Most trading content sells you a strategy. This teaches you how to judge one, any one, including strategies you build yourself. Expectancy, drawdown, overfitting, sample size, these aren't abstract statistics concepts, they're the actual tools that separate "this looks promising" from "this is genuinely worth trading." Here's the full framework, with real formulas and worked numbers, not just definitions.
What makes a trading strategy rules-based?
A rules-based strategy defines its entry, exit, and position sizing through specific, mechanical conditions, decided in advance, that don't require in-the-moment judgment calls. "Buy when price closes above the 20-day high on volume 1.5 times the 20-day average" is rules-based, it's checkable, repeatable, and testable. "Buy when it looks like it's about to break out" isn't, since two people, or the same person on two different days, could interpret that completely differently.
This distinction matters because only a rules-based strategy can actually be backtested or judged with any rigor. Vague, discretionary ideas can't be tested, since there's no fixed logic to actually run against historical data.
What is backtesting?
Backtesting means running a rules-based strategy's logic against historical price data to see how it would have performed in the past, trade by trade, calculating what each signal would have produced. It's the first, necessary step in judging a strategy, though on its own it isn't sufficient, covered under overfitting and sample size below.
How do traders calculate risk-reward and expectancy? What is expectancy?
Expectancy tells you the average result you can expect per trade, combining how often you win with how much you win and lose when each happens. The formula:
Expectancy = (Win rate × Average win) − (Loss rate × Average loss)
This single number matters more than win rate alone, and here's a worked example showing exactly why. Say a strategy wins 40% of the time, with an average win of ₹3,000, and loses 60% of the time, with an average loss of ₹1,000.
Expectancy = (0.40 × ₹3,000) − (0.60 × ₹1,000) = ₹1,200 − ₹600 = ₹600 positive expectancy per trade.
Now compare a strategy that wins far more often, 70% of the time, but with a poor risk-reward profile, an average win of just ₹500 against an average loss of ₹1,500.
Expectancy = (0.70 × ₹500) − (0.30 × ₹1,500) = ₹350 − ₹450 = −₹100 negative expectancy per trade.
The second strategy wins nearly twice as often as the first, and still loses money on average. This is exactly why judging a strategy by win rate alone is misleading, expectancy accounts for the full picture, both how often you're right and what each outcome is actually worth.
Win rate vs risk-reward?
These two numbers trade off against each other, and the expectancy examples above show exactly how. A high win rate can still produce a losing strategy if the average loss dwarfs the average win. A low win rate can still produce a profitable strategy if wins are large enough relative to losses. Neither number alone tells you whether a strategy actually works, only expectancy, which combines both, gives you that answer.
This is also why a strategy's risk-reward ratio, target distance divided by stop distance, gets decided alongside your entry and stop-loss, before the trade, since it directly shapes what your win rate would need to be for the strategy to have positive expectancy at all.
What is maximum drawdown?
Maximum drawdown is the largest peak-to-trough decline your account experiences during the test period, expressed as a percentage or rupee amount. Say your virtual account grows from ₹1,00,000 to a peak of ₹1,50,000, then falls to ₹1,20,000 before eventually recovering and growing further.
Drawdown = (₹1,50,000 − ₹1,20,000) ÷ ₹1,50,000 = 20%.
This matters separately from your final return, since two strategies can end up at the same final profit while taking completely different paths to get there. A strategy with a smoother equity curve and a 10% maximum drawdown is generally more tolerable to actually trade than one with the same final return but a 40% drawdown along the way, even though a backtest's final number alone wouldn't show you that difference.
What is overfitting? How can I avoid overfitting a trading strategy?
Overfitting means tuning a strategy's specific rules or parameters so closely to a particular historical dataset that it captures noise specific to that data, rather than a genuine, repeatable pattern. It's the single most common reason a backtest looks excellent and the live version disappoints.
Warning signs include: a strategy with many finely tuned parameters relative to how many trades it's actually tested on, parameters that don't have any logical or economic reasoning behind their specific values, they just happened to work on this dataset, and results that look dramatically better than simpler, more robust variations of the same idea. To actually avoid it, keep your rules as simple as the underlying logic genuinely requires, favor round, sensible parameter values over precisely optimized ones, and validate on out-of-sample data, a separate stretch of historical data your strategy wasn't built or tuned on, before trusting any backtest result.
How large should a test sample be?
A commonly cited practical minimum is 30 to 50 completed trades, spanning different market conditions, trending, choppy, volatile, rather than just one favorable stretch. Treat this as a reasonable starting benchmark, not a rigorously derived statistical proof, larger, more volatile strategies generally benefit from an even bigger sample before you trust the results.
Beyond raw trade count, split your data properly, build and tune your strategy's rules on one stretch of historical data, then test it, unchanged, on a separate stretch it's never seen. A strategy that only performs well on the exact data it was built on hasn't actually been tested yet, it's just been fit to a single dataset.
Why can live results differ from backtests?
Mainly because backtests typically assume better fills than you'd actually get live, rarely include realistic transaction costs unless you add them manually, and can't capture the psychological pressure that changes how consistently you follow your own rules once real money is genuinely at stake. Two companion guides on this site cover this gap in real depth, one focused specifically on why simulator and live results diverge, another walking through the full testing process on a live-condition simulator before real capital gets involved.
How do I know whether a trading strategy actually works?
Pull all of the above together into one checklist. Does it have positive expectancy, calculated properly, not just a high win rate? Is the maximum drawdown something you could genuinely tolerate, financially and emotionally, if it happened with real money? Was it tested on a real sample, 30 to 50 trades or more, across varied market conditions, and validated on data it wasn't built on? And have you accounted for realistic costs and slippage, not just the raw backtest numbers?
A strategy that clears all four checks is worth taking seriously. One that fails even one of them isn't necessarily worthless, but it tells you exactly what to fix before trusting it with meaningful capital.
Backtesting tells you how a strategy would have performed historically. Forward testing, running it live against real-time conditions with virtual money, is the necessary next step before trusting a strategy with real capital. Neostox's paper trading lets you do exactly that, testing any rules-based strategy on live NSE and BSE market conditions across equities, futures, and options, building the real sample size this framework depends on.