Trading model terms, defined plainly
Short, jargon-free definitions for the words used across these lessons. Where a term has a whole lesson behind it, the entry points there.
Trading model
A fixed set of rules that maps market inputs to a decision - entry, exit and size - applied the same way every time so the results can be counted and tested.
Rules-based trading
Trading by a specified process rather than by judgement on the day, so the same situation always produces the same call.
Discretionary trading
Trading by judgement in the moment, where the same chart can produce different calls depending on the trader's read - powerful but hard to test or repeat. See: rules-based vs discretionary trading.
Mean reversion
The tendency of a price stretched far from a typical level to drift back toward it; the engine behind most of the models this site discusses. See: what a trading model is.
Testability
The property of a model that lets it be run against history and forward in time to produce a measurable record, rather than merely asserted after the fact. See: the testability lesson.
Measurability
The property that a model's results can be counted in full - decisions, wins, losses and drawdown - rather than recalled selectively. See: the measurability lesson.
Repeatability
The property that a model produces the same call from the same inputs, removing the trader's mood from the decision. See: the repeatability lesson.
Holding clock
How long a model carries a position - from a single session to a long horizon - which is a design choice, not a different idea, and is scored separately for each model.
Conviction grade
An A-to-D label marking how strong a call is relative to its own model's measured return distribution; there is no E grade. See: how grades map to position size.
Drawdown
The deepest peak-to-trough fall in a model's equity over a period - a more honest read on risk than the headline return.
Win rate
The share of a model's calls that closed profitably - meaningful only with the total decision count and the average loss shown beside it. See: reading a record without being fooled.
Cryptographic timestamp
A hash of a call written to a public ledger at publication, proving the entry, target, stop and grade existed in that exact form before the outcome was known. See: repeatability, fixed in advance.
Overfitting
Tuning a rule so tightly to past data that it captures noise rather than a real edge, which is why a forward, timestamped record matters more than a flattering backtest. See: how to build a simple model.