Why Your Algo Backtest Looks Profitable But Your Live Trades Aren't
The slippage problem that costs Indian intraday traders ₹15,000–₹20,000 every month — and how to fix it with tick-level data.
If your algo strategy looks great on paper but consistently underperforms live, you probably don't have a strategy problem. You have a data problem.
This is the story of how Raghunath — an intraday momentum trader from Hyderabad — spent eight months blaming bad luck before he found the real culprit.
The Strategy Was Fine. The Data Wasn't.
Raghunath ran a momentum strategy on mid-cap NSE stocks. Fast entries in the first 45 minutes, tight stops, exits before 11am. Clean, rule-based, no discretion. He'd backtested it thoroughly on 1-minute OHLC data and the numbers looked solid.
But every single month, his live P&L came in ₹15,000–₹20,000 short of what the backtest projected.
He adjusted position sizes. He tweaked entry conditions. He switched brokers once. Nothing moved the needle.
What 1-Minute Data Doesn't Show You
When you backtest on 1-minute OHLC data, your strategy executes at the candle open. Clean, simple, the way most retail traders set it up.
What it doesn't show you:
- The actual bid-ask spread at the moment your order hits the market
- How deep the order book is at that price
- The 3–4 second delay between signal generation and exchange execution
- How much the price moves in those 3–4 seconds on a volatile mid-cap stock
Each trade might slip by ₹0.50–₹2.00. Across 200 trades a month, that's your entire month's edge — gone. Raghunath was losing roughly ₹600 per day to slippage he couldn't see because his data wasn't granular enough to show it.
The Fix: Backtesting on Tick Data
A trader in his community mentioned TradeMade's tick-level data offhandedly. Raghunath looked it up out of curiosity.
Tick data captures every single transaction — every bid, every ask, every executed order on the exchange. When you backtest on tick data, there's no hiding behind candle opens. Your strategy has to survive the real microstructure of the market.
He re-ran his strategy on TradeMade using tick-level data for the same 6-month period. The simulated results dropped. Which was actually good news — because it meant the model finally matched reality.
"The backtest wasn't wrong about the strategy. It was wrong about the world the strategy was operating in."
💡 Running an intraday strategy on 1-min data?
There's a good chance your live results are consistently underperforming your backtest for the same reason.
Test your strategy on tick dataTwo Small Changes. A Very Different Result.
Once Raghunath could see the real microstructure, the fixes were obvious:
Tightened entry conditions to only trigger when a minimum volume threshold was present at the entry price level — not just when price hit the level.
Added a small spread buffer to limit orders placed in the first 30 minutes, when bid-ask spreads on mid-caps are widest.
Neither change was dramatic. Both required seeing tick-level data to even know they were needed.
The Numbers, Six Weeks Later
| Metric | Before | After |
|---|---|---|
| Monthly backtest vs live gap | ₹15,000–₹20,000 | Under ₹3,000 |
| Live performance vs backtest | ~82% | 94–97% |
| Strategy changes needed | Zero | Zero |
The strategy didn't change. The edge was always there. He'd just been measuring it wrong.
What This Means for Your Strategy
If you're running an intraday algo and your live results are consistently 10–20% below backtest, here's the honest checklist:
- Are you backtesting on 1-min or 5-min OHLC data? If yes, you're not seeing slippage.
- Are you trading mid-cap or small-cap stocks? Spreads are wider. Slippage is higher.
- Is your strategy high-frequency? More trades = more slippage events compounding.
- Have you modelled execution delay? Even 2 seconds matters in momentum trading.
Tick data won't fix a broken strategy. But if your strategy has a real edge and you're still losing to the market, the data is usually where the answer is.
TradeMade provides tick-level historical data for NSE and BSE instruments. Built specifically for Indian retail traders and quants who are serious about the gap between backtest and live.
Have a similar story about slippage or backtest gaps? We'd genuinely like to hear it — reach out here.