Every trade is a forecast. Brier replays a real historical setup with the future hidden, you commit a decision before you can see the outcome, and the server scores your process — calibration, discipline, expectancy. Not your luck.
Trading gives slow, noisy, self-flattering feedback. Four reasons most traders never find out where they actually stand:
A handful of good trades feels like mastery. Variance is large and samples are small, so outcome tells you almost nothing about process in the short run.
Self-reported journaling apps record the story you tell yourself. If your read of your own discipline is wrong, the journal is wrong in exactly the same direction.
Most study tools only hold currently-listed tickers. The names that diluted, reverse-split or went to zero are quietly deleted — so every backtest is a flattering lie.
Getting better requires seeing, in numbers, where your confidence and reality diverge. Almost nothing in a trader's workflow forces that confrontation.
One honest repetition at a time. The loop is the product; everything else supports it.
Define a universe and point-in-time conditions. Brier scans years of history — delisted names included — and returns every occurrence as a moment to study.
The chart is drawn up to the signal bar; the future is fogged. Read the setup, decide whether you'd take it, and write a pre-commit note.
Entry, stop, target, direction, size. The Gate checks the plan against your own rules and flags any violation before you commit.
The future un-fogs and the trade resolves into an R multiple — with realistic fill logic, minute-resolved when a daily bar spans both stop and target.
The server scores the decision: Brier score, execution gap, decision quality. Your cockpit updates. Nothing self-reported.
Lock promising setups and validate expectancy on held-out data, corrected for how many variants you tried. Keep what survives. Repeat.
The future is fogged from the signal bar. You decide without seeing the outcome — the only honest way to practise judgment.
Calibration, execution gap, expectancy and decision quality are computed server-side from your logged decisions. You can't fool yourself.
Daily history back to ~2000 including delisted names — the ones that diluted, reverse-split or went to zero. You study the true distribution, blow-ups included.
Double-click any daily candle, from any year, to drop into that exact day's 1-minute tape and see how the setup really resolved.
Lock a setup and validate expectancy on held-out data, with multiple-testing (Bonferroni) correction. Quant rigour for a discretionary trader.
One rigorous tool across both, daily and minute level, splits and dividends adjusted consistently. The same honest measurement everywhere you trade.
Define a universe — date range, exchange, sector, market-cap tier, minimum price, asset class. Add point-in-time conditions and match ALL or ANY. A plain-language summary restates your query, with a live universe count.
In study mode the chart stops at the signal bar and the future is hidden. In test mode you commit a plan; the Gate checks it against your rules and flags violations before you commit. Then the future un-fogs and the outcome resolves.
The Brier score and its decomposition — reliability, resolution, uncertainty — plus a calibration curve, an overconfidence index, execution gap and decision quality. The reliability curve shows where your stated confidence and reality diverge.
Lock a setup and run validation: expectancy with a confidence interval, corrected for how many variants you tried. If the lower bound clears zero on held-out data, the edge survived. If it didn't, you learned that cheaply.
A measuring instrument is only as truthful as what it measures. Brier is built on a dataset most tools quietly avoid.
When a company dilutes to nothing, reverse-splits, or delists, most data vendors drop it. Study on that data and you are quietly practising in a world where failure was erased. Brier keeps the delisted names — so when you study a setup, the blow-ups are in the sample, exactly as they were in real time.
point-in-time · split & dividend adjusted · equities + crypto
You want to actually get better, not chase tips. Brier gives you an honest practice range and a calibration score that moves only when your judgment does.
You have to prove calibration and discipline to pass an evaluation. Rehearse the gate, log the reps, and walk in with a measured record instead of a hope.
You want an objective measurement layer for your students. Brier turns "I think you're improving" into a number you can both point at.
Brier is a measuring instrument and a practice range. It is not any of these, and never pretends to be:
No alerts, no calls, no "buy this." It measures your decisions; it does not make them for you.
No course funnel, no secret. Just blind reps and the numbers they produce.
It raises the odds you improve and removes guessing. It cannot and does not guarantee money.
Start with a short calibration baseline. See where you actually stand, then begin moving it.
No. A backtester scores a mechanical rule over history. Brier scores you — your blind, discretionary decisions — and reports your calibration and discipline, not a curve-fit equity line.
In study and test, the chart is drawn only up to the signal bar; later bars are fogged in the interface and withheld server-side until you commit. You cannot see the outcome before you decide.
If your dataset only holds survivors, every study is biased toward success. Keeping delisted, diluted and dead names means the failures are in your sample — so your measured edge is honest.
A standard measure of forecast calibration: how well your stated probabilities match observed frequencies. Lower is better. It is the spine of how Brier judges process.
Both, at daily and minute resolution, with splits and dividends adjusted consistently — so the same honest measurement applies wherever you trade.
No. Brier has no signals and no opinions on individual names. It is an instrument for measuring and training your own judgment.