Case study, fintech product
Backtesting used to mean Python or spreadsheets. I made it a sentence you type, and a report you can actually trust.
Retail traders in India have plenty of ideas and almost no way to test them. Checking a strategy against real history meant writing code, wrangling spreadsheets or paying for a terminal.
So most people skip the test and find out with real money.
1 of 3
The hard part wasn't the maths. It was making a quant tool feel as simple as typing a sentence, without hiding the risk.
There's no feature grid above the fold. The hero, pictured at the top of this page, is a single input already filled with a real strategy: “Buy when RSI drops below 30 and price is above the 200-day average.”
A visitor understands the product before reading a word about it. Right under the input, a link lets them read a real report without creating an account.
Most backtesting tools lead with the return. I framed every test as a dive log: gains above the line, depth below it, and the worst loss printed next to the return at the same size.
The copy does the same job: “Your 3-year test never left the shallows. The loss row above is the part you keep.” Every charge is itemised and every result is labelled as simulated, because in a finance product trust is the interface.
The daily market summary opens with a sentence, not a table: how the index moved, how many stocks came along, and what kind of day that makes it. The numbers inside the sentence are highlighted so they still scan.
Every reading is written by rules from that day's data, and claims only count when three time windows agree. Beginners get the plain-English line. Experienced traders get the evidence underneath it.
Financial dashboards tend to paint everything red and green. Here, colour is reserved for direction. The 52-week range bars are deliberately one colour, so their length is the only thing you read.
2 of 3
I built the whole stack solo: the interface, the server, the backtest engine and the data behind it.
EdgeTest has a public side, with market pages, a screener and sample reports anyone can open, and a signed-in app where people write and run their own strategies. Both talk to the same backend.
A Next.js frontend handles the interface and API routes. An AI layer turns a sentence into precise trading rules, and a Python engine runs those rules across ten years of NSE prices, returning a full report in under ten seconds.
Public pages
Market Today, Options Today, IPO watch, screener, sample reports
Signed-in app
Strategy chat and saved results, behind Google sign-in (Supabase)
Next.js server
Pages, API routes, auth checks
AI strategy parser
Plain English into exact entry and exit rules
Python backtest engine
Trades, charges and drawdowns across the test window
Market data
3,519 listed companies: a core of about 500 with a median of ~18 years of history, the rest from 2016. Refreshed every trading day
3 of 3
A backtesting tool has no audience on day one. So the product had to bring in its own.
The public pages are useful on their own: a daily market read, an options summary, IPO tracking and a stock screener you can use in plain English, with no sign-up. Each one is written to answer a real search, like “free AI stock screener for Indian stocks”.
Every page ends a step closer to the product: read a sample report, then sign in with one Google button to run your own test. The daily Edge Meter has a share button, so each day's reading can travel on its own.
1
Search
Someone looks for a screener, an IPO date or today's market.
2
Free tool
They get a useful answer with no account.
3
Sample report
A real backtest shows what the product can do.
4
Google sign-in
One button. No forms, no password.
5
First backtest
Their own idea, tested on ten years of data.
Home, screener and Market Today on mobile.
“The hardest products make something complex feel as simple as typing a sentence. That was the entire design goal.”
Next case study
Coaching ERP
An admissions website and a role-based back office for coaching institutes.
