01 · A different kind of trader

From clicking to building

Picture a trader watching a screen for the same setup, again and again. Now give that setup explicit rules and let software monitor it. The work moves upstream: deciding what the rules mean, testing them, and checking what happens when they meet a live market.

The useful shift is from reacting to a chart to designing a repeatable process.

Seven chapters. Two interactive experiments. Start with the idea, then test the assumptions.

02 · Make the idea explicit

An algorithm needs a rule it can actually follow.

“Buy when momentum looks strong” leaves a lot unsaid. Which data? At what time? How is momentum measured? When is the signal invalid? What happens if the position is already open?

Algorithmic trading uses software to generate and submit orders when specified conditions are met. It does not have to use machine learning. A fixed rule can be an algorithm too. NSE’s definition ↗

Writing the rule exposes gaps that discretion can hide. It also makes those gaps testable. The specification is part of the strategy.

03 · Where the work moves

Automation changes the job description.

A broker API gives software a way to submit and track orders. A strategy platform can package parts of that workflow into an interface. Neither connection is evidence that the underlying idea makes money.

The opportunity is a more inspectable process: rules can be tested, decisions can be logged, and execution can be compared with intent. The trader still chooses the assumptions and remains responsible for deciding when the system should run.

Technical context: Kite Connect order API documentation ↗

04 · An edge has to survive friction

A good-looking backtest can disappear in the fills.

A backtest asks what a set of rules would have done on historical data under a model. Change the fees, fill assumptions, or data, and the answer can change. Repeatedly tuning the rules on the same history can also fit noise rather than a durable pattern.

Move the sliders to see how much of the assumed edge survives. Even with the same assumed win rate and payoffs, a small gross advantage can turn negative. Live results also vary because outcomes arrive in an uncertain order; positive expectancy does not prevent drawdowns.

Keep some data outside the development process, use realistic costs, and check how sensitive the idea is to small changes. None of these checks proves future profitability.

Further reading: QuantConnect on slippage and overfitting and out-of-sample testing ↗

05 · The market has its own state

“Order sent” is only the beginning.

Software can receive an order identifier without receiving a fill. The order may still be pending, rejected, cancelled, or partly filled. The strategy’s internal state must keep up with the broker’s actual state.

This is why execution logs, position reconciliation, alerts, and recovery rules belong in the design. A strategy can be sensible while its implementation produces unintended orders or misses the intended exposure.

Sources: Kite Connect order states and order updates ↗

06 · Keep a human in the loop

Discipline can be automated. Judgment needs an owner.

Rules can reduce ad hoc decisions during execution. They do not remove the choices made earlier: which idea to test, which results to trust, how much exposure to permit, and what counts as a reason to stop.

For Indian retail traders, the broker, provider, and exchange framework also matters. SEBI issued its retail-algo circular in February 2025; NSE publishes implementation standards and subsequent operational requirements. Check the current official material and your broker’s supported setup before connecting a system.

SEBI circular ↗ · NSE requirements ↗

07 · The new skill set

Build a process you can explain—and interrupt.

Algorithmic trading can bring research, execution, and review into a single workflow. That shifts attention toward data literacy, clear rules, realistic testing, and the ability to diagnose a system when something goes wrong.

The strongest version of this change is a trader who can explain why a rule exists, show how it was tested, account for every order, and recognize when the evidence no longer supports running it.

Technology makes a process easier to repeat. The quality of that process remains the real question.