Der wöchentliche KI-Coach der TIC-App: ein Sonntagsreview, das sich erinnert

It is an Elite-plan feature, and it reads only one thing: the trades you logged in your journal over the previous seven days.

What Does the Weekly AI Coach Actually Review?

Your last seven days of journal entries, weighted towards how you executed rather than what you earned. It produces a one-line headline, a key metric, one to three things you did well, one to three things to watch, a single concrete focus for the coming week, and a short letter underneath.

The two inputs it leans on hardest are the ones most performance reports leave out. The first is playbook execution: which named setup each trade belonged to, and how each of those setups performed separately. The second is the mood you tagged the trade with, cross-referenced against the outcome. If your trades tagged as FOMO have a 22% win rate and your confident ones sit near 60%, that is a far more actionable finding than your overall win rate, and it only exists because you logged it.

Praise goes to the process, not the profit. A week that made money while you broke your own rules still earns the watch-out, which is deliberate: rewarding a badly executed profitable week teaches exactly the wrong lesson. Where your journal has risk logged, results are cited in R-multiples rather than pips, because R is the figure that compares across instruments.

Can It Make Up Numbers?

No, and the design reason is worth stating plainly. Every statistic in the review is computed in advance by the app's own code, in plain arithmetic, before any AI model is involved. The model receives a finished block of numbers and is instructed to narrate them, not to count anything itself.

This matters because a language model asked to tally rows will occasionally tally them wrong, confidently. Taking the arithmetic away from the model removes that failure entirely. If the coach says your win rate was 57%, that figure came from your journal by division, not from a model's impression of your week.

Two further guardrails sit alongside it. Your free-text journal notes are passed to the model as data and explicitly never as instructions, so a note you wrote to yourself cannot redirect what the coach does. And the coach is barred from predicting markets or recommending specific trades or levels: its scope is process, psychology and execution, and nothing else.

What Happens When You Barely Traded?

It tells you, rather than inventing a narrative. The coach needs at least three scored trades in the week to run at all, and every review carries a confidence rating that follows the sample honestly: low under five scored trades, medium under fifteen, high at fifteen or more.

A four-trade week gets a low-confidence review that says so in as many words. This is the opposite of how most analytics features behave, and it is the same reasoning behind the signal scorecard's unmatched bucket and the discipline streak's refusal to count profit: a number presented with more certainty than the data supports is worse than no number.

Even a high-confidence week is one week. Seven days of trading is a tiny sample against which to judge a strategy, which is the whole subject of how many trades a track record needs. What a single week can legitimately tell you is about your behaviour, and behaviour is what the coach is pointed at.

Why Does It Follow Up on Last Week?

Because advice nobody checks is entertainment. If last week's focus was to cut entries you tagged as FOMO, this week's letter opens by grading that against this week's numbers: you logged two, down from five, and that is real progress. If the data cannot settle it either way, it says it cannot tell, rather than manufacturing a verdict.

Every review is stored, and the history goes back up to a year, so a focus you were set in March is still there to be found in September. A coaching note you read once and never see again is not coaching, and keeping the record is what lets you see whether you are the same trader you were six months ago.

Who Is It For?

Traders who already keep a journal. The coach is a reader of your journal, and it can only be as good as what you put in it. If you log entry, exit, the setup you were trading, the risk you took and how you felt, you get a genuinely specific review. If you log nothing, there is nothing to coach.

The trading journal itself is free in the TIC app, as is the app. The Weekly AI Coach sits on the Elite plan, alongside the on-demand AI trade review that reads a longer window when you want it rather than on a schedule.

None of this predicts markets or promises returns, and the coach is explicitly built not to try. It is a structured way to look at your own execution once a week. The same preference for verified evidence over assertion is why TIC's Myfxbook record is published in full at /results. The app is free to download, and the journal it reads costs nothing.