Methodology

This page explains where FundSpec's data comes from, how each model works in plain terms, how often each figure refreshes, and what each one cannot tell you. It is written so that a careful reader can judge the numbers rather than take them on trust.

Data sources

Company fundamentals are read from the financial statements filed with the US Securities and Exchange Commission and published on SEC EDGAR. FundSpec parses the filings themselves, not a vendor's summary of them, and keeps every reported period so that ratios are computed on a trailing twelve month basis from the actual quarters. Values are point in time: a figure for a given date uses only filings that were public on that date.

Prices, realtime quotes, index levels, options prints, analyst rating actions, price targets, earnings estimates and reported results, insider transactions and news come from licensed market data providers. Social sentiment is inferred from public social media posts that carry a stock's cashtag.

Refresh cadence

Quotes and index levels update during market hours. Unusual options prints are collected continuously through the session; free accounts see them twenty minutes after they print, and FundSpec Unlimited removes the delay. The AI trading models publish their calls as they are made during the session.

Fundamentals, financial ratios, technical indicators, momentum, fair value and the sector and industry relative value figures are recomputed every night after the close, so they are current as of the previous session. Analyst actions and the earnings calendar refresh every trading day.

Momentum

FundSpec Momentum classifies every covered stock and ETF as Bullish, Bearish or Neutral from the strength and consistency of its recent price path, and shows a confidence level that reflects how firmly that trend is established. It is a description of what the price has been doing. It is not a forecast, and FundSpec never presents it as one.

The Trader Profile setting changes the horizon the label reads. Conservative follows the long trend and changes rarely, so it suits position trades. Balanced reads the two to four week swing, matches the momentum chart in the app, and is the recommended default. Aggressive reads the current week's move and updates most often, so it suits short holding periods. The three horizons disagree on a large share of days by design: a bounce inside a broader downtrend is Bullish at Aggressive while Balanced stays Bearish. Whatever profile you choose applies to every momentum label and chart in the app.

Stocks priced below one dollar are not labelled at all, and very thinly traded names show Neutral at the profiles that demand more liquidity, because their price series are not reliable enough to describe.

FundSpec fair value

The FundSpec fair value is an estimate of what a company would be worth if the market priced its fundamentals the way it prices comparable companies. It blends two independent estimates.

The first asks what the company would be worth at its industry's typical pricing of earnings, sales, book value, tangible book value, operating cash flow and enterprise value to EBITDA. Peer groups are built from the SIC industry code in each company's own SEC filings, narrowing to the most specific group that has enough members. Benchmark multiples are set by the larger companies in each group, because that is where price discovery concentrates, and extreme values are trimmed before the typical multiple is taken. Every company is scored against its group whatever its own size.

The second asks what the whole market's cross section implies a company with these fundamentals should be worth. A gradient boosted model is trained each month on the accounting characteristics of every listed filer, with market value as the target and no price based inputs. The fitted value is the warranted market value for that set of fundamentals. All evaluation is walk forward: a month is scored only by a model trained on earlier months.

The two estimates are blended in log space, with more weight on the industry estimate. The gap between the fair value and the actual market value is then adjusted for the market as a whole, so that a period when every stock trades above its warranted value does not make every stock read as overvalued. The FundSpec rating labels a meaningful gap Undervalued or Overvalued and a large gap Significantly Undervalued or Significantly Overvalued; a small gap is Fairly Valued. The fair value range shown with the estimate is calibrated on history so that the price has typically fallen inside it.

Sector relative value and industry relative value are the first estimate on its own, at the sector and industry group widths, with the exact peer cohort used for each stock available on its page.

Fair value is an automated estimate from public filing data. It does not know about a pending acquisition, a product launch or a lawsuit, and it cannot value a company whose filings are missing or too old. Where the data is insufficient, the rating is Not Rated.

AI trading models

FundSpec's trading models are reinforcement learning agents trained one per ticker on that ticker's price history. An agent learns by trading a simulated market built from historical bars: at each bar it calls the ticker bullish, bearish or neutral, and it is rewarded when the price confirms the call with a large enough move over the chosen horizon and penalised when it does not. How large a move counts is set by the trader profile, so a conservative model demands a decisive move before it signals. The agent is then tested on a slice of history it never trained on. Models that pass are released and publish their calls during the live session.

Every released model carries its full record: the closed trades it has taken on its ticker, how often it wins, the average size of a winning move, and total gains against total losses, the profit factor. Wins and losses appear in the same list. Everyone can see the recent winning trades closed by the models; FundSpec Unlimited shows the calls as they are published.

Subscribers can train a model of their own on any US stock or ETF. The builder lets you choose daily bars for swing trading or intraday bars for day trading, the learning algorithm (PPO, A2C or DQN), the trader profile, how much history the agent sees per decision, the reward horizon, and how long to train. Training takes hours, and the finished model publishes signals to you in the same way the released models do. FundSpec does not place trades for any model.

Top Picks are daily lists built from the nightly data: growth at a reasonable price, dividend consistency, and the strongest bullish and bearish momentum readings.

Unusual options activity

An unusual options print is a single options trade whose size, premium or execution stands out from the normal activity in that contract. FundSpec tracks these prints across the US market through the session and scores each one. The premium is the money behind the trade. A sweep is an order split across several exchanges to fill quickly, which suggests urgency. Conviction tiers, from Notable through High Conviction to Whale, combine premium size, size relative to open interest, sweep execution and how the order was filled.

The pulse view sums every tracked print for the day: net premium is bullish premium minus bearish premium, tracked premium is the total money behind those prints and not total market volume, the put to call ratio compares put premium with call premium, and the sweeps figure is the share of prints executed as sweeps. The whales board ranks the largest prints for today and the trailing week. Flow screeners filter prints by expiry, aggressor side, cost basis, open interest, size, strike, trade count and volume, and subscribers can save screeners and receive alerts when a print matches.

Options activity shows where large sums are being committed. It does not show why. A large call purchase may be a directional bet, a hedge against a short position, or one leg of a spread whose other legs printed elsewhere.

Technical indicators

Technicals are computed nightly from adjusted daily closes: simple and exponential moving averages over five, ten, twenty, fifty, one hundred and two hundred sessions, the relative strength index, William %R, Chaikin money flow, the average directional index with its positive and negative directional indicators, MACD with its signal line, Aroon, the stochastic oscillator, and pivot points with two support and two resistance levels derived from the previous session's high, low and close. Chart Studio overlays moving averages and an oscillator pane on the price chart and adds trendlines, levels and measurement tools.

Analyst ratings

Analyst actions are the rating and price target changes published by covering firms: upgrades, downgrades, initiations and reiterations, each with the firm, the previous and current rating, and the price target. The ratings screen groups the market into the most widely covered stocks, the names with the most bullish and most bearish analyst positioning, the largest potential upside and downside against the median price target, and the battleground stocks where analysts disagree most.

Earnings

The earnings calendar lists the companies reporting on each date with the consensus estimate, the reported result once it is out, the surprise against the estimate and the price reaction. Each stock page keeps the history of past reports so that a company's tendency to beat or miss, and how its stock has responded, can be read at a glance.

Screener and compare

The screener filters US stocks by sector, industry, market capitalisation and FundSpec rating, then by any tracked fundamental, entered either as a raw value or as a percentile rank across the whole market. Percentiles are recomputed nightly. The compare view lines up stocks and ETFs on trailing returns rebased to a shared start, the fifty two week range, realised volatility, the worst drawdown over the past year, typical dollar volume, beta and correlation against SPY, analyst price targets, fundamentals with their market percentile, and, for subscribers, fair value and momentum.

FundSpec AI

FundSpec AI is a conversational assistant that answers questions about US stocks and ETFs from FundSpec's own data: fundamentals, valuation ratios, prices, analyst ratings, earnings estimates and, for subscribers, momentum and fair value. Its answers are grounded in that data rather than in general web knowledge, and it remembers the conversation so follow up questions work. Like every language model it can be wrong, and its answers are a starting point for research, not a conclusion.

Limits

FundSpec provides information for research and education. Nothing on the site or in the app is a recommendation to buy or sell any security, and no model's past record guarantees its future results. Data providers occasionally publish errors and corrections, and figures derived from filings depend on what companies chose to report and how. If a number looks wrong, tell us through the contact page and we will check it.