Make the assumptions
visible.
The scanner is a research tool built from daily observations, fixed rules and historical comparisons. Here is how to interpret it.
01 / Measuring a price extreme
The scanner fits an ordinary least-squares trend to log prices over 150 prior daily observations, also requiring at least six calendar months. The current close is excluded from the fit. Residuals are centered on their median.
The current deviation from the extrapolated trend is divided by residual standard deviation. An absolute z-score of 2 is “extended”; 3 is “extreme.” A negative value is below trend; a positive value is above. These labels are descriptive, not Gaussian tail probabilities.
The robust score uses median absolute deviation. “Turning” requires both a shrinking absolute z-score and a close moving toward the center. A one-day turn is not an established trend reversal.
02 / Comparing historical events
Analogs use the same instrument, extension direction, severity bucket (2–3σ or ≥3σ) and alignment with the fitted trend. Event starts are separated by at least 21 observations. Only completed outcomes before the latest observation are eligible.
The simulated trade enters at the next open, targets the frozen median, stops at two ATR from entry, or exits after 20 sessions. If the next open has already crossed the target, it skips entry. Adverse opening gaps fill at the open. Daily bars hitting both barriers are resolved stop-first.
Target-hit frequency includes a 95% Wilson interval. The median time to target describes successful hits only. Expected value is the sample mean net strategy return under the displayed cost assumption. Fixed 5-, 10- and 20-session returns use separate holding rules without target or stop barriers.
03 / Reading daily variability
ATR(14) uses Wilder smoothing to summarize daily trading range. Daily return sigma is the standard deviation of 150 log returns. Neither is a loss ceiling. A fitted residual half-life is a descriptive diagnostic, not a reliable target arrival date.
04 / Position scenarios in NOK
The default scenario uses NOK 50,000 of total market exposure per position and at most six positions. This is not NOK 50,000 of margin. Scenarios assume 28 calendar days and unchanged conversion FX. Cash positions are long-only.
Commission, conversion, custody, spread, slippage and financing assumptions affect the reported net outcome. Cash historical returns translate principal and P/L; derivative historical returns translate P/L. Actual Saxo account terms, tradable instruments, minimum lots and contract rolls still require verification.
“All targets” and “all stops” are hypothetical joint outcomes, not expected returns. Six similar positions can move together. Stops can be exceeded by gaps, and the NOK volatility illustration excludes currency translation.
05 / Sources and observation dates
Market histories come from Yahoo Finance via yfinance, or supplied CSVs. Stock discovery uses ApeWisdom and Stocktwits, with Finviz context. Risk inputs are separate daily histories and Finviz breadth counts. Missing inputs cannot establish low risk.
Rows dated today in UTC are excluded. The scanner currently uses a four-calendar-day price-age threshold and a 36-hour breadth threshold; these are approximate rather than full exchange calendars. Fetch time, build time and the date of a market observation are different things.
06 / What remains unvalidated
No held-out or prospective evidence establishes the predictive accuracy of these thresholds or analogs. Selection effects, correlated assets, regime changes, continuous-futures rolls and revised provider histories limit interpretation. Past results are not evidence of future profitability.
The implementation audit identified issues still requiring correction: display-time freshness, cached stock eligibility, aligned credit dates, historical exit-time FX conversion, and acceptance of large inconsistent OHLC ranges. Nondefault generic cost settings can also disagree with stock analog labels. A visual redesign does not fix these calculations.
The risk model is a US equity environment filter, not a covariance-based portfolio model. Its output is not a probability of loss. Historical simulations use research proxies and incomplete execution assumptions.
Public redistribution rights for the research feeds have not been confirmed. Data availability and accuracy can change. Report suspicious observations to a@turquet.trade.
Open the research scanner ↗