Transparency

Data Sources

Where every number on this site comes from, what it powers, and how often it updates.

Every dashboard is built from pre-computed data so pages load fast. The table below lists each underlying source, how current it is, and any limitations to keep in mind. Sources marked Planned are not live yet.

COT — Commitment of Traders

Live

Source: U.S. Commodity Futures Trading Commission (CFTC) — official Legacy, Disaggregated, and Traders-in-Financial-Futures reports.
Access: Public, free, no account required.
Powers: The Market Dashboard, COT Shortcut, Market Detail, and the GOLD card's COT sub-tab — every COT positioning indicator (COT Index, OI Stochastic, WILLCO, COT Movement, Composite COT, Spec/Comm ratios).
Update cadence: The CFTC publishes once a week (Friday, covering the prior Tuesday's positions). We refresh from the CFTC archive and rebuild the dashboards on that schedule.
Notes & limitations: A ~3-day reporting lag between the position date and publication is inherent to the CFTC report itself — it is not a delay in this app.

Gold Fear & Greed (Sentiment)

Live

Source: OnOff.Markets public Gold Fear & Greed Index (the 0–100 score plus its component sub-signals).
Access: Public, free, no account required.
Powers: The GOLD card's Sentiment sub-tab — the gauge, the full history chart, and the component breakdown.
Update cadence: Snapshotted automatically every day and archived. OnOff only publishes a rolling 1-year window; because we keep each daily value, our history keeps growing past that one year over time.
Notes & limitations: Covers gold only. The value reflects OnOff's own methodology; we redraw and archive it, we do not recompute it.

Options flow (GLD Put/Call)

Planned

Source: To be selected — the GLD volume Put/Call ratio and Put/Call open-interest ratio (e.g. Barchart's GLD Put/Call Ratio, OptionCharts, or Interactive Brokers).
Access: Pending — most options-data providers require a paid subscription or a brokerage account.
Powers: Will power the GOLD card's Options-flow sub-tab, ideally with OptionCharts-style gauges.
Update cadence: To be defined once a source is chosen.
Notes & limitations: Not yet live. Waiting on a decision about which data provider to use (see the provider comparison below) and on the corresponding access/terms. The sub-tab currently shows a placeholder.

Gold price history (cycle analysis)

Live

Source: Yahoo Finance — continuous front-month COMEX gold futures (GC=F), full available daily close history (from 2000).
Access: Public, free, no account required.
Powers: The GOLD card's Cycle sub-tab — the FFT periodogram, the MESA (maximum-entropy) spectrum, and the Bartels significance test on the daily, weekly and monthly series.
Update cadence: Rebuilt automatically whenever the frontend data is rebuilt, i.e. on the weekly COT refresh.
Notes & limitations: The same instrument as the GOLD COT market (CFTC 088691), so the two sub-tabs describe one market. Weekly and monthly series are resampled from these daily closes rather than fetched separately, so the three timeframes cannot disagree — but they are also not three independent samples. Being a continuous front-month series spliced with no roll adjustment (see the Strategy price history row), it carries contract-roll discontinuities.

Strategy price history

Live

Source: Yahoo Finance (daily/monthly price series for the futures and ETFs used in the strategy back-tests, e.g. Corn, Gold ETF, Silver ETF, S&P 500).
Access: Public, free, no account required.
Powers: The ETF, index and FX-spot strategy cards (Monthly MA Crossover, GLD/SLV strategies, 0DTE Moves, Tesla breakout, the Kaufman studies, S&P 500 mean reversion) and four of the twenty-six market cards: the three roll-free tickers — Bitcoin BTC-USD, US Dollar Index DX-Y.NYB, Lithium & Battery ETF LIT — plus Aluminium ALI=F, which neither futures feed covers. The other futures cards moved to the QuantConnect feed on 2026-08-20 and to the Norgate feed on 2026-09-24 (client decisions); Corn KAMA stays on this feed.
Update cadence: Refreshed when a strategy's back-test is rebuilt; results are pre-computed and cached. These payloads are NOT on the weekly refresh — they are rebuilt by hand.
Notes & limitations: The continuous futures tickers (=F) on this feed are Yahoo's front-month chain, spliced with no roll adjustment — not back-adjusted (Panama) and not ratio-adjusted (verified empirically 2026-08-13 by matching the chain against individual contracts) — which is exactly why the futures cards moved to adjusted feeds. The one futures card still on this feed, Aluminium, therefore prints a phantom jump at every contract roll: absolute performance numbers carry that roll noise, comparative readings (rule vs rule, period vs period) run on the same seams on all sides, and the card is labelled. Separately, Yahoo Finance is a free, unofficial feed and can occasionally have gaps or revisions; back-test outputs are recomputed on each rebuild. COMEX aluminium is largely stale before 2020 — 95.6% of 2018's closes are unchanged — so that card starts in 2020. Lithium has no price series on this feed, so its card tracks the LIT equity ETF and is labelled as not being the lithium price.

Strategy price history — QuantConnect ratio-adjusted futures Since 2026-10-06 the Aluminium card (ALI=F) is net of futures trading costs ($5 + 1.5 ticks per round turn); Bitcoin, the US Dollar Index and LIT are not futures and carry none.

Live

Source: QuantConnect Datasets — AlgoSeek US Futures (the 157 most-liquid contracts; daily bars measured to start 2008-01-02 for most markets, better than the documented May 2009), served as continuous contracts with backwards-ratio roll adjustment (DataNormalizationMode.BackwardsRatio; open-interest roll rule, calendar rolls for Copper and the Brazilian Real).
Access: Requires a QuantConnect account (free tier suffices: the series are exported through their research environment, a ~10-minute manual routine per refresh; see market_data/README.md in the repository. Automated refreshes would need their paid API tier).
Powers: Since 2026-09-24 one of the twenty-six market cards — the Brazilian Real 6L=F, which Norgate does not cover. From 2026-08-20 to 2026-09-24 this feed powered the Corn MA and Corn Donchian cards, sixteen market cards and the CTA tab; all of them moved to the Norgate feed below (client decision: the Panama method).
Update cadence: Historical record of the 2026-08-20 decision; since 2026-09-24 only the Brazilian Real card uses this feed, rebuilt with --price-source quantconnect. The same hand-rebuilt cadence as every strategy payload: refresh the CSVs per market_data/README.md, then rebuild with `python -m strategies.corn_ma_optimization.build_frontend_data --price-source quantconnect --as-of <settled date>`, `python -m strategies.corn_donchian_breakout.build_frontend_data --price-source quantconnect`, `python -m strategies.multi_market_ma.build_frontend_data --price-source quantconnect` and `python -m strategies.cta_trend_following.build_frontend_data --price-source quantconnect`. Not on any automatic schedule. The 2026-08-20/21 switch rebuilt every section of every card in one run — including the Reset / No Reset re-entry comparison (section 13) and Corn's dated trade log (section 14) — so, unlike after the 2026-08-04 section-13 addition, no section currently sits on a later snapshot than the sections beside it; if a future partial rebuild reintroduces such a seam, each card's Reset / No Reset section reports the drift by comparing its own Immediate column against section 7's shipped table.
Notes & limitations: Historical record of the 2026-08-20 decision; since 2026-09-24 only the Brazilian Real card uses this feed, rebuilt with --price-source quantconnect. Backwards-ratio adjustment preserves percentage returns across every contract roll — the right convention for this site's compounding percentage equity curves. Two consequences travel with it: point/dollar values in older history are scaled units rather than the nominal prices of their day, and the whole back history legitimately rescales whenever a new roll occurs. Measured and decided on the exports of 2026-08-20 (per-market data-quality verification, including a fill-forward audit and a roll-mapping comparison): 17 of the 24 futures switch to QuantConnect — 12 at full history from 2008-01-02 (the full 2008 crash included; Corn, Crude Oil, Natural Gas, Gold, Wheat, the three Treasuries, Lean Hogs, Live Cattle, Copper on calendar rolls, and Soybeans, whose 2014-15 vendor data hole is disclosed on its card) and 5 with a trimmed start where the early data is gap-ridden (Silver from 2013, British Pound and Japanese Yen from 2017, Mexican Peso from 2019, Brazilian Real from 2016 on calendar rolls). Seven cards stay on the Yahoo feed with its disclosed seams: Palladium (QuantConnect's palladium daily data is chronically sparse) plus the five ICE softs and Aluminium (no QuantConnect data). Versus Yahoo, roughly eight years of history (2000–2008) fall away and the CTA in-sample tab (window to 2015-12-31) keeps 8 years of data; the GOLD card's long-cycle analysis keeps its longer Yahoo series (QC gold starts 2012-11). Since 2026-10-06 its results are net of trading costs too; with no real price in a ratio-adjusted series the cost is charged against the adjusted price, which sits below the real one in early years (about 0.56x in 2016 on the Brazilian Real), so early trades pay up to ~1.8x their real cost share — disclosed on the card.

Strategy price history — Norgate Panama-adjusted futures

Live

Source: Norgate Data futures package (norgatedata.com): about 100 futures markets (112 continuous series including micro and session variants) across 11 exchange groups (CME Group, ICE, Eurex, Euronext, ASX, SGX and others), end-of-day OHLC with volume and open interest, individual contracts plus continuous contracts. History starts at the market's first trading day or around 1978-1983 for the long-established US contracts (measured on the export: corn and gold 1978, crude oil 1983, 10-Year T-Note 1982), 1997 for the E-mini S&P and 1999 for the Eurex markets. The continuous series used here is back-adjusted by the difference (Panama) method — verified on the export itself: the gap to the real price is constant within every contract and steps only at a roll. Norgate rolls one business day before last trade or first notice, whichever comes first — later than most trend followers roll, and later than the QuantConnect open-interest rule it replaces.
Access: Paid personal subscription (Futures package, about USD 270 a year). Norgate's software runs only on Windows, so the series are exported from a Windows virtual machine on the subscriber's own computer. The licence is personal use only and forbids redistribution, so the price files are never committed to this repository or served by the site — only back-test results computed from them are published.
Powers: Live since 2026-09-24, replacing the QuantConnect feed (client decision, 2026-09: the Panama method with absolute values): the Corn MA and Corn Donchian cards (Corn KAMA stays on Yahoo) and twenty-one of the twenty-six market cards — Crude Oil CL=F, Palladium PA=F, Live Cattle LE=F, Cocoa CC=F, Coffee KC=F, 2-Year T-Note ZT=F, Silver SI=F, Gold GC=F, 10-Year T-Note ZN=F, 30-Year T-Bond ZB=F, Wheat ZW=F, Soybeans ZS=F, Natural Gas NG=F, Lean Hogs HE=F, High Grade Copper HG=F, Sugar No. 11 SB=F, Orange Juice OJ=F, Cotton CT=F, British Pound 6B=F, Mexican Peso 6M=F, Japanese Yen 6J=F (tickers are the cards' configured identifiers; the underlying series are Norgate's continuous contracts) — including the six the QuantConnect feed could not serve (Palladium and the five ICE softs), plus the CTA tab's in-sample copies. Aluminium and the Brazilian Real have no Norgate market and keep their current feed.
Update cadence: Hand-rebuilt, like every strategy payload: update Norgate in the virtual machine, run the export, then rebuild with --price-source norgate and NORGATE_CONTINUOUS_ROOT pointing at the export. Not on any automatic schedule — Norgate cannot run in the cloud.
Notes & limitations: A Panama series keeps point differences real but not price levels: over a long history the back-adjusted price reaches zero or below on 27 of Norgate's 112 markets (heating oil on 91% of its days). So signals, ATR, stops, targets and profit in points are computed on the adjusted series, while every percentage — trade returns, the compounding equity curve, annualized volatility, buy-and-hold — is taken over the real price of the contract held that day (on a roll day, the incoming contract's). A trade is a fixed stake at entry in the held contract's real value, carried across rolls — the same convention the ratio-adjusted feed gave, so a long held throughout equals buy-and-hold. Bars are no longer dropped for a negative adjusted price; only a non-positive real price is, and the one such bar (crude oil, 2020-04-20, the expiring May contract at -37.63) is disclosed on its card. Two things move the numbers against the previous feed and should be read before comparing: the roll rule (measured over each card's previous QuantConnect window — 2008-2026, later for gold, silver and the currencies — buy-and-hold differs by up to 9 percentage points a year on lean hogs and 6 on natural gas, by 1-2 on copper, crude oil, corn, live cattle and soybeans, by about 1 on wheat and 0.7 on the Mexican peso, and by under half a point on gold, silver, the Treasuries, the pound and the yen), and the history length — most cards now start 1978-1983 instead of 2008, and the 1980s-90s trends lift trend-following results considerably, so a full-history figure is not comparable to a figure from the old 2008+ window. Trading costs (client 2026-10-06): every futures result on the strategy cards is net of $5 commission + 1.5 ticks of slippage per round turn per contract, at today's exchange tick sizes (the 2-year note's and the yen's ticks have halved since the 1980s, so their early slippage is slightly understated); a position held across a contract roll is not charged an extra round turn. Spot, index and ETF series carry no cost.

Strategy price history — CTA Trend Following (in-sample, to 2015-12-31)

Live

Source: The same feeds as the market cards — the Norgate Panama-adjusted continuous futures, with Yahoo for the roll-free tickers — truncated at 31 December 2015. No separate feed: one fetch per market, cut at that date before any indicator or back-test reads it.
Access: See the two rows above.
Powers: The CTA Trend Following tab — the concentration analysis, the funded ensembles, the cross-market book, and the same twelve sections as the market cards rebuilt over the shorter window. Twenty-four markets since the 2026-09-24 switch to Norgate: COMEX aluminium is absent (its usable history begins in 2020), and so is the Brazilian Real (its QuantConnect history starts 2016). The British Pound, Japanese Yen and Mexican Peso are back, because Norgate's history reaches 1979 (pound, yen) and 1995 (peso).
Update cadence: Static. This is deliberately NOT on the weekly refresh — the as-of date is the point of the tab, so its payloads are rebuilt only when the method, the feed or the market list changes, never on a schedule. Rebuild with `NORGATE_CONTINUOUS_ROOT=<export dir> python -m strategies.cta_trend_following.build_frontend_data --price-source norgate`.
Notes & limitations: Every data caveat from the rows above still applies to these cards, and the window adds its own. On Norgate most futures now have an in-sample window starting 1978-1983 (the 2-Year T-Note and natural gas 1990, the peso 1995), where the QuantConnect era gave them 2008-2015 — three to four decades including the 1980s trends, so these results are not comparable to the tab's previous figures. Bitcoin has roughly fifteen months of history by the cut-off (its feed starts September 2014), below the 750-bar floor this tab uses, so it gets a page but is never funded. The US Dollar Index reaches back to 1971 on its Yahoo cash-index series and Lithium only to 2010, so their windows differ from the futures around them. The data after 2015-12-31 exists and is untouched — it is the out-of-sample test for whatever comes out of this work, and no calculation on this tab reads it.

Orange Juice futures

Live

Source: Yahoo Finance, with Nasdaq Data Link (ICE Orange Juice futures) as a cross-check/fallback.
Access: Public. The Nasdaq Data Link cross-check can use an optional free API key for higher rate limits.
Powers: The Orange Juice seasonality strategy card.
Update cadence: Refreshed when the back-test is rebuilt.
Notes & limitations: Two independent sources are used so the price history can be validated.

Saxo OTC FX options

Live

Source: Saxo Bank's public FX Options web tool (at-the-money volatility, risk reversals, pin risk).
Access: Public web data, no account required.
Powers: The Saxo OTC dashboard.
Update cadence: Snapshotted on demand; the live Saxo figures move intraday.
Notes & limitations: Coverage is the major currency pairs across standard tenors (1W–1Y).

Options flow — choosing a data provider

Decision needed

The Options-flow sub-tab (GLD volume Put/Call ratio, Put/Call open-interest ratio, and OptionCharts-style gauges) needs a paid options-data source — there is no free or public feed for it. There are three realistic options, trading off cost, how closely they match the requested gauges, and how cleanly we can automate the daily update.

Provider	What you get	Needs	Monthly	≈ Weekly	≈ Daily	Automation
Interactive Brokers (IBKR)Suggested	The full GLD option chain (volume + open interest per strike). We compute the Put/Call volume and open-interest ratios ourselves.	An IBKR account + the OPRA US-options market-data subscription, accessed via IBKR's API.	≈ $0–10	≈ $0–2.50	≈ $0–0.35	Full API — clean automated daily collection.
OptionCharts.io	The same GLD option-overview gauges the client referenced, plus Put/Call ratios. Premium is 15-minute delayed with CSV export; Ultimate adds real-time data.	A paid OptionCharts subscription (Premium or Ultimate).	≈ $20–30 (Premium)	≈ $5–7	≈ $0.70–1.00	No official API — CSV export (semi-manual) or scraping (against their terms).
Barchart OnDemand	The exact 'GLD Put/Call Ratio' the client cited, including historical Put/Call ratios via a proper REST API (back-history available).	A paid Barchart OnDemand API plan (usage-based / custom).	from ≈ $500	≈ $115	≈ $16	Proper REST API — cleanest automation, plus historical back-fill.

Prices are approximate, as of June 2026 — confirm with each provider. Every provider bills monthly; the weekly/daily figures are just the monthly cost spread out for comparison. Sources: IBKR market data, OptionCharts, Barchart OnDemand.

Why each one

Interactive Brokers (IBKR). Cheapest and most legitimate if you already trade at IBKR: the OPRA fee is small and is often fully waived once monthly trading commissions pass a low threshold. We own the calculation, and — like the sentiment archive — history accumulates from the day we start collecting. Trade-off: it's the most build effort (an IBKR data connection), and you don't get back-history, only forward.

OptionCharts.io. Cheapest way to match the exact gauges in the client's screenshot, with low cost and a 15-minute-delayed feed that's fine for a daily ratio. The catch is there's no clean API: automated ingestion means a CSV-export routine rather than a real feed, so it's more manual to keep fresh.

Barchart OnDemand. The most 'correct' source for the precise metric the client named, and the only one that hands us historical Put/Call ratios out of the box (no waiting to accumulate). But at roughly $500/month it's overkill unless options flow becomes a core, paid part of the product.

Suggestion: if you already trade at Interactive Brokers, start there — it's the cheapest and most legitimate, and we build the ratios ourselves (history accrues from when we start). If not, and you mainly want those gauges at low cost, OptionCharts Premium is the pragmatic choice. Reserve Barchart for when options flow needs to be a core, historical data product.

Data is refreshed by re-running the build pipelines and publishing the updated site. Pricing and positioning figures are for research and information only and are not investment advice.