What is Quantitative Trading?
Reviewed by Karthikeyan Anbarasan (Founder, TradeLyser) · Updated 19 Oct 2026 · Editorial policy
Quantitative trading is less about code than about precision. A rule is quantitative when two people given the same data would take exactly the same trades. That precision is what makes a strategy testable, and testing is where most of the work goes.
How a quantitative rule is built and tested
- Write the idea as a hypothesis: for example, Nifty tends to keep moving in the direction of its 20-day average once that average is above the 50-day.
- Turn it into exact rules: entry, exit, stop, position size, which contract, what time of day, and what happens on gaps and holidays.
- Split the data. Use one stretch (in-sample) to build and adjust the rule, and keep a later stretch (out-of-sample) untouched.
- Run the backtest with realistic charges and slippage, and record trade count, win rate, average win and loss, expectancy and max drawdown.
- Run the fixed rule once on the out-of-sample stretch. If it holds up, paper trade or trade it at minimum size before scaling.
- Keep comparing live results with the test. A live expectancy well below the test is a signal to stop and review, not to add size.
Worked example: costs on a Nifty futures rule
Charges decide whether many quantitative rules survive. Take one Nifty futures lot of 65 at an illustrative 25,000, so the contract value is ₹16,25,000 on each side. A round trip (buy and sell) costs roughly the following, with a flat ₹20 per order assumed for brokerage.
| Charge (illustrative, one lot round trip) | Basis | Amount |
|---|---|---|
| STT | 0.05% of the sell side | ₹812.50 |
| Exchange transaction charge | 0.00183% of ₹32,50,000 turnover | ₹59.48 |
| Stamp duty | 0.002% of the buy side | ₹32.50 |
| SEBI fee | ₹10 per crore of turnover | ₹3.25 |
| Brokerage (assumed) | ₹20 × 2 orders | ₹40.00 |
| GST | 18% of brokerage, exchange and SEBI charges | ₹18.49 |
| Total | ≈ 14.9 Nifty points | ₹966.22 |
A rule that looked like +20 points per trade in a cost-free backtest keeps about a quarter of that edge. Add one or two points of slippage per side and it can turn negative. The shorter the holding period, the larger the share that costs take.
How traders use it, including discretionary traders
You do not need to run an automated system to borrow the method. A discretionary trader can quantify filters (gap under 0.5%, entry before 11:00, stop no wider than 1 × ATR) and log them with each trade. Once 50 to 100 trades carry the same numeric tags, you can test whether each filter actually helped. Record a version number for the rule set on every trade, such as "ORB v3", and start v4 when a parameter changes; mixing versions in one tag makes the statistics describe a rule you no longer trade.
Where quantitative testing misleads
- Overfitting: trying many parameter sets and keeping the best one finds the settings that suited past noise.
- Look-ahead errors: using the day’s close to decide a trade that, in the rules, happens at the open.
- Ignoring regime: a test run only on a trending year says little about a sideways one.
- Survivorship: testing a stock rule only on today’s Nifty 50 members leaves out stocks that were dropped after falling.
- Small samples: as Figure 1 shows, 30 trades leave a win rate uncertain by about 18 points either way.
Quantitative trading
Algorithmic trading
Reference guide
| Context | Value | Reading |
|---|---|---|
| Rules fixed before the test, then checked on unseen data | Honest test | The out-of-sample result is the one to plan around, even when it is weaker |
| Fifty parameter sets tried, the best one kept | Overfitting risk | The winner may simply fit past noise; expect it to do worse live |
| Backtest without charges or slippage | Overstated edge | On Nifty futures, statutory charges alone are about 15 points per lot round trip in this example |
Questions
What is quantitative trading in simple words?
Trading from rules that are written in numbers and tested on past data, so the same market conditions always lead to the same decision. It removes judgment calls from entries, exits and size.
Is quantitative trading the same as algo trading?
No. Quantitative describes how the rules are made: numerically and tested. Algorithmic describes how orders are sent: by software. Many quant strategies are automated, but a tested rule can be traded by hand.
How many trades does a backtest need?
There is no fixed number, but under 100 trades the win rate and expectancy are very uncertain. At 100 trades a measured 50% win rate is still uncertain by about 10 points either way.
Is algo trading legal for retail traders in India?
Yes, through brokers that follow SEBI’s framework for retail algorithmic trading via APIs. The rules on registration and order tagging have been updated over time, so check the current SEBI circular and your broker’s terms.
Why do backtests do better than live trading?
Common reasons are charges and slippage left out of the test, parameters tuned to past data, look-ahead errors, and a market regime that has since changed. Testing on unseen data and paper trading first narrows the gap.