Every trading day is logged with its profit/loss, win/loss count, and running account balance. On days the regime filter signals YELLOW or RED, the system sits out — those days appear with $0.00 P/L and a documented reason. No results are excluded, edited after the fact, or selectively published.
A volatility measure — the average daily price range over 14 days. Used to set stop distances.
The mean number of shares traded per day over a lookback period. Higher = more liquid.
A unit of measurement equal to 1/100th of a percentage point (0.01%). Used to express small differences in execution cost, slippage, or yield. For example, 20 basis points equals 0.20%. In trading systems, small basis-point differences in fill quality compound into material differences in annual performance.
How often an AI model gives the correct answer without any project-specific context — relying only on its training data. In our contradiction detection test, we presented false claims about our system and measured whether models caught them cold. Higher blind accuracy means the model is less likely to accept whatever it's told — it pushes back based on what it learned during training, rather than deferring to context.
Buying on margin means using money borrowed from your broker to purchase more shares than your cash balance alone would allow. While margin can amplify gains, it equally amplifies losses — and if your account falls below the broker's minimum, you face a margin call, forcing you to deposit cash immediately or have positions liquidated at the worst possible moment.
A trading maxim meaning that choosing not to trade is itself an active decision. Traders can be long, short, or flat — and flat (cash) is a deliberate position. Sitting out a choppy market is as valid as any entry.
The news event driving a stock's move — earnings, FDA approval, contract win, etc. Gaps without catalysts tend to fade.
A hard maximum-loss exit rule that closes a position immediately if it drops below a predefined threshold from entry. Unlike a trailing stop, the catastrophic floor is measured from the entry price and does not move. It exists to cap worst-case single-position losses, not to optimize returns.
Context Compliance Delta — our metric for measuring AI sycophancy. Calculated as: Score(with KB context) minus Score(blind). A high CCD means the model changed its answer dramatically when given context, suggesting it follows whatever it's told rather than reasoning independently. Lower CCD = more trustworthy judgment.
The designated human decision-maker in a human-AI collaborative system. All strategic decisions, production code changes, and spending require CHA approval. Ensures human oversight of autonomous AI operations.
A trailing stop that hangs from the highest high, using ATR to set distance. Prevents cutting winners short.
A statistical measure of effect size for categorical data. Ranges from 0 (no association) to 1 (perfect association). V = 1.0 means the independent variable completely determines the outcome. Named after Swedish mathematician Harald Cramér.
Buying and selling stocks within the same trading day, closing all positions before 4:00 PM ET.
Peak-to-trough decline in portfolio value. Max drawdown measures worst-case scenario.
Technology-driven innovation in financial services, encompassing digital payments, robo-advisors, algorithmic trading platforms, and blockchain-based systems.
The number of shares available for public trading. Low float stocks move faster on volume.
A day trading strategy that buys stocks gapping up on high volume, riding momentum through the trading day.
When a stock opens higher than its previous close. A 5% gap-up means it opens 5% above yesterday's closing price.
A checkpoint in the research flywheel where findings are evaluated against pre-defined acceptance criteria before advancing to the next stage. The Chief Human Agent makes the go/no-go decision at each gate. Negative results are documented with the same rigor as positive ones. No parameter reaches production without passing all gates.
A validation methodology that splits historical data into two non-overlapping periods. The in-sample period is used to develop and tune a strategy; the out-of-sample period tests whether the findings generalize to unseen data. This guards against overfitting — discovering patterns that worked historically but have no predictive power.
An order to buy/sell only at a specific price or better. Used to control entry/exit prices.
An order to buy/sell immediately at the current best available price.
The tendency of a price or indicator to return toward its historical average over time. A mean-reverting strategy bets that extreme moves (like large gaps) will partially reverse — the opposite of a momentum or trend-following approach.
The tendency of a stock moving in one direction to continue moving that way. Measured by rate of change.
The price range established in the first 5-15 minutes of trading. Breakouts above/below signal direction.
The dollar or percentage gain/loss on a trade or portfolio.
Simulated trading with fake money to test strategies before risking real capital.
How much capital to allocate per trade. We use equal $10,000 positions across 5 picks.
Trading that occurs before the regular 9:30 AM ET market open. Used to gauge overnight demand.
Gross profits divided by gross losses. Above 1.5 means winners significantly outpace losers.
A traffic light system (GREEN/YELLOW/RED) that gates trading activity based on broad market signals.
Technology solutions that help financial firms comply with regulations efficiently. Includes automated compliance monitoring, KYC/AML, and real-time reporting tools.
Risk arbitrage, also called merger arbitrage, is a trading strategy that attempts to profit from the price gap between a company's current stock price and the price offered in a pending acquisition or merger. The "risk" refers to the possibility the deal falls apart, which can cause the stock to drop sharply.
The training process that makes AI models helpful and conversational. Human raters score model responses, and the model learns to produce answers that get higher scores. The side effect: models learn that agreeable, confident answers score higher than honest, uncertain ones — which is why more RLHF training correlates with more sycophancy.
Today's volume compared to the 15-day average. RVOL > 1.0 means heavier than usual trading — a sign of institutional interest.
The practice of recording what would have happened on days the system chose not to trade. On YELLOW and RED regime days, the scanner still runs and logs hypothetical picks, tracking their intraday performance without placing orders. This produces a parallel dataset proving the regime filter's value.
Risk-adjusted return metric. Above 2.0 is excellent, above 4.0 is elite for a retail algo.
The difference between expected and actual fill price. A key cost in any trading system — wider spreads and volatile conditions increase slippage.
The difference between the best bid and ask prices. Tighter spreads mean lower trading costs.
A pre-set price that triggers an automatic sell to limit losses. Our system uses ATR-based stops.
A statistical distortion that occurs when backtests only include securities that still exist today, excluding companies that were delisted, merged, or went bankrupt. This inflates apparent performance because the worst-performing stocks are missing from the dataset. MorningEdge tests against 16,000+ tickers including 7,000+ delisted securities to eliminate this bias.
An AI model's tendency to agree with or validate what the user says, even when the user is wrong. More capable models tend to be more sycophantic because their training (RLHF) rewards user satisfaction, and agreement satisfies users. In our experiments, the largest models agreed most — the smallest pushed back most.
Limiting to 1 pick per correlated group (e.g., max 1 crypto stock) to avoid concentrated risk.
A stock's short symbol (e.g., AAPL for Apple). Used to identify securities on exchanges.
An exit execution strategy that first attempts to close positions using liquidity-aware limit orders at favorable prices, then falls back to market orders as a safety net if limits are not filled before market close. This minimizes slippage on exit compared to closing entirely with market orders.
The average price weighted by volume throughout the day. Institutional benchmark — above VWAP is bullish.