| ATR | Average true range: how far price usually travels in one bar, gaps included. |
| AUC | Chance a random "up" case scores higher than a random "down" case; 0.5 is no skill. |
| Balanced accuracy | Average of the up-call and down-call hit rates; drift cannot inflate it. |
| Base rate | How often "always the majority side" is right; the bar any forecast must clear. |
| Black-Scholes | The textbook formula for an option's fair price, given volatility. |
| Bollinger Bands (BB) | Average ± 2 standard deviations of closes. |
| Calibrated probability | Share of past readings like this one that actually went up. |
| Cholesky | A fast, exact way to solve the small linear systems behind ridge regression. |
| Classic SD / v1 | The paper's indicator: 21-bar compression, AO momentum, yes/no active window. |
| Compression tier | 3 high, 2 mid, 1 low, 0 none: how deep inside the Keltner Channels the bands sit. |
| Condor (iron) | Sell a call and a put out of the money, buy further ones as protection; pays if price stays in range. |
Depth q | 2σ/ATR; the continuous squeeze measure behind the tiers. |
| Drift | Stocks' tendency to rise over time; makes "always up" look smart. |
| Effective entry | What a limit plan really pays on average: the limit if filled, the window's last close if not. |
| Energy | Squeeze depth summed over the squeeze's life; how hard the spring is wound. |
| Era | One of the three date ranges (Train before 2014, Valid 2014–2018, Test 2019+) that never mix. |
| Expected move | SD v2's forecast of the size of the next 21-bar move, in percent. |
| FeatureFrame | All of SD's measurements for one symbol, one column per measurement, one row per bar. |
| Fill probability | Share of history (in this SD bucket) whose price traded through the limit within the window. |
| Fire | The moment a squeeze's tier drops below its highest tier. |
| First touch | Which of ±1.5 ATR price hits first; "which way did it break?". |
| Full period | A month or year traded from flat to flat: decide at each close, fill next open (stocks) or that close (options), mark daily, close everything at the end. |
| Group A / B / C | v1's definitive call / exhaustion call / squeeze lifecycle symbols. |
| Implied volatility (IV) | The movement that option prices imply; the market's own forecast. |
| Intensity | Where today's expected move ranks, 0 to 1. |
| IRLS | Iteratively reweighted least squares: how the logistic model is fitted. |
| Keltner Channels (KC) | Average ± k × ATR. |
| Lean | Blended momentum × time-decay strength; the side of 0.5 SD sits on. |
| Lift | Hit rate minus base rate; the part a forecast can take credit for. |
| Long-term P(up) | Chance of a higher close in 126 or 252 bars; walk-forward, it equals the market's drift. |
| Look-ahead | Using information from after the decision date; the bug every rule in the backtester exists to prevent. |
| Market weather | SPY's trend, momentum, volatility ratio and recent returns; sets SD's side. |
| Phase | τ/θ: how far through its expected life a fire is. |
| Profit factor | Gross profit ÷ gross loss over closed trades; above 1 makes money. |
| Proxy IV | IV estimated from realized volatility when no real index exists; blind to earnings. |
| Purge | Dropping samples whose labels would cross from one era into the next. |
| Real-price scale | Real price ÷ adjusted price on a date; lets the backtest buy whole contracts at prices that were actually printed. |
| Ridge | Least-squares regression with a penalty that keeps weights small. |
| Sample | One (symbol, day): what was known at the close, and what happened afterwards. |
| Scorecard | The row of hit / base / lift / balanced / ρ numbers that grades a set of calls. |
| SD bucket | Squeeze state (coiled, fired, quiet) × intensity decile; 30 in all, used by the timing tables. |
| Seed | The number that fixes a random stream, so a noisy run can be reproduced exactly. |
| Sentiment | Analyst consensus, revisions, target upside and headline tone, as one number. |
| Sharpe / Sortino | Annualized mean daily return ÷ its standard deviation (Sortino: ÷ downside deviation only). |
| Side | Which side of 0.5 SD sits on: the sign of P(up) − 0.5 from drift plus market weather. |
| Squeeze | Bollinger Bands inside Keltner Channels: unusually quiet price. |
| Straddle | Buy a call and a put at today's price; pays on a big move either way. |
| Time decay | v2's strength curve: full for 1.5θ, then halving every θ. |
| Variance risk premium | Options usually charge for more movement than is delivered. |
| Vol edge | SD's expected move ÷ the implied move; ≥ 1.3 in a coil is the straddle setup. |
| Volatility guard | min(σ₂₀, σ_robust): noisy prints cannot inflate SD's expected move. |
| Walk-forward | Fit on the past, grade on the next slice, slide forward, repeat. |
| Walk-forward choice | Picking each year's model by its record over the previous years only, then grading it on that year. |
| z_min / z_max | Lowest low / highest high over the window, in σ·√N units of normal movement. |
| θ (theta) | Average squeeze length over the last 125 bars; the fire's expected life. |
| τ (tau) | Bars since the squeeze fired. |