Models suggest.
Evidence decides.
We use machine learning to search widely and quantitative discipline to choose narrowly. People design the process, review the evidence and decide what is allowed to trade. Every rule has passed through the same four gates, in the same order.
From raw data to a decision.
Data
Every model is only as honest as its inputs. We collect market prices, full options chains, news and macroeconomic releases, align them to the moment they were actually known, and keep the raw history so any result can be rebuilt from scratch. Look-ahead bias is treated as a defect, not a nuance.
Models
Machine learning is how we search a very large space of possible ideas quickly. A model's output is a hypothesis: a candidate rule with a stated entry, exit and holding period. It earns nothing by being clever, and it does not reach the book on the strength of a good chart.
Proof
Each candidate is tested on years of history it was never fitted to, in rolling walk-forward windows, after realistic spreads, commissions and slippage. It must also beat placebo versions of itself (the same rule fired at random times) and hold up in both halves of the sample. Most ideas fail here, which is the point.
Risk
Sizing, exits and exposure limits are set by a separate risk layer that can overrule any signal. Losses are capped per trade and per day, positions are reconciled against broker records, and trading halts automatically when anything does not match, until a person has reviewed it.
A powerful tool, held to a strict standard.
Artificial intelligence lets a small team search far more ideas than it could by hand. It also makes it easy to fool yourself. Here it is a research tool, not the decision-maker: people set the rules, review every change and own every outcome. These are the standards we hold it to.
Risk has the last word.
Between every signal and the market sits an independent risk layer. It sizes positions from the amount at risk rather than from conviction, enforces hard per-trade and daily loss limits, caps exposure by market and in aggregate, and reconciles our records against the broker's. When anything does not reconcile, trading halts automatically until a person has reviewed it.
Liquid markets, one platform.
We research and trade liquid, exchange-listed markets: US equities and equity options, index and commodity futures, and major currencies. Liquidity is a requirement, not a preference; a strategy that cannot be traded at a fair price is not a strategy.
All of it runs on TipSheet, the research and execution platform we built and operate ourselves. It collects and cleans the data, runs the tests, monitors live rules against their records and enforces the risk limits on every market day, with the results reviewed by us daily.