Algorithmic trading (Wikipedia)

The encyclopedic overview of algorithmic-trading — the neutral, third-party (T2) anchor for the spoke, whose three founding sources are all author-published T3 repos (banbot, tensortrade, kronos-financial-foundation-model). It grounds the field’s scale, its strategy taxonomy, and the backtesting discipline in a source with no product to sell.

Scale — and how far past crypto it reaches

The founding corpus is crypto-exchange-centric, but algorithmic execution dominates the established markets too, which directly answers the spoke’s “beyond crypto” open question:

US equities — HFT firms were ~60–73% of volume by 2009, settling near ~50% by 2012; ~2% of firms drive the bulk of that volume.
Foreign exchange — roughly 80% of orders algorithmic by 2016 (up from ~25% in 2006); a 2019 study put FX algo participation around 92%.
Options ~40% (2016); futures and, increasingly, bonds follow.

So the crypto bots of the founding trio are a recent entrant to a field already standard in equities and FX.

Strategy taxonomy

The article catalogs the strategy families a trading system‘s decision layer implements: arbitrage, pairs / statistical arbitrage, mean reversion, scalping, market making, momentum / trend following, index-rebalancing arbitrage, and event arbitrage (merger arb, regulatory or contract triggers) — the last being the market-level cousin of the event-driven engine concept.

The validation ladder

It frames strategy development as three stages, which sharpens this spoke’s backtesting page:

Backtesting on historical data, with input-optimization (and risk hygiene: perturb inputs ±10%, Monte Carlo, outlier removal, model slippage and commissions).
Forward testing on out-of-sample data — the explicit guard the article stresses against over-fitting (the encyclopedic framing of the lookahead/over-fit hazard, even though it doesn’t use the term “lookahead bias”).
Live testing, comparing real fills against the backtested and forward-tested models.

Machine learning’s place

Early algo trading ran hand-coded rules on technical indicators; the article notes deep reinforcement learning (reinforcement-learning-trading) as the adaptive successor — “learns and optimizes iteratively,” strongest in volatile regimes where static rules falter — plus directional-change methods. This independently corroborates the predict/learn split the founding trio embodies.

Why it matters here

A neutral T2 base under the spoke: it lifts the floor off all-T3, sizes the field (algo trading is the norm in equities/FX, not a crypto novelty), and grounds the backtesting/over-fitting claims the self-published founding sources assert about themselves.

algorithmic-trading · backtesting · event-driven-trading · reinforcement-learning-trading · synthesis