Every chart in Turning Tide traces back to a public regulatory or exchange source — or, where we apply licensed third-party data and internal models, we say so. This page is the audit trail. We name primary regulatory sources verbatim (SEC EDGAR, the Federal Reserve, the U.S. Treasury, FINRA). Where we use commercial data partners under license, we describe them by category — what they cover and how fresh they are — rather than by vendor brand, because the underlying primary source is what carries the audit weight.
The table below pairs each major feature with its source category, freshness, coverage, and delivery class. Freshness is the elapsed time between the underlying event (a filing, a trade, a release) and the moment that record becomes available in Turning Tide.
We label every feed honestly. Paid tiers carry near-real-time quotes in place of the Free tier's standard 15-minute delay (required for unentitled retail data access). Full tick-level streaming — the full-fidelity exchange feed on Pro and Elite — is on the near-term roadmap and is labeled as such until it ships.
Intraday means the feed updates throughout the trading day on a defined cadence — typically 15-minute snapshots for options chains, or near-real-time for off-exchange print tape (subject to FINRA's standard reporting delay).
End-of-day means the feed refreshes once per session — fundamentals, short interest, valuation models, and most macro series fall here.
Modeled means the value is computed from upstream inputs rather than read directly off a primary feed. DCF valuations, composite scores, and AI summaries are all modeled — and we label them as such everywhere they appear.
The backtester (Pro and Elite tiers) is built to avoid the most common ways backtests lie:
No lookahead bias. Every signal at time t is computed only from data available at time t. Earnings revisions, analyst targets, and any other point-in-time inputs use the version of the value that was known on that date — not the restated number visible today.
No survivorship distortion. The universe includes delisted tickers. A strategy backtest run against today's S&P 500 alone would silently exclude every company that went to zero — the result is fake. We keep the dead names in.
In-sample / out-of-sample split + walk-forward.Strategies are tuned on an earlier slice of history and tested on a later, untouched slice. Walk-forward iterates this across rolling windows to confirm the parameters aren't fit to a single regime.
Monte Carlo paths. We resample trade sequences across 1,000 paths to estimate the distribution of returns and drawdowns — not just the single realized backtest curve.
Transaction costs are modeled. Default assumptions (commission per share or per contract, bid-ask half-spread, slippage on aggressive entries) are visible in the backtester settings panel. They are not zero.
Backtested performance reflects historical conditions and an internally-modeled trade execution. Future returns are unknown. We say this loudly because it matters.
AI research briefs are generated using our AI research engine — a frontier large-language model. The briefs synthesize public filings, recent flow signals, and macro context into a structured summary: what changed, the bull case, the bear case, signal conflicts, and what to research next. Each brief cites the source documents it processed.
The AI does not predict prices. It does not recommend buys, sells, or position sizes. It does not see your watchlist, your portfolio, your account, or your trades. We do not train models on user data.
Failure modes are real. Large language models can hallucinate — invent a quote that isn't in the filing, misattribute a transaction to the wrong insider, or summarize a guidance change incorrectly. Every brief shows the source documents it processed; treat the brief as a research starting point and verify load-bearing claims against the linked source.
Knowledge cutoff. The model carries a training cutoff date. For events past that cutoff, we pass the AI fresh source text from the live feeds — the model is summarizing what we hand it, not recalling it from training.
Turning Tide is not a broker-dealer. We do not hold customer assets, route orders, or execute trades. Connect your existing broker; Turning Tide is the research layer above it.
Turning Tide is not a registered investment adviser. Nothing in the product or on this site is investment advice or a recommendation to buy or sell any security. Signals are research inputs; the interpretation, the decision, and the risk are yours.
We don't run a signal service, a chat room, or a paid trade-call feed. We don't accept payment to feature tickers. We don't have affiliate relationships with brokers that influence what we show you.
Not investment advice. All content on Turning Tide is research and education, not a recommendation to buy or sell any security or to follow any particular investment strategy. Past performance is not indicative of future results.
Backtests are not predictions. Historical or simulated performance does not predict future returns. Live trading conditions — including slippage, fills, borrow availability, and tax — diverge from idealized backtests.
Options carry substantial risk. Options can expire worthless. Multi-leg, short, and complex strategies can lose more than the initial premium. Read the OCC's Characteristics and Risks of Standardized Options before trading options.
Short selling carries unlimited loss potential.Borrow availability, recall risk, and forced buy-ins are real execution risks not modeled in any backtest.
Off-exchange print ambiguity. Large block prints reported through FINRA's TRF do not by themselves identify direction (buy vs sell), aggressor, or actor (institutional vs retail-aggregated). Off-exchange flow is a research input that needs price and volume confirmation.
AI summaries can be wrong. See the AI section above for failure modes. Verify load-bearing claims against the linked source documents.
Report data errors or methodology questions to data@turningtide.io. We treat data quality as a first-class concern; corrections route directly to engineering, not customer support triage.