Data Sources
Every number on the dashboard comes from a named source with a named cadence. Nothing here is live: prices are end-of-day, option quotes are a snapshot taken at the refresh, and anything delayed is labelled as delayed. The dashboard rebuilds on weekdays at 16:45 ET; the Markets, Sentiment and Macro panels also show how old their own snapshot is.
1. Where each number comes from
| Source | What it feeds | Cadence | Delay / notes |
|---|---|---|---|
| Yahoo Finance (yfinance) | Daily closes for the 6 watched tickers; Bollinger Bands, 200-day SMA, regime, breadth, the Markets trend score | Weekday 16:45 ET refresh | End-of-day closes. Intraday quotes on the site are not live — the only live price surface is the TradingView chart widget, delayed ~15 min for most users. |
| CBOE daily statistics | VIX, equity put/call ratio | Weekday 16:45 ET refresh | Published after the close. |
| CBOE delayed quotes | Option chains, put/call walls, GEX snapshot | Weekday 16:45 ET refresh | ~15-min delayed, not live. The dashboard never claims live option quotes. |
| CFTC — Commitments of Traders | Non-commercial net futures positioning: S&P 500 (E-mini), Nasdaq-100, gold, silver, crude oil, euro, yen | Weekly, released Friday afternoon | The dashboard shows the latest published report, never a forecast of the next one. |
| FRED | Macro series: CPI, unemployment, 10-year and 2-year yields, T-bill rates, the 10Y–2Y spread | Per series, on its official release schedule | The dashboard shows the latest published value. The macro panel's "vs prior" is the change against the previous published value, not against a consensus forecast — no genuine free consensus-forecast source exists. |
| AAII sentiment survey | Weekly bull / neutral / bear percentages and the bull-minus-bear spread | Weekly, Thursdays | The survey week is printed on the card ("Week of Nov 11"). A stale week is shown as-is and never relabelled as fresh. Full AAII history is a paid product and is not used; the site accumulates its own history from Nov 2026 onward. |
| MarketWatch RSS | Headlines panel | Accumulated across the week | Matching stories from the week, de-duplicated; nothing is generated or rewritten. |
| StockTwits | Bullish share of messages, SPY and QQQ only | Per refresh | Small sample — sentiment colour only, never a trade signal on its own. |
| TradingView | Live chart widget | On demand | Delayed ~15 min for most users; the widget is the only live price surface on the site. |
| Screener (yfinance fundamentals) | Scanner rows: price, change, RSI, ATR, IV, yield, buffer | Per refresh | Fundamentals are daily, not intraday. |
2. The Market Radar — how each factor is scored
The Radar table on the Wheel page scores every watched ticker and every cross-asset market on the same −2…+2 scale. Blue is bullish, red is bearish, grey is no data. A cell is never left blank: grey "—" means the factor does not apply to that row (for example, a band distance on a Markets row) or the snapshot is missing (for example, the GEX column reads "—" when the gamma snapshot is older than 3 days — the evening GEX scan is a manual run, so a stale file is shown as missing, never carried as fresh).
| Factor | Input | +2 | +1 | 0 | −1 | −2 |
|---|---|---|---|---|---|---|
| Band | Distance to the lower Bollinger Band (20, 2), % | ≤ 1.5 | ≤ 5 | ≤ 10 | ≤ 20 | > 20 |
| Regime | Close vs the 200-day SMA | Above: +2 | — | Below: −2 | ||
| Trend | Markets trend score, 0–100 | ≥ 70 | ≥ 55 | ≥ 45 | ≥ 30 | < 30 |
| VIX | VIX level | < 15 | < 20 | < 25 | < 30 | ≥ 30 |
| Put/call | Equity put/call ratio | < 0.70 | < 0.85 | < 1.00 | < 1.15 | ≥ 1.15 |
| IV rank | IV rank over the recorded history in iv_history.json, % | < 30 | 30–<45 | 45–<55 | 55–<70 | ≥ 70 |
| Walls | Nearest put wall below spot, % under the price | ≤ 1.5 | ≤ 5 | ≤ 10 | ≤ 20 | > 20 |
| Flow | Unusual-flow dominant side by premium (45–75 DTE prints) | — | Calls dominate | No prints / tie | Puts dominate | — |
| AAII | AAII bull-minus-bear spread | ≥ +15 | ≥ +5 | > −5 | > −15 | ≤ −15 |
| COT | COT percentile rank (52-week) | ≥ 80 | ≥ 60 | ≥ 40 | ≥ 20 | < 20 |
| StockTwits | Bullish share of messages, % | ≥ 60 | ≥ 45 | ≥ 35 | ≥ 25 | < 25 |
| Breadth | % of S&P 500 names above their 50-day average | ≥ 70 | ≥ 60 | ≥ 50 | ≥ 40 | < 40 |
| Regime verdict | Weekly regime verdict tone (regime.py) | Good tone: +2 | Caution tone: 0 | Bad tone: −2 | ||
| 10Y–2Y | 10-year minus 2-year Treasury yield, pp (FRED) | > +0.50 | 0 to +0.50 | — | −0.50 to 0 | ≤ −0.50 |
| HY OAS | High-yield option-adjusted spread, % (FRED) | < 3.5 | 3.5–5 | — | 5–6.5 | > 6.5 |
Bias pill
| Row | Bias | Rule |
|---|---|---|
| Watched ticker | Very Bullish / Bullish / Neutral / Bearish / Very Bearish | The Market Score verdict: ≥ +6 Very Bullish, +3…+5 Bullish, −2…+2 Neutral, −5…−3 Bearish, ≤ −6 Very Bearish. |
| Markets row | Very Bullish / Bullish / Neutral / Bearish / Very Bearish | The 0–100 trend score: ≥ 70 Very Bullish, 55–70 Bullish, 45–55 Neutral, 30–45 Bearish, < 30 Very Bearish. |
The Bias pill on a watched ticker is the Market Score verdict — the composite, not a single trend read. Ticker rows are ranked by Market Score, highest first; Markets rows sit below them in scorecard order and carry no Market Score (their Score cell is "—"). The "Market-read" line above the table is a fixed template filled with the day's numbers (nearest-to-band ticker, count of bullish regimes, VIX, put/call, AAII spread) — a sentence, not an analysis.
3. The Market Score — the published formula
Market Score = Technical + Volatility + Options + Sentiment + Macro, each factor −2…+2, so the score runs −10…+10. The weights are equal — every factor counts once. There is nothing to copy from the commercial scanners: their weights are unpublished. Ours are public by design, and the arithmetic is checkable by hand from the Radar cells — each factor is the clamped sum of its inputs below, using the same thresholds as §2. The score is computed once per refresh, server-side; the browser only renders it.
Verdict bands: ≥ +6 Very Bullish, +3…+5 Bullish, −2…+2 Neutral, −5…−3 Bearish, ≤ −6 Very Bearish. The Bias pill on a ticker row is this verdict.
| Factor | Inputs | How it is computed |
|---|---|---|
| Technical | Distance to lower band · Regime (vs 200-day) · Markets trend score | Each input −2…+2 per §2 (band, regime, trend); factor = sum clamped to −2…+2. The trend score comes from the Markets scorecard (SPY→S&P 500, QQQ→Nasdaq, IWM→Russell 2000, DIA→Dow); NFLX and BMNR have no trend row, so that input is "—" for them. |
| Volatility | VIX level · IV rank · Equity put/call | Each input −2…+2 per §2; factor = sum clamped to −2…+2. IV rank is (latest − low) / (high − low) over the recorded history in iv_history.json; high IV reads like high VIX (fear, negative). The file covers the screener universe, not the watched ETFs — today only NFLX has a rank, and its history is thin (6 prints), labelled as such. |
| Options | GEX regime · Wall distance · Unusual flow | GEX regime is the sign of net gamma exposure only: positive (dealers long gamma, range-friendly) +1, negative (dealers short gamma, hedging amplifies moves) −1, deliberately ±1 not ±2. The snapshot comes from the evening GEX scan (gex.json); it feeds the score only while it is ≤ 3 days old, otherwise the input is "—". Wall distance is the nearest put wall below spot, scored like the band (§2 walls); the nearest call wall is shown for context. Unusual flow is the dominant side by premium among the day's 45–75 DTE prints: calls +1, puts −1, none or a tie 0 — deliberately ±1, not ±2, because one day of prints is a weaker signal than a level. |
| Sentiment | AAII spread · COT percentile rank · StockTwits bull% | Each input −2…+2 per §2; factor = sum clamped to −2…+2. COT maps SPY→S&P 500 (E-mini), QQQ→Nasdaq-100 and IWM→Russell 2000 (E-mini) futures; DIA, NFLX and BMNR have no matching future, so that input is "—". StockTwits covers SPY and QQQ only. The Sentiment page also shows the COT Index (each market's net as % of its 3-year range) and the week-over-week net change — evidence, not scored inputs: the percentile rank already scores the net, and scoring both would double-count it. |
| Macro | Regime verdict · Breadth · 10Y–2Y spread · HY OAS | Each input −2…+2 per §2; factor = sum clamped to −2…+2. The regime verdict is the weekly verdict tone: good +2, caution 0, bad −2. The 10Y–2Y spread and HY OAS come from FRED. |
One score, many inputs. The Markets trend score (0–100: trend 50 + momentum 25 + RSI 25) is an input to the Technical factor — it is not a second score. The site has exactly one Market Score per ticker.
Missing inputs. An input with no data contributes nothing and is shown "—" in the scorecard evidence table. A factor with no inputs at all contributes 0. The score is a market-direction read, not a wheel-entry verdict: closeness to the lower band scores positive (the entry-zone logic the Radar already uses), so a ticker can read "Bullish" on the score while its Setup panel says WAIT — the two surfaces answer different questions.
History. The last 60 daily scores per ticker are kept in data.json (scores[t].history), accumulated one entry per refresh. The score-under-price chart on the Signal map tab plots them as a stepped line with dashed ±4 thresholds and flip triangles (▲ turned bullish, ▼ turned bearish, ◆ back to neutral); range pills filter the window (1M / 3M / 6M / 1Y / All) and the header states how many days of history exist, honestly — the first weeks say "collecting history — N days" instead of inventing confidence.
Hit-rate. For every day the score was ≥ +4 (a bullish signal) or ≤ −4 (a bearish signal), the 30-day forward return is checked: a bullish signal wins when the price rose, a bearish signal wins when it fell. The hit-rate is the share of resolved signal days that won, stored as scores[t].hit_rate with n (the count of resolved days), wins and horizon: "30d", and shown on each ticker's scorecard — always with n. Days whose 30-day window has not elapsed yet do not count. The history starts thin (n is 0 or small for the first weeks) and the scorecard says so — "not enough history yet" — instead of publishing a percentage from a handful of days. Like the Lab, the hit-rate measures the past; it does not say the next signal will win.
Risk gauge (separate from the score). The Risk On/Off gauge is its own read, not a sixth score factor: five components, each −2…+2 — SPY vs TLT and HYG vs LQD (the ratio's 1-month % change; stocks or credit beating bonds is risk-on), Gold (GLD, inverted — a rising safe haven is risk-off) and the Dollar (DXY, inverted — a rising dollar is risk-off), and the VIX level (<15 +2, <20 +1, >25 −1, >30 −2). The components sum to −10…+10: Risk On ≥ +4, Risk Off ≤ −4, otherwise Mixed. The compact gauge sits in the Radar market-read line; the full breakdown (all five components) is on the Macro page. A component that fails to load contributes nothing and is shown "—".
Events (earnings + ex-div). Each watched ticker's next earnings date and next ex-dividend date come from the yfinance calendar first, with the screener's earnings dates as the fallback for S&P names. ETFs have no earnings. The ETFs' next ex-dividend date has no free source (the Nasdaq dividends calendar lists common stocks only, and the per-symbol endpoint returns the last ex-div, not the next) — it reads "—" rather than a computed date. The scorecard flags earnings less than 21 days out ("do not sell into earnings") and shows the ex-div date when one is known.
4. Colour rules
- Blue / red is direction: blue = bullish or positive, red = bearish or negative. That covers the Radar cells, the Bias pills, price changes, regime labels, and the signal-map entry dots.
- Green / red is money only: profit and loss, premium, yield, IV. A green number is money made or premium earned; a red number is money lost or premium paid.
- Red appears in both vocabularies — the context disambiguates: a direction cell and a money cell never share a component.
- Green is also the interaction colour: hover borders and glows. Hover green is not a data colour.
5. What this site does not do
- No live quotes. Daily closes are end-of-day; option bids are a snapshot at run time and can be stale or move before any trade.
- No forecasts. Nothing here predicts a price. The Radar is a snapshot of where things stand, and the Lab measures how often a signal fired in the past — it does not say it will fire again.
- No recommendation, no positions. Nothing here says to buy, sell, or hold any security, and no actual trades are tracked or claimed.
- No dark-pool prints or signed options flow. Those feeds are paid; the site uses a free weekly proxy (FINRA ATS volume) and a volume-vs-open-interest flow proxy instead.
6. Scenario backtests — what happened after past events
The Strategy Lab's Scenarios section and the top-3 line on the Signals page show base rates, not forecasts: for five events, what followed on average over the next 7, 14 and 30 trading days, measured over 10 years of history. Every figure carries its n (the count of past occurrences); a small n is labelled thin, and an event with no history yet shows "not enough history yet" instead of a number.
| Event | Definition |
|---|---|
| Lower-band touch | Day's low touches the lower Bollinger Band (20, 2, population std) while the close is above the 200-day SMA. Every signal day counts. |
| Score flips to +4 | Market Score crosses upward through +4. From observed score history only — never backfilled, so n starts at 0 and grows one day at a time. |
| VIX above 25 | First day of each run with the VIX close above 25. Runs count once, so overlapping windows do not inflate n. |
| Put/call ≥ 0.85 | First day of each run with the equity put/call ratio at 0.85 or higher (CBOE daily options data). Runs count once. |
| GEX flips sign | Day net GEX changes sign vs the previous day. The GEX history file starts 2026-11-02, so n is thin at first. |
Inputs. Ten years of Yahoo daily closes (from 2016-01-01) for the six watched tickers and the VIX; CBOE daily options files for the put/call history (backfilled to ~1 year, stored up to ~3 years); the GEX history file; the Market Score history. Cadence: recomputed on Mondays from the full 10-year window, and the payload lands with the Monday refresh — other days carry the previous Monday's numbers forward. Where shown: the Lab's Scenarios section (one table per event, plus monthly seasonality) and the top-3 pooled line on Signals, which is the n-weighted mean of the per-ticker 30-day averages for the three events with the strongest follow-through.