What's shipped, what's coming, what's on the table.

When something failed, the app used to hand you the server's raw English sentence: on signup, at checkout, during a CSV import, when a save went wrong. Every one of those now reads in the language you set, with the recovery step named. Where a limit is the reason, it says which limit — a monthly AI allowance is not the same thing as "slow down for a moment", and it no longer pretends to be.

shipped this week

Twelve product panels became six, and the four-step explainer became three steps plus one look at what connecting actually does. In their place the page now takes the demo account's numbers apart: 64 winners bring in more trades than 56 losers, and the account still ends down — which is the whole argument for looking at your payoff instead of your win rate. Every figure comes from the demo account you can open without signing up. What PerpLog may and may not do with a connected exchange account now has its own section, ahead of the price rather than buried in the FAQ, and the "most popular" badge is gone because nothing backed it.

shipped 1 week(s) ago

The lab only judges trades whose market context it has computed, and the button that computes it was quietly broken. It always took the newest 200 closed trades, whether or not they were already done — so on an account past 200 trades the counter never moved, and the oldest trades could never be computed at all. It now works through them in batches and counts up as it goes, so you can see it advance and it picks up where it left off. The number on the button counts the trades that actually need work, account-wide, instead of the trades in your current filter.

shipped 1 week(s) ago

"Am I getting better?" is two questions, and a journal that answers them as one will flatter you. The development page now separates what you control from what came out of it. Each habit — following the plan, leaving the stop alone, sizing inside your cap — is priced in money: how much its change was worth over the trades of the recent half. Three things make that number worth reading. Habits that happen after the close are shown but never priced, because a review cannot have caused the trade it describes. Trades where two habits broke together are not counted twice. And the page ends by admitting how much of the change its habits do not explain at all, with a link to the place that might. If nothing can be priced yet, it says so instead of printing a zero that looks like a finding.

shipped 2 week(s) ago

A strategy that stops working is easy to see afterwards and hard to see in time. The page used to hand you four tools — a chart with eight lines, a heat map with ninety-six cells, a slope column — and left the conclusion to you. It now answers the question instead. Each strategy is measured against the rest of your book over the same months, so a bad quarter for everything no longer reads as one strategy decaying. Strategies are grouped by what they are doing, not sorted by a column you picked. And the page is honest about its own limits: at two to five trades a month almost every change fits inside its error band, so the verdict rests on the money and on how many of the last months actually lost — the part no error band explains away. It says that out loud rather than dressing a coin flip up as a finding.

shipped 2 week(s) ago

Every rule in a trading journal has the same weakness: you write it down, and then you are the one who has to remember whether you kept it. The rules the Strategy Lab proposes do not need that. It already knows, for every trade, which market condition it was taken in — so it can simply check. Set a rule at the end of a finding and the Playbook reports what it changed: breaches per ten trades before you set it against since, and the money in those breaches on both sides. A trade whose market context was never computed breaks no rule and counts in no denominator, because a number nobody measured should not read as discipline. And below five trades since you set it, there is no verdict at all — just what is still missing.

shipped 2 week(s) ago

Eight indicator cards of equal weight let you pick whichever one flattered you, and four of them were reading the same trend under different names. The page now answers one question: which market condition separates your good trades from your bad ones, and may you believe the separation? Splits are ranked by the money already lost in the weaker bucket. Every gap is printed next to its own error band, and because eight splits are tested at once the bar for a real finding rises from 95% to 99.4% — so the page will tell you, with a full history in front of it, that nothing is proven yet. Each split is re-measured in both halves of your history, the three trend duplicates are shown against the main split with what they actually add, and the finding ends in a rule with a target and a check date.

shipped 2 week(s) ago

There used to be a checkbox called “SL → break-even”, and it was wrong twice over. It logged a stop-loss move at exactly your entry price — a stop that, for most people, is not where they actually put it. And it had to be maintained next to the stop-loss history, so the two records could disagree with each other. If you take partial profit and then trail your stop a little past entry on market structure, the checkbox had no way to say so. Now the stop-loss history is the record and the answer is read off it: a stop at or beyond your entry means the trade can no longer close red on price. The line names the stop it used, so you can check it — and if your stop-loss field has drifted away from the history, it says that too instead of quietly picking one.

shipped 2 week(s) ago

The question the page exists for — at which size does my edge survive — is a comparison, so it stopped being a form you re-submit. Five sizes are simulated in one pass and stand on one scale, with the size you actually trade marked on it. A size is only recommended if it passes two tests, not one: ruin alone waves through a size whose normal run is a 25% drawdown. The horizon is your own yearly trade cadence rather than a round number, because a −50% drawdown is arithmetically impossible over 60 trades at 1% risk. And the assumption that binds — that trades are drawn independently — now reports your measured clustering instead of a disclaimer.

shipped 2 week(s) ago

A 0–100 score you cannot check is an opinion with a number on it. Each of the three skills now prints what goes into it: every input with its current reading, its reading at the baseline, the sub-score it earns, its weight and its contribution — then the sum, then the rounding to the tile’s integer as a separate line. The overall composite is gone on purpose, and the page says why: the average would hide which of the three moved. The baseline is the average of the first third of the window rather than a single month, and “what got better / what got worse” replaces the badge shelf, filtered to the window you are looking at.

shipped 2 week(s) ago

Plan adherence only ever described the trades you graded. The page now leads with both numbers — how often you followed the plan, and how many of your closed trades were never reviewed at all, priced per trade. Process against outcome is one 2×2 (good trade / bad luck / lucky / lesson) with a checklist line you can move between 60, 70 and 80 percent; the four boxes account for every scored trade at each setting. Click a box and you get its trades with the checklist step that was actually skipped. Where the data cannot support a claim — no answer recorded, no checklist stored, a planned-risk figure that dwarfs every real loss — the page says so instead of printing a zero.

shipped 2 week(s) ago

The page now opens with a verdict: which window pays for your day, which one gives it back, and what that costs per trade. The four six-hour windows are only the default view — switch to "from your data" and the windows are cut wherever your result flips sign. When the fixed grid disagrees with your data, the page says so outright, because an afternoon block averaging out to roughly break-even can hide a losing stretch that starts at four. Opening a window shows the hours inside it, the habits tagged there and the trades themselves, and a rule you wrote is measured against the day you wrote it — frequency and result reported separately, because they move independently.

shipped 2 week(s) ago

Instead of nine analyses stacked down the page, Analytics now opens with a ranked list of findings: what the behaviour costs, how many trades back it, and the one change to make. The headline figure is net of trades that two findings share, with the gross and the double-counted part shown underneath rather than hidden. Every confidence label states its own threshold, and a finding with too little data is separated from one that never occurred at all — those are different problems.

shipped 2 week(s) ago

Strategies are a dense table ranked by net result with a trend line per row, and your leaks are ranked by the money they cost rather than by category — grouped into "fix these first" and a long tail. Every number recomputes for the period you pick. Where a strategy has fewer than five closed trades, the win rate and profit factor stay hidden instead of dressing up noise as an edge, and each countermeasure you wrote shows whether it has actually reduced the breaches since you wrote it.

shipped 2 week(s) ago

The settings nav now has a search box that also understands what you mean — type "api key", "csv" or "dark" and the right section surfaces even when its name doesn't contain the word. Every section carries an icon, and on settings pages the main sidebar collapses to a slim icon rail so the settings themselves get the width.

shipped 2 week(s) ago

The risk page now opens with the one number that matters when you're underwater: your distance to break-even, with the two levers that close it — and a jump straight into the sizer to simulate them. The position sizer was rebuilt risk-first: enter your stop, get the size; plan partial exits per TP level, see the whole trade as a proportional price ladder with risk and profit zones, and copy the finished plan.

shipped 2 week(s) ago

The trade page now walks you through the review — evaluate, emotion, lesson — with a progress footer that shows exactly what's left. The chart carries the story itself: your running P&L as an overlay, best and worst excursion marked in money at the exact candle, every partial fill dotted in, and how long you sat in the trade written over the position zone. Screenshots live in a permanent evidence row instead of a buried tab.

shipped 2 week(s) ago

The weekly report now opens with your review agenda and an honest blocker when trades are still unjournaled. P&L attribution reads as one equation with a segment bar, daily P&L shows all seven days, and best/worst trades link straight to the trade itself — in the monthly and quarterly views too.

shipped 2 week(s) ago

The top of the page is now one statement: your readiness with the most urgent signal spelled out, one action, and today's state beside it. The daily session sits right under it as a slim strip. Below: KPI tiles that no longer clip their own text, and a new third row — cost & risk, your journaled emotions vs. their results, and P&L across the four global trading sessions, greyed while a session has too few trades to mean anything. Fresh accounts get honest dashed placeholders instead of fabricated zeros.

shipped 2 week(s) ago

Above the table there is now one plain sentence about the trades you are currently filtered to, plus five figures backing it up: trades, net P&L with the fees and funding behind it, win rate, average R, and how many are still unjournaled. Active filters show as removable chips, and every number follows them.

shipped 2 week(s) ago

When your edge is negative, Kelly and risk-of-ruin correctly go blank. The risk page now answers anyway: how far you are from break-even, in win rate and in payoff, and which group of your own trades explains the gap.

shipped 3 week(s) ago

Sync your Hyperliquid perps from a public wallet address. No API key to create, nothing to configure — paste the address and your history builds itself.

shipped 4 week(s) ago

Share a stats snapshot as an image, or enable your opt-in public profile — with proper preview cards when the link lands in X or Discord.

shipped 5 week(s) ago

Step-by-step guide with real screenshots for creating a read-only key — inside onboarding and as a public guide page.

shipped 6 week(s) ago

Explore PerpLog on a populated read-only account at /demo. No email, no API key, nothing to connect.

shipped 6 week(s) ago

One adaptive verdict above the dashboard, fused from tilt signals, discipline, open reviews and your daily loss limit. The journal that tells you when not to trade.

shipped 3 month(s) ago

What's shipped, what's coming, what's on the table. No hidden plans.

shipped 4 month(s) ago

New accounts get full Pro access for 14 days. No card on file. Decide once you have data on the screen.

shipped 4 month(s) ago

Ask a question right from the app — no Discord, no external chat. Reply within 48h.

shipped 4 month(s) ago

Step through your closed trade bar by bar — see entry, exits, MFE/MAE in context, not in a static screenshot.

shipped 4 month(s) ago

External monitoring of API and frontend uptime. Linked from the in-app help section.

shipped 4 month(s) ago

Realized PnL grouped by tax-relevant period, FIFO-compliant CSV ready for tax software or Steuerberater.

shipped 4 month(s) ago

Quantified, monthly view of which behaviors are improving and which keep biting you. Not vibes — numbers.

shipped 4 month(s) ago

When did each strategy work and stop working? Annotate market regimes, see the cause-and-effect.

shipped 4 month(s) ago

See whether your edge actually beats just holding BTC across any window you pick. Honest yardstick.

shipped 4 month(s) ago

Send 5–20 trades to the AI at once and get pattern observations across them, not just per-trade comments.

shipped 5 month(s) ago

Session-level replay (multiple trades in one view), more drawing tools, faster on slow connections.

Replace the current chart layer with TradingView Advanced — same indicators you already know, integrated with your trade data.

If a trader you refer signs up for Pro, you both get a free month. No multi-level nonsense.

Quotes from real traders with their X-handle so you can check they exist. No anonymous hype.

Binance Futures is the most requested next. Each venue done right — execution-level rebuild plus reconciliation — rather than many done approximately.