SIGNAL9 logoSIGNAL9

Standards · Updated August 2026

Editorial policy & methodology

This page explains exactly how SIGNAL9 produces what you read: where the data comes from, how the AI pipeline turns it into a signal, how we grade our own results, how we handle mistakes, and how the site is funded. If something here is unclear or wrong, tell us at contact.

1. What SIGNAL9 is — and is not

SIGNAL9 is an independent research and education tool. It scans equities and cryptocurrencies, combines price data, technical structure, news flow and sentiment, and publishes a written thesis with an entry zone, a target and a stop-loss. It also runs ATLAS, an autonomous paper-trading bot whose every trade is logged publicly.

We are not a broker, not a licensed investment adviser, and not a fund. We never take custody of money, never place trades for users, and never publish personalised advice. Every output is general information produced by software, and it can be wrong. Read the risk disclaimer before acting on anything.

2. Data sources

  • Price and volume — end-of-day and intraday quotes from public market-data endpoints (Yahoo Finance). Quotes are delayed for most venues; we never present them as real-time execution prices.
  • News — headline feeds from mainstream financial press and, for crypto, the major crypto-native outlets. Headlines are used as inputs, not reprinted wholesale.
  • Sentiment — public social chatter and news tone, scored in aggregate. We do not quote individual private users.
  • Fundamentals — publicly filed figures such as revenue growth, margins and valuation multiples where the data provider supplies them.

Where numbers are estimates or model outputs rather than reported facts, we say so on the page. We do not invent statistics, testimonials, ratings or track-record figures.

3. How a signal is produced

  1. Universe scan. The scheduler pulls a defined universe — US large caps, a hype / small-cap set, EU listings and crypto majors — rather than an arbitrary list.
  2. Technical layer. Trend structure, moving-average alignment, RSI, MACD, VWAP and relative volume are computed from the raw candles, not from a third-party opinion.
  3. Context layer. Recent news and sentiment are summarised and checked for catalysts, earnings dates and known event risk.
  4. Reasoning layer. Two independent frontier models review the same evidence pack. Where they disagree materially, confidence is reduced.
  5. Confluence score. The signals are collapsed into a single 0–10 score. The method is documented in Confluence Score, Explained.
  6. Conviction gate. Only setups clearing our confidence and reward-to-risk floors are published as actionable. Everything else is shown as watch-only or discarded.

4. How results are graded

Every published ATLAS trade is written to the public ledger at entry, with its target and stop visible from the start. Positions are marked against market prices and closed by the same rules that opened them — target hit, stop hit, or timeframe expiry. Nothing is removed, re-dated or quietly re-graded after the fact. Losing trades stay on the record.

The bot record and the user record are kept strictly separate. Human paper trades on the leaderboard never touch ATLAS's statistics, and ATLAS never contributes to user XP. You can inspect the raw history on the bot track record and the methodology write-up in how our AI performs.

5. Written guides

Guides on the blog are written and edited by the SIGNAL9 team, with AI assistance for drafting and data summarisation. A human reviews every published guide for accuracy before it goes live. We date each guide and update it in place when the market facts change — exchange hours, index membership, or a strategy's measured behaviour — rather than republishing it under a new date to look fresh.

We do not accept payment for editorial coverage of a ticker, a broker or a competing tool. Comparison articles state our own product's weaknesses as well as its strengths.

6. Corrections

If we get something wrong, we fix it in place and note the change in the article's update line. Material corrections — a wrong number in a track record, a mis-stated exchange rule — are called out explicitly rather than edited silently. Report errors to corrections@getstockwhisper.com.

7. How SIGNAL9 makes money

Two ways: an optional Pro subscription that unlocks deeper scans and the full signal detail, and display advertising on public pages. Advertising is served by third-party networks and is labelled. Advertisers have no influence over which tickers the scanner surfaces, what a guide says, or how a trade is graded. We do not sell leads and we do not sell personal data — see the privacy policy.

8. Limitations we want you to know

  • Models are probabilistic; a high confidence score is not a guarantee.
  • Quotes can be delayed, and thin small caps can gap past a stop-loss.
  • Backtested or paper results exclude slippage, spread and taxes.
  • Coverage is limited to the universes we scan; absence of a ticker is not a verdict.
Advertisement