How Coach Alpha Reviews Your Trading (and What It Will Not Do)

"AI trading analysis" can mean two very different things. One is a model reading a pile of trades and writing confident prose about them. The other is software that computes what actually happened and then asks a model to explain it. The first reads well and is often wrong. The second is less magical and much more useful.

Coach Alpha, the AI reviewer in ohYaaa, is built the second way.

Find first, then explain

A language model is a strong writer and a weak calculator. Hand it two hundred trade rows and ask "what's your pattern?" and it will produce something plausible, which is not the same as true. So the pattern-finding is done in deterministic code before the model sees anything. Every figure that comes out carries its sample size.

The checks include:

• After a loss versus after a win: how your next trade performs in each case, and how often you increase size right after a loser.
• Trade number in the day: first, second, third-or-later trades compared, and your longest losing streak in a day.
• Time of day: results by hour, when enough trades have an entry time logged.
• Position size: your smallest third of trades against your largest third.
• Checklist: trades after a completed pre-market checklist against trades after a skipped one.
• Self-rating: whether the trades you rated highest actually did better.
• Day level: days with a written plan against days without, results by mood and market difficulty, and days you completed your habits against days you did not.

Then it reads what you wrote

The model is then given those facts, together with your own words: daily notes and plans, with the time each thought was written, and trade notes, setups, mistakes and triggers. Its job is to connect the two: "Your trades after a loss average X, and your 10:41 note says you felt you needed to make it back." That connection between what the numbers say and what you said at the time is something a spreadsheet cannot produce.

It scales to your data

The coach is told how much data there is. With ten trades, it should say the sample is thin rather than invent a pattern. With three months, it can say more. A good review tells you when it cannot tell.

What it will not do

• It will not tell you what to buy or sell. It reviews your behavior and results, not the market.
• It will not guess where the data is missing. No entry times, no time-of-day analysis.
• It will not grade you on P&L alone. Process quality is part of the grade.

Ways to use it

• Weekly and daily reports, weekly on a schedule and daily on demand.
• Custom reports: ask a specific question about a date range, such as "why do my Thursdays lose?"
• Ask-a-report: a question box on top of an existing report.
• Comparison: put two reports side by side and see what changed.

All of this is in AI Reports, free. The more you log, with times, tags, notes and ratings, the more the coach has to work with.