Retail trading runs on claims nobody checks. We freeze a rule with a date before we look, grade it forward on public data, and publish the verdict either way. We started with our own screener — it failed, and we published that too.
Rule frozen before the fact • Graded forward, winners and losers • Recomputable from public data • Published either way
Free, one email per verdict, nothing else. Privacy · Or just read the experiments ↓
One finished. Four running in public right now. Each is a claim someone is selling, written down as an exact rule with a date, then graded forward against a control it has to beat.
We pre-registered the edge we believed in and ran it forward for seven weeks. It lost — and at 30 trades it had looked like it was working. That early read was noise.
The most published retail setup of the last twenty years, always sold on its win rate. We're testing whether a high win rate survives contact with expectancy.
The most widely taught indicator signal in retail trading — which makes it the least likely to still contain an edge, and exactly the claim nobody bothers to check.
A 40-year-old calendar claim still sold in seasonality newsletters. It was the strongest survivor of our 25-test sweep — in which, corrected honestly, nothing survived.
Over 27 years, essentially all of QQQ's return arrived overnight. Firms sold this as "night effect" ETFs and the ETFs died. Registered as attribution, not a tradeable edge.
These are not results we are advertising. They are the published outcome of Experiment 01, recomputed in your browser from the raw pick log every time this page loads.
Both of these are our own errors. We publish them because a record that has never issued a correction isn't a careful record — it's an unaudited one.
This site has one load-bearing claim: every pick is written down before the market opens. On seven occasions our scheduler drifted past the bell and logged anyway — and we didn't notice for six weeks. An outside reader found it in our public CSV. What broke, what it did to the numbers, why we won't claim it flattered us, and the gate that makes it impossible now.
A stock's price history is not a fixed fact — it gets rewritten every time a company splits. Mix a stored price with one you download later and you invent returns that never happened, and in cheap stocks the error is always in your favour. We found it in our own code, where it had turned a −2.9% baseline into a +8.0% one.
We never execute trades, hold your money, or touch your brokerage — ThePickLog is the scoreboard, not the casino. How your data's handled → The 7 checks any service should pass → Score a service against them →
Five rules. They are the whole method, and they are what makes a verdict here mean something.
Every frozen rule is in HYPOTHESES.md; every verdict is recorded in the audit log; the raw data is in picks.csv and outcomes.csv. Nothing here has to be taken on faith.
Before it was a testing lab, this was a scanner. Every weekday morning a cloud job screens the market's tiniest stocks — "low-float" names with very few shares available to trade, which can spike 20% or 100% in a day and collapse just as fast — and ranks the ones showing early signs of ignition on a simple published score.
That screen is Experiment 01, and it failed. We still run it, log it and grade it in public, for two reasons: the log is the raw material every rule here is tested against, and a scanner you can watch fail in real time is a more honest demonstration of the method than any description of it. The screen is what we point the honesty machine at. The honesty machine is the product.
Every name we log, live prices, a public leaderboard, and your own graded record next to ours. It costs nothing and it never touches a brokerage — the point is to find out what your instincts are actually worth before they cost you anything.
Everything we publish, and everything the site does. No card, no trial.
For people who want the lab to keep running. Planned, not for sale today.
Not early access to verdicts. Those stay free and public the day they land — a verdict only some people can read isn't published.
We'll send one confirmation email — you're not on the list until you click it. After that, one email when it opens. Nothing else, ever. Privacy
Point the machinery at your own rule. Planned, not for sale today.
We'll send one confirmation email — you're not on the list until you click it. After that, one email when it opens. Nothing else, ever. Privacy
The paid tiers are shown so you can see where this is headed — they are not available, nothing on this site is for sale today, and neither tier will ever include picks, alerts or a signal. They open only if and when the tooling is genuinely worth it and legal review clears. We'll say so out loud when that changes.
Create an account to trade the screen with play money and follow every experiment as it grades. Or send us a claim someone is selling and it goes in the public queue.
Educational / informational use only. ThePickLog publishes an impersonal, objectively-screened watchlist on a regular schedule. Nothing here is investment advice or a recommendation to buy, sell, or hold any security, and no content is tailored to any individual's circumstances. We are not a broker-dealer or registered investment adviser. The operators may hold positions in screened securities. Low-float / low-priced stocks are highly volatile and carry a substantial risk of loss. Past performance does not predict future results.
The low-float universe ranked by the published ThePickLog score. Click any row to see exactly why it screened.
The highest-momentum names, screened for business quality — so you can see at a glance which movers are real businesses and which are likely to rug you. Click any row for the full scorecard.
Fundamentals sourced from SEC EDGAR (public filings); live price/market-cap is Yahoo-derived and may be delayed. Not affiliated with the SEC.
| # | Ticker i | Tier i | Score i | Price i | Gap % i | RVOL i | Float i | Quality i | Watch level i | Trade (paper) i |
|---|
Fundamentals are read from SEC EDGAR public filings. Live quotes come from a market-data vendor whose terms don't permit public redistribution, so today's screen shows clearly-labeled sample prices to visitors — the banner at the top of the page says so whenever that's the case. The pick log, the track record and every experiment on this site are unaffected: they have never used this feed, and they re-derive from picks.csv and outcomes.csv. One known limit worth stating: float here is approximated from shares outstanding; true public float and short interest need a paid feed.
For the deployed app, set FMP_API_KEY in your Vercel
environment variables. Quotes and fundamentals are then proxied server-side — the key never reaches
the browser (same pattern as the Alpaca keys). Live quotes are owner-only: because the data
vendor's terms don't permit public redistribution, the proxy is gated (FMP_UI_TOKEN) and
visitors see clearly-labeled Sample data for the live screen. The pick log and track record are
unaffected — they never used this feed.
Don't paste a key here on the public URL — a browser-side key is visible to anyone. Use the Vercel env var above for the live demo; this field is only for local testing with a throwaway key. Note: float is approximated from shares-outstanding in the free FMP quote; true public float and short interest come from a paid feed / FINRA in production.
Owner-only. This panel drives the site owner's Alpaca paper account; every request
must carry the owner token (matching ALPACA_UI_TOKEN in Vercel), which is saved only in the
owner's browser. Everyone else: sign in on the Compete tab and trade with $100,000 in play
money instead. Orders route through a server-side proxy (/api/alpaca) — Alpaca keys never
reach the browser. Educational tool — the owner is responsible for every order placed.
A buy is blocked if its dollar value exceeds your per-trade cap or % of buying power, or if today's P&L has hit your daily loss limit. The Risk % drives the "Size to risk" button in the order ticket (shares = risk $ ÷ distance to your stop). Settings save in this browser.
Educational / informational only — not investment advice or a recommendation. Screened by objective criteria, identical for all users.
Your simulated Compete portfolio — balances, open positions, and order history.
| Ticker i | Qty i | Avg entry i | Current i | Mkt value i | Unreal. P&L i | Close i |
|---|
| Submitted i | Ticker i | Side i | Qty i | Type i | Status i | Cancel i |
|---|
| Submitted i | Ticker i | Side i | Qty i | Type i | Avg fill i | Status i |
|---|
Educational / informational only — not investment advice. Paper-trading only, no real money: Compete portfolios are a $100,000 simulation; the owner's Alpaca paper account is separate and owner-only.
Make a call the market can't let you take back. Freeze a rule with today's timestamp and it gets graded out-of-sample against the same append-only pick log the whole site runs on. The app gives you no way to edit, delete, or backdate it — only calls logged after you register count. Beat the baseline and you climb. Whiff and everyone sees that too. That's the point.
Build your rule from pick-time fields only (price, float, tier, regime…) plus an exit. No code, no future data, no take-backs.
Stamped and frozen — the app gives you no way to edit, delete, or backdate it, and it's scored only on picks logged after today. That's what makes your score mean something.
The public log scores you against the same-day open→close baseline, out-of-sample. Climb the board or eat the loss in the open.
| # | Hypothesis | State | n | Win% | Avg net | Δ vs baseline | Live |
|---|
No "proprietary AI." ThePickLog runs two transparent, deterministic models on every name. Neither is a buy signal — together they help you separate "this is moving" from "this is actually worth the risk." Below is what every number means and how to use it.
This page is about how a stock scores on today's screen. How a claim gets tested — pre-registration, the out-of-sample window, the tamper-evident log — is a different subject, and it lives in the method.
Each pre-market, the scanner ranks tiny low-float stocks by four observable facts: how few shares exist, how unusual today's volume is, how far the price has gapped, and the share price itself. A second, independent check reads the company's actual financial statements to ask whether there's a real business behind the ticker. Then the honesty machinery takes over: every pick is written to a permanent, append-only log before the market opens — enforced by a hard gate that refuses to log anything after 09:20 ET, added after we found seven sessions where it hadn't been — and graded automatically at prices you could realistically get — including how badly it fell, not just how high it spiked. Any new rule must be registered with a date in advance and proves itself only on future picks. In one sentence: a robot that picks volatile stocks every morning and grades its own homework a week later, in a log it can't edit.
Everything below is the technical detail of those two models — the exact formulas, weights, and what each metric means.
Answers "is this moving right now?" A weighted blend of four pre-market facts — float, relative volume, gap, and price. It's what ranks the Watchlist. High score = the stock is unusually active for its size, nothing more. Tiers A–D rank that intensity, not quality: in the live log the hottest tiers (A/B) have shown the deepest drawdowns — not better returns — so read a high tier as a downside flag, not a green light (Finding A).
Answers "is this a real business?" A fundamentals-driven read of the financial statements that scores quality, value and risk, and tags the name Investable / Speculative / Too Hard. Click any ticker (or its Quality chip) on the Watchlist to run it.
Float — shares available to trade. Smaller float, more violent moves. Sweet spot under ~3–5M shares; scored down toward a 50M ceiling.
RVOL — today's volume vs. the stock's own average. The single best tell that something unusual is happening. ~10× maxes the input; 5×+ is notable.
Gap % — move from prior close into the pre-market. Bigger gaps = more attention and momentum; ~20% maxes the input.
Price band — strategy targets roughly $0.50–$10. Outside the band is down-weighted, not auto-excluded.
A deterministic, rule-based second opinion — no AI in the score — layered on as a defensive filter. It reads the income statement, balance sheet and cash-flow statement and rolls seven categories into one 0–100 score and a risk label.
Financial health, business quality & valuation — revenue trend, margins, free cash flow, leverage, liquidity, returns on capital, and conservative value multiples.
Management alignment centers on share-count discipline — dilution is the #1 way low-float names destroy holders, and it's read straight from the filings.
The "too hard" filter tags each name Investable, Speculative, or Too Hard (pre-revenue, biotech, mining) — because knowing what not to analyze is half the edge.
Plain-English definitions for every metric in the screen and the deep-analysis scorecard — what it is, and how to read it.
Shares a company has freely trading in the market (excludes insider / locked-up shares).
Read it: smaller float = thinner supply = sharper moves. Under ~5M is the strategy's sweet spot.Today's volume divided by the stock's own average volume.
Read it: 1× is a normal day; 5×+ means unusual interest; very high RVOL on no news can be a pump.The percentage move from the prior close into the pre-market / open.
Read it: bigger gap = more overnight attention. Direction matters — chasing a huge gap is how you buy the top.The strategy's target range, roughly $0.50–$10 per share.
Read it: inside the band scores full; outside is down-weighted, not banned.An intensity band of the momentum score: A ≥75, B ≥60, C ≥45, D below.
Read it: a "how hot is the setup" shorthand — not a grade of the business. The hottest tiers (A/B) have shown the deepest drawdowns in the live log (Finding A), so treat a high tier as a downside warning, not a buy.A reference price set +20% above the screen price.
Read it: a yardstick for grading the screen later — not a price target or a recommendation.The smoothed annual growth rate of sales over the years available.
Read it: ~10–20%+/yr is healthy; a negative rate (shrinking sales) is a yellow flag.Profit left after cost of goods (gross), after running the business (operating), and after everything incl. tax/interest (net).
Read it: higher and positive is better. Negative operating/net margin = the business loses money on its core operations.Cash from operations minus capital spending — the cash a business actually generates.
Read it: positive FCF is the single best sign of a real business; persistent negative FCF means it survives on raising money.Cash produced by day-to-day operations, before capital spending.
Read it: negative for three straight years is a serious cash-burn red flag.Total debt divided by shareholder equity — how leveraged the company is.
Read it: ≤0.3 is conservative, ~1 is moderate, >2.5 is a high-leverage red flag.Current assets ÷ current liabilities — can it cover the next year's bills?
Read it: ≥1.5 is comfortable; below 1 means short-term obligations exceed liquid assets.Profit earned per dollar of capital put into the business.
Read it: ≥15% signals a genuinely good business; negative means it destroys capital.Net profit as a percentage of shareholder equity.
Read it: ≥20% is strong, but check it isn't just from heavy debt.How steady (and positive) year-over-year growth has been.
Read it: steady growth scores higher than the same average delivered in lumpy, unpredictable jumps.A stand-in for durable advantage: high, stable gross margin plus positive returns on capital.
Read it: high here suggests pricing power; it's an estimate, not a guarantee of a moat.Market value relative to annual revenue.
Read it: ≤1 is cheap, >6 is rich. Most useful when a company isn't yet profitable.Price relative to net profit.
Read it: ≤10 cheap, >35 rich; negative P/E = no profits, so the multiple is meaningless.Price relative to the cash the business throws off.
Read it: ≤10 cheap, >35 rich — harder to fake than earnings.Enterprise value (incl. debt, minus cash) vs. pre-interest, pre-tax operating earnings.
Read it: ≤6 cheap, >20 rich; accounts for debt that P/E ignores.Cash on hand minus total debt.
Read it: positive (more cash than debt) is a balance-sheet cushion and a plus for value.A conservative value (latest FCF per share × 14) and how far below it the price sits: (value − price) ÷ value.
Read it: a positive margin of safety (≈30%+) means the price is meaningfully below a cautious estimate. Needs positive FCF to compute.The change in share count over the years available.
Read it: flat-to-shrinking is great; +15% is a yellow flag; +50% is a red flag and the top way pennies erode holders.Percentage of the company held by executives and directors.
Read it: ≥10% means management eats its own cooking. Needs a paid feed — shown as "not checked" otherwise.Green / Yellow / Red / Black summary of the risk category and critical flags.
Read it: Black = a critical disqualifier (e.g. going-concern) and overrides the score regardless of how the rest looks.Investable, Speculative, or Too Hard (pre-revenue, biotech, mining, opaque financials).
Read it: Too Hard isn't an insult — it's the discipline of skipping what you can't reliably value.The methodology is provided for education. A high score is not a recommendation to buy; it describes how a stock ranks on these public criteria. Both models are deterministic, rule-based research tooling — not investment advice, and not a buy/sell/hold signal. Verify everything in primary filings before acting.
Every pick logged before the open, graded on that day's open→close return — all of them, winners and losers. This is the page most scanners won't show you.
Loading the live pick log…
Want the deeper cut? The model validation dashboard tracks every pre-registered hypothesis (exit rules, filters, a Bayesian read-out) out-of-sample — recomputed live in your browser from the same raw CSVs, so nothing here has to be taken on faith.
In August 2026 we found that our scheduler had drifted past the opening bell on seven occasions and logged 128 picks late. Those picks are excluded from every number here and the rows were kept, not deleted. Read the correction →
Does a higher score actually mean a better outcome? The real test is whether A beats B beats C on mean net return — so that is the column below, next to the median, the number of distinct companies behind it, and a ticker-clustered confidence interval. This is how you'd know, and how you'd kill the product honestly if it doesn't.
| Date i | Ticker i | Tier i | Score i | Screen px i | Watch i | Net i | Worst dip i | Result i |
|---|
Hypothetical research outcomes — may not reflect real execution, liquidity, or costs. Past performance does not predict future results. Educational / informational only.
A built-in guide for understanding what you're looking at — metrics, why a name screened, the quality lens — and for walking through a decision step by step. It answers only from the data this app has fetched and ThePickLog's published method. It will not pick stocks for you or give financial advice.