Algorithmic Trading Dashboard with HTML and LLM: A No-Code Case Study
Algorithmic trader Dmitry Ovchinnikov overcame Excel limitations—400,000 rows of data were slowing down analysis of 100+ strategies. Using an LLM, he created an HTML dashboard: dynamic P&L history, capital management, and browser-based visualization. The prototype was built in a day without writing code.
Excel and MetaTrader Analysis Challenges
Data from MetaTrader was exported to CSV and processed with VBA macros. The system worked, but lacked history: a missed daily export meant lost data forever. Volume reached 400,000 rows over months, causing lag even on powerful PCs.
MetaTrader reports are inaccurate: they distort margin and don't account for a unified cash position. No third-party services fully cover these needs.
Key pain points:
- No dynamic trade history.
- Difficulties calculating open positions.
- Lack of real-time capital overview by algorithms and instruments.
Dashboard goals: track fund allocation, strategy status, and daily P&L from the start of trading.
Algorithmic Trader Dashboard Requirements
The tool must provide operational and analytical management:
Operational:
- Where the money is: by algorithms, instruments.
- Strategy status (active/inactive).
Analytical:
- Breakdown of P&L by days/hours.
- Strategy performance analysis.
- Capital reallocation.
- Identifying weaknesses.
Result calculation for algorithm+instrument pairs includes:
- Closed trades (direct profit calculation).
- Open positions (daily snapshot).
The dashboard recalculates full history from CSV files on each load, unlike old MQL scripts with instant snapshots.
Prototype Creation with LLM
Dmitry described tasks to an LLM (Claude): load CSV from MetaTrader, process with JS in browser, visualize. No formal spec—iterative development via screenshots and examples.
Result: a single HTML file with data, JS logic, and charts. Works on PC and mobile without installation.
Process:
- Load CSV → parse trades.
- Group by algorithms/instruments.
- Calculate daily P&L (closed + floating P&L).
- Visualization: charts, tables, filters.
Problems and solutions:
- Correct floating P&L: several iterations, analyzing current HTML.
- Switching chats at context limit: model read the file, restored logic.
LLM comparison:
| Model | Result |
|----------|----------------------------|
| Claude | 1–3 iterations, precise code |
| DeepSeek| Loses context |
| Gemini | Breaks structure |
| ChatGPT | Lengthy spec discussion |
Prototype ready in a day, refinements iterative.
Functionality and Advantages
The dashboard provides a complete picture:
- Capital allocation tables.
- P&L dynamics charts by strategy.
- Filters by date, instruments.
- Algorithm status (detects inactive ones).
Changes for the user:
- Faster capital allocation decisions.
- Quicker response to strategy failures (previously took months).
- Transition from Excel+VBA to LLM→HTML.
The approach is universal: applicable to sales, personal finance—anywhere dashboards from tabular data are needed.
Key points:
- Dynamic P&L calculation with floating positions for each day.
- Single HTML file: portability, CSV updates.
- Iterative development via LLM without code.
- Scale: 400k+ rows, 100+ strategies.
- Time saved: prototype in a day.
Future Development
The current dashboard covers basic analysis. Possible additions:
- Risk metrics (VaR, drawdown by strategy).
- Performance heatmaps by instrument.
- Time-of-day analysis.
- Correlations between strategies.
- Real-time integration (WebSocket to broker).
— Editorial Team
No comments yet.