What's new

The old look was built on a purple-tinted ground, a purple-to-pink gradient and a neon glow, which is the stock look of most software shipped this year. The palette is now near-black with a single brass accent, and the rule underneath it is that the only colour in the interface is the market's: candle green and red are reserved entirely for price direction, so they never compete with decoration. Headings are set in a serif against a humanist body face and the existing monospace figures. Contrast was computed for every text and semantic colour pair in all three themes rather than eyeballed, and every pair clears WCAG AA.

New questions span market structure, candlesticks, smart-money concepts, risk, psychology and discipline, written to sit at the ceiling of what each topic can fairly ask against a chart. Difficulty is tiered per batch so a drill cannot fill up with easy rows, and a guard rejects any question whose wording asks you to predict what price will do next, because that is forecasting rather than chart reading.

Options were ordered with a random-sign comparator, which is not a shuffle: measured over 200,000 runs the correct answer landed in an end slot 67% of the time. The same comparator picked which questions a session drew, so with a stable pool order the same rows were starved every time and parts of the bank were close to unreachable. Both now use a seeded Fisher-Yates, and tests fail if any answer slot or any question's draw rate drifts off uniform.

Real OHLC is the summary of a price path: the high and the low are wherever price actually travelled. The old builder picked a close and bolted on a high and low as independent random numbers, so wick length meant nothing. Volatility now clusters into quiet stretches and bursts, momentum persists so candles come in runs, trends move in impulses and pullbacks rather than a constant drift, and every pattern chart in the app inherits it.

Pick a setup type, drag a box over the price and time area where you see it, and the server grades it on overlap against setups you never see. Correct marks score, incorrect and duplicate marks cost, and near-whole-chart boxes are rejected so the format cannot be brute-forced. Every challenge is validated before it ships: all six setup types present, no zone too thin to hit or wide enough to be free, none stacked ambiguously on another, and setups spread across the timeline.

You read the chart, commit to a decision, and only then see what actually happened next. Your process is scored; the outcome is not, and it never feeds a score. This is the one surface with genuine forward price data, which is why counterfactual replay exists here and nowhere else.

Mark as many taught zones as you can before time runs out. Speed and accuracy are both scored, and your best runs are kept so there is something to beat rather than just something to complete.

Rank moves on the quality and consistency of your training, not on how much you paid or how much a simulated account made. Seasons run monthly with their own standings, and trophies and titles carry onto your public profile.

The dashboard leads with a single daily run rather than a menu of options. A streak freeze covers one missed day a month, because a streak that punishes one bad day teaches you to fear the app rather than use it.

It reports whether your process scores are trending, whether the calls you were confident about actually scored better than the ones you were not, and names a single focus for the week ahead. Every number in it is computed from your own attempts. Where there is not enough data to say something honestly, it says so instead of filling the space.

Crews are small groups with their own standings. Sprint Duels put two people on the same seeded chart so the comparison is fair by construction: the same setups, the same time limit, and no advantage from a luckier draw.

Practice placing and managing trades on live prices with virtual money. The score is a Discipline Score: planned stops, journalling, and whether you respected your own guardrails. A hard 2% risk cap is enforced server-side. No real money is ever involved.

Your reasoning is stored alongside the decision, which is what makes recurring weakness detectable at all. The journal is yours; nothing in it is published anywhere.

Lessons are ordered beginner to advanced, and recap checkpoints reintroduce the topic you are weakest at rather than letting you leave it behind. Coach Mode turns that into an explicit plan for the week.

Quizzes, Spot the Setup, Marking Mode and Decision Reps are all graded server-side against data the browser never receives, so scores and leaderboards reflect training rather than developer tools.

Your profile shows what you have trained and earned. It never shows money, returns, or anything that could be read as a performance claim.

Beta feedback goes straight to the people building this, and a fair amount of what is listed above started as a message on that page.