STOCKPEAK HANDBOOK

Every day, a computer reads the financial reports and price history of roughly 7,000 US-listed stocks and narrows them to a shortlist of about 150. A separate AI analyst can then study the shortlisted names in depth and say whether each price looks too low, about right, or too high. This handbook explains how each step works, what every label on the screen means, and what the system is deliberately not doing.

If a word has a dotted underline, it is a technical term — click it to see what it means without leaving the page. A complete, searchable index of every term lives in the glossary section.

Behind the page is an ordered chain of jobs that re-runs automatically on a schedule via GitHub Actions. The order is enforced and the data is checked at each step, so stale or partial data cannot quietly corrupt the shortlist. In plain words:

1. Data comes in. Company filings from the SEC, prices and analyst forecasts from Yahoo Finance, and economic series from FRED (the Federal Reserve's data service).
2. Safety filters (“Tier-1 hygiene”). Stocks nobody could sensibly buy are removed: names that cannot be traded, companies that are tiny (under $300 million) or priced under $3 a share, thinly traded stocks, shell companies, companies with no usable filed fundamentals, companies with chronic losses and heavy debt, and smaller companies with strong signs of accounting manipulation. About 3,000 stocks are left to be scored.
3. The dual-door screen. Each remaining stock is scored through three Doors — compounder, value gap and trend leader — and the best cases go on the shortlist.
4. Bands. The shortlist is split into Research now and Watchlist. Everything else is Pass, or Vetoed if a safety filter removed it.
5. The AI analyst (when it is running). It studies shortlisted names in depth and gives each a verdict. It runs on a local computer, not in the cloud.
6. The gate. A verdict counts only if it passes the gate's rules. The ones that fail stay visible, marked as blocked, with the reasons.
7. Publish. Verdicts and reports are published to this site.
8. Paper books. Three simulated portfolios follow the shortlist, the AI verdicts and your own holdings, every day.
9. Grading. Each verdict is later checked against what the price actually did.

Everything is Point-in-time: each logged signal uses only information that existed at that moment. That discipline is what makes the track record trustworthy.

The desk is the main page. At the top is one switch that decides whose view you see:

• AI — the AI analyst's verdicts, sorted into the sections below.
• Quant — the screen's own ranking: pure math over financial statements and prices, with no AI involved.

Colour key — one meaning per colour

• ●Blue — the list: how far a stock has come through the screen (strongest for Research now). Also marks what you have selected.
• ●Violet — Quality: a high-quality business, and the scores behind it.
• ●Teal — Value: cheap against the growth it has delivered, and the scores behind it.
• ●Pink — Momentum / trend: a strong, steady price trend.
• ●Green — good: undervalued, a gain, a verdict that passes the gate.
• ●Off-white — fair: the price sits inside the value band.
• ●Coral — bad: overvalued, a loss.
• ●Amber — a warning: look closer (forensic warnings, the notice that a verdict is blocked, desk notices).
• ●Grey — does not count: the numbers of a blocked verdict, vetoed stocks, no data.

The sections of the AI lens

• Research now — names whose verdict is undervalued and that are not blocked (see the gate).
• Watchlist — names whose verdict is fair or overvalued and that are not blocked.
• Blocked by the gate — verdicts on record that fail the rules. They are shown as a record, with the reasons, and are never shown as picks.
• Awaiting — shortlisted names the analyst has not studied yet.
• Vetoed — names removed by a safety filter, and rows where the analyst produced no usable verdict (Not usable).

The value band strip

Next to a verdict you will see a thin strip. The shaded band is the range of values from the analyst's runs, the tick is the median (the middle value), and the white line is today's price. If the line sits left of the band, the price is below every run's value; inside the band, it is within the range; right of it, above every run. That is the whole verdict. See Value band (IV band).

The screen does not blend everything into one score. A company can deserve attention by being excellent, by being mispriced, or by being in a strong, steady uptrend — and averaging those into one number describes none of them well. So there are three Doors. A stock needs to clear only one.

How the doors compete

Each of the first two doors turns its score into a Percentile, and the better of the two is what a stock competes on. A Champion — a stock in the top 10% on both doors — gets a small bonus of +2, because being excellent and cheap at once is rare. A falling knife cannot be a champion.

Staying power

A name already on the shortlist stays until it falls clearly out. This is called Hysteresis (sticky shortlist): it stays in Research now while its rank is 60 or better, and on the shortlist while its rank is 150 or better. Without that buffer, stocks sitting on the cut line would hop in and out with every small price move.

Fair comparisons

• Most grades are Sector-neutral — a bank competes with banks, not with software companies.
• Momentum factor is the exception, in a useful way: it is also read across the whole market, so a boom that lifts an entire sector stays visible instead of looking ordinary.
• For cyclical businesses — oil, gas, mining, shipping and similar — whose cash flow swings with the cycle, the screen averages cash flow over several years (8 years for oil, gas and mining; 3 years for other cyclicals) instead of using one good or bad year.

Sector tilt is off

Every sector gets the same base Sector quota of shortlist places. The macro engine could give extra places to sectors that suit the economy, but its sector picking has not proved itself, so the tilt is switched off until it does.

The result of the screen is four Bands:

• RESEARCH NOW — the top of the shortlist, roughly the first 50 to 60 names. Worth your research time today.
• WATCHLIST — the rest of the shortlist, which holds about 150 names in all.
• PASS — scored, but not shortlisted.
• VETOED — removed by a safety filter. No band, and the reason is written on the red chip.

What removes a stock

A Veto is a hard disqualifier, applied before the doors score anything. The reasons are: the stock cannot be traded, it is too small or too thinly traded, it has no usable filed fundamentals, it is a shell company, it has chronic operating losses together with heavy debt, or it shows two Forensic warnings red flags at once. The last two apply only to companies worth under $10 billion.

What only warns

On large companies the forensic tests are warnings: Beneish M-score (likely earnings manipulation), Accruals (profits not backed by cash), Altman Z (a distress score) and heavy Share dilution / issuance from issuing shares. A warning never removes a stock worth $10 billion or more. On smaller companies a stock is removed only when two red flags agree. This keeps fast-growing leaders from being thrown out for looking unusual, while still showing you the warning so you can look closer.

Notes are not warnings

Data notes — for example that a stock's momentum was worked out from monthly prices, or that its latest annual report is old — describe how a number was built. They say nothing bad about the company and never remove a stock.

A shortlisted name can be studied in depth by an AI analyst — like getting a careful second opinion. The work is divided so that each part does what it is good at:

• The AI does the research and chooses every valuation input. It runs on a local model (LLM (large language model)), not a cloud service.
• Python does every calculation. An AI is a good analyst but an unreliable calculator, so no number is left to its arithmetic.
• Code checks the answer against outside anchors: the professional analysts' forecasts and their price-target range (the Street fence).

Each stock gets 2 or 3 independent runs, and each run ends in an estimate of what the business is worth. The verdict is where today's price sits against the band of those values: below it means undervalued, inside it means fair, above it means overvalued. A wider band means the runs disagreed, and that lowers the suggested size. See Value band (IV band).

A verdict counts only if it passes the gate. Every verdict is marked Actionable (yes or no), and when the answer is no, the desk gives the Gate reasons in plain words:

• Made by the old analyst — it was produced before the current rules, by an analyst that was ruled invalid.
• Runs disagree too much — the runs' values are spread too widely to trust.
• Only one usable run — a single run cannot show how much the analyst agrees with itself.
• Outside the analysts' price-target range — the value falls outside the street fence.
• Implausibly far above the price — the value is so much higher than the price that it is treated as suspect.
• No analyst price-target range — there is no fence to check the answer against.
• Analyst data could not be fetched — the lookup failed, so the check could not be made.
• Mine / oil & gas producer not yet supported — the analyst cannot yet value these properly, so they are blocked from being a buy.
• Calculator not used — the analyst never used the valuation calculator for its answer.
• Test run — the row came from a test, not a production run.

A few other technical reasons exist; the desk spells out each one in words. A verdict can have more than one reason.

These four things arrive with the relaunch. They are not live today.

• Entry timing. Each verdict says whether it is a “buy now” or “wait for momentum”, with a stated condition that would flip the call.
• Value basis vs momentum basis. A strong, fundamentally backed uptrend can be held at half or quarter size even when the analyst's value is below the price. See Position basis.
• Thesis monitor. Each verdict states its own invalidation rules. They are re-checked against current data, and a broken thesis is flagged. See Thesis status.
• Caution for uncovered stocks. A stock that no professional analyst covers gets a smaller suggested size and a stricter margin of safety, instead of being treated like a covered one.

A separate engine reads the state of the US economy and publishes three things:

• A probability for each economic season. These probabilities are the main output (see Probability vector). The single season label is only a summary and drives no number.
• Shock alarms for fear, credit, interest rates, oil, the dollar, jobs and inflation (see Shock register).
• A turbulence-risk flag, which turns on when the fear index is 30 or higher.

Its numbers come only from FRED. News is used as context for the story, never as a number. Its sector picking has not proved itself, so the screen does not use it.

Instead of showing a flattering Backtest, the system Paper tradings every day with real prices and real Transaction costs (bps). If the system is wrong, this page will say so.

The three books

• Equal-weight — holds every Research now name in equal parts. It is the pure stock-picking test, with no AI and no sizing.
• AI book — follows the analyst's verdicts that pass the gate (AI book (rn_depth)). Its history so far comes from the old analyst, which was ruled invalid. It has held only cash since 2026-09-24 because no verdict passes. When the new analyst goes live, the AI record restarts from zero and the old history is archived.
• Mine — your own saved holdings, tracked like a fund (Unitization) so that adding money never fakes performance.

Honest rules

• A trade happens at the first close after the signal, because a signal can only be acted on after it exists.
• Costs are included on every trade.
• Benchmarks (IWM and SPY) use the same dates as the trades.

How to read it

• Equal-weight vs IWM — the stock selection works if the picks beat the small-cap Benchmark.
• AI book vs Equal-weight — the AI analyst earns its keep only if its book beats the plain shortlist.
• Mine vs the others — your own deviations show up as the Behavior gap.

Sharpe ratio and CAGR appear only after enough days of live data; early on this page is deliberately boring.

The Portfolio tab is for your actual holdings. Each one is checked against the screen's Band and against the AI verdict, and the result is shown next to it. A holding whose AI verdict is blocked is shown as blocked — never as a signal. Your holdings also feed the Mine book on Track Record.

The Lenses button opens the older screens: the 100-bagger screen, Reverse, Paradigm and YouTube. They are kept for reference. They are not the current system, and nothing on the desk is built from them.

Three habits keep the system honest:

• Rules before results. Pass/fail rules are written down before the results are seen, so they cannot be tuned afterwards to look good (that would be Overfitting).
• Grading verdicts. A grader checks each AI verdict against what the price did over 30, 91, 182 and 365 days, compared with IWM, SPY and QQQ. A horizon is graded only once it has fully elapsed, and the benchmarks use the same dates as the verdict.
• Paper books. The Track Record books trade daily with real prices and costs.

• SEC filings — 10 years of as-filed fundamentals via SEC Company Facts (the ground truth for quality, forensic warnings and demonstrated growth), kept Point-in-time.
• Yahoo Finance — prices, analyst Analyst estimates, and coverage.
• FRED — US economic series behind the macro engine.
• News — narrative context only. It is never turned into a number.

The scoring chain re-runs on a schedule via GitHub Actions. The AI analyst runs separately on a local computer.

A searchable index of all 100 defined terms. Click any term chip to open its definition — or click a term inline in any section above.

This site is a research tool. It is not financial advice, not a trading bot, and not a crystal ball. The screen works on average over years, not on every stock every month, and the AI analyst is still being rebuilt. The system's own track record is honest and often humbling by design. Past performance — including paper-traded performance — does not guarantee future results. Do your own research.

Back to the Stockpeak desk