, onchain rails, and the models an agent uses to trade. Plain-English definitions, worked intuition, and why each one matters when you hand a strategy to an autonomous agent.
Probability & Statistics
The math primitives an agent runs on — distributions, entropy, hypothesis tests, and Bayesian updates.
Linear Algebra & Optimization
Vectors, matrices, and convex problems — the toolkit behind portfolio weights and gradient descent.
Time Series & Stochastic Processes
Anything that moves over time — stationarity, autocorrelation, regimes, and random walks.
Machine Learning
Supervised, unsupervised, and RL — from logistic regression to LLM-driven trading agents.
Market Mechanics
How venues clear trades — order books, makers, takers, spreads, and adverse selection.
Onchain Mechanics
Where crypto rails diverge from TradFi — perps, funding, oracles, and prediction markets.
Risk & Sizing
How much to bet — Kelly, drawdown, Value at Risk, and fractional sizing under uncertainty.
Agents & Operations
Building autonomous strategies that survive — lifecycle, governance, and capital allocation.
Recently added
Adverse Selection
When your resting limit order gets filled instantly, the counterparty is often better-informed than you. Adverse selection is the systematic loss makers absorb on those picked-off fills.
Agentic Trading
Agentic trading is the use of LLM-driven agents to generate, evolve, and operate trading strategies — not just to produce one signal among many, but to run the loop end to end under human-set policy.
Calibration
A forecaster is calibrated when things it says are 70% likely happen 70% of the time. Calibration measures whether stated probabilities match reality — the foundation of honest sizing.
Funding Rate
Perpetual futures never expire, so exchanges run periodic funding payments between longs and shorts to tether the perp price to spot. The funding rate is both a cost and a signal.
Gradient Descent
Gradient descent minimizes a function by repeatedly stepping in the direction of steepest decrease. It is the workhorse that fits nearly every model an agent relies on — and portfolio weights too.
Hawkes Processes
A Hawkes process is a self-exciting point process: each event temporarily raises the probability of the next. It models the clustering and bursts seen in order arrivals and trades.