Inspiration
I'm an economics student who's been building my own trading tools — a position-size calculator, a growth simulator, a manual trade journal — while learning forex and crypto trading. The thing I kept running into was simple: I could log my trades, but I couldn't actually see the patterns in them. I knew I had a discipline problem somewhere in my losses, but spotting it by scrolling through rows in a spreadsheet never worked. I wanted my own journal to talk back to me.
What it does
Trade Coach takes your trade history — date, pair, direction, entry/exit price, size, P&L, and your own notes — and turns it into an honest coaching report:
- Pattern detection — automatically flags revenge trading after losses, stop-loss discipline issues, and losing streaks, pulled straight from your notes and numbers.
- An equity curve — a live chart of your account's cumulative performance, trade by trade.
- Kelly-fraction position sizing — calculates what fraction of your account your own win rate and reward:risk ratio actually justify risking per trade, plus a safer half-Kelly recommendation.
- Risk-of-ruin estimate — a probability of blowing the account if current behavior continues unchanged.
- A plain-English coach's summary — a written readout that ties the numbers back to what to actually do differently next week.
- A downloadable report — so you can save or share your analysis.
How I built it
It's a single-page HTML/CSS/JavaScript app with no external dependencies — trades are parsed client-side, pattern detection runs on regex-matched notes and P&L sequences, the equity curve is drawn with the native Canvas API, and the position-sizing/risk-of-ruin math is built on the Kelly criterion and a simplified gambler's-ruin approximation. Everything runs in the browser; nothing needs a backend or a database, which meant I could focus the build time on getting the analysis logic and design right rather than infrastructure.
Built with the help of Claude (Anthropic) for code generation and debugging — I directed the design, the trading-domain logic (what patterns to detect, how position sizing should work, what the coaching output should say), and the overall product decisions, while Claude wrote the implementation.
Challenges I ran into
Getting the pattern-detection logic to feel genuinely useful — not just "you lost money" but why — took the most iteration. I also had to think carefully about the position-sizing math: a naive Kelly calculation can recommend dangerously large position sizes, so I built in the half-Kelly recommendation that professional traders actually use, since the goal was to make someone a more disciplined trader, not a riskier one.
Accomplishments that we're proud of
- Going from a rough idea to a fully working, deployed app in one sitting — this is my first hackathon submission and my first time deploying anything live on GitHub Pages.
- Getting the Kelly-fraction position sizing and risk-of-ruin math actually correct and working off real, arbitrary trade data — not hardcoded numbers, genuine calculations that respond to whatever you paste in.
- Keeping the whole thing dependency-free — no frameworks, no backend, no API keys required to run — so anyone can open the live link and use it immediately, no setup.
- Turning something I built for myself (understanding my own trading mistakes) into a tool that's actually usable by anyone with a trade log, not just me.
What I learned
Real trading discipline problems are visible in the data if you look at the right cross-sections — losses immediately following losses, notes correlating with outcomes, win rate split by instrument. Turning that into automatic, readable findings instead of raw numbers is what actually makes a tool like this usable day to day.
What's next for Trade Coach, Personalized Instant Journal Analyzer
- Wiring in a real LLM call (the coaching summary is currently rule-based) so the written feedback adapts to nuances a fixed rule set can't catch
- CSV upload support instead of paste-only
- A version connected to a live trading journal so the analysis updates automatically as new trades are logged
Log in or sign up for Devpost to join the conversation.