Home / Research Process / Performance Review
ReviewIntermediate

Performance Review

Performance review is the stage in which a validated strategy's results are judged honestly against a relevant benchmark and the original hypothesis, on a risk-adjusted basis and after realistic costs, in order to reach a disciplined keep-or-kill decision.

Quick answer: Performance review is the stage in which a validated strategy's results are judged honestly against a relevant benchmark and the original hypothesis, on a risk-adjusted basis and after realistic costs, in order to reach a disciplined keep-or-kill decision.

In simple words

Performance review is deciding whether a strategy is actually worth trading once the tests are done. That means comparing it to a fair benchmark, checking whether it did what your hypothesis said it would, and looking at risk and costs, not just the headline return. The hardest part is being willing to kill an idea you have grown attached to when the evidence says so. A high return that only came with huge drawdowns, or that a simple index beat, is not a success.

Purpose

This stage exists because raw profit is a misleading judge: a strategy is only worth keeping if it beats a fair benchmark on a risk-adjusted, cost-aware basis and confirmed the specific edge its hypothesis predicted.

Visual explanation

Performance Review

Where performance review sits in the strategy lifecycle, feeding a keep, refine or kill decision.

Strategy LifecycleHypothesisBuildBacktestValidateDeployMonitorDecay /Retireiterate

Professional explanation

Judge against a fair benchmark, not against zero

A positive return means little on its own; the real question is whether the strategy beat what you could have earned with less effort and risk. The correct benchmark is a relevant, investable alternative, such as buying and holding the Nifty for an equity strategy or a risk-free deposit rate for a market-neutral one. A strategy that returned a healthy figure but underperformed a simple index, after accounting for the extra risk and effort it demanded, has not earned its place. Choosing the benchmark honestly, and before seeing the result where possible, prevents the common trick of picking whichever comparison makes the strategy look best.

Risk-adjusted, not raw, performance

Two strategies with the same return are not equal if one reached it through far larger drawdowns. Honest review therefore looks at risk-adjusted measures such as the Sharpe or Sortino ratio, the maximum and average drawdown, and the length of the recovery period, alongside the return itself. A strategy whose equity curve lurches through deep drawdowns may be untradeable in practice because no human would sit through them, regardless of its final figure. The blind spot of any single number must be stated: a high Sharpe over a short or single-regime sample, for instance, can be an accident of the period rather than evidence of durable quality.

Did it confirm the hypothesis, or just make money

A crucial and often-skipped check is whether the strategy earned its return through the mechanism the hypothesis predicted, or through something incidental. If your hypothesis was a weekly mean-reversion edge but the profit actually came from a single large trending move, the hypothesis was not confirmed even though the account grew. Attributing the return to its source, by examining the trade distribution and when the gains occurred, distinguishes a genuine, repeatable edge from a lucky by-product. A strategy that made money for reasons unrelated to its thesis has no reason to keep working, and should be treated as unvalidated.

Costs, capacity and the shape of returns

Review must use net figures after realistic brokerage, taxes such as STT, and slippage, because a gross edge that costs consume is not an edge at all. It should also consider capacity, whether the strategy still works at the size you intend to trade, and the shape of the return stream, whether profits came from many independent trades or a handful of outliers. A result driven by two or three exceptional trades is fragile, because removing them collapses the edge, and it should be reviewed with far more scepticism than a broad, evenly distributed set of gains. The distribution of outcomes matters as much as their sum.

The keep, refine or kill decision

The review culminates in a decision, and the discipline is to make it against criteria set in advance rather than to rationalise whatever result appeared. Keep applies when the strategy beat its benchmark on a risk-adjusted, net basis and confirmed its hypothesis with a robust, well-distributed edge. Kill applies when it failed the benchmark, contradicted its thesis, or depended on a few lucky trades, and killing should be the default under doubt because the base rate of genuine edges is low. Refine is the narrow middle path, permissible only if the change is a new, pre-committed hypothesis validated on untouched data, never an excuse to keep tuning a failed idea until it passes.

Guarding against confirmation bias in review

By the time a strategy reaches review, the researcher has usually invested effort and formed an attachment, which makes confirmation bias acute: the temptation is to weigh favourable evidence heavily and explain away the rest. Countermeasures include writing the keep-or-kill criteria before seeing final results, seeking a second reviewer who did not build the strategy, and deliberately arguing the case for killing it. The healthiest research cultures reward killing bad strategies as much as launching good ones, because a disciplined kill protects capital that an attached, optimistic review would have quietly put at risk.

Honest review vs flattering review

AspectHonest reviewFlattering review
BenchmarkFair, chosen in advancePicked to make results look good
Metric focusRisk-adjusted and net of costsHeadline gross return
Return sourceAttributed to the hypothesisAccepted regardless of cause
Reliance on outliersChecked and discountedIgnored
Default under doubtKillKeep and rationalise

Practical example

Illustrative example (Indian market)

A Nifty swing strategy on capital of Rs 5,00,000 shows a three-year net return that grew the account to Rs 7,35,000, a CAGR of about 13.7 percent, which looks satisfying in isolation. On review, the researcher compares it to simply holding the Nifty over the same period, which returned more with a smaller maximum drawdown, so on a risk-adjusted basis the strategy did not beat its benchmark. They also attribute the return and find that two trades during one trending quarter produced most of the gain, while the mean-reversion mechanism the hypothesis predicted contributed little. Because the strategy failed its benchmark, leaned on a couple of outliers, and did not confirm its thesis, the disciplined decision is to kill it, despite the positive headline number, since keeping it would mean trading an edge the evidence does not support.

Benchmark choice is consequential on NSE: an equity long strategy should be judged against a total-return Nifty index, and a costed comparison must include STT, stamp duty and exchange charges on both sides. A strategy that beats a price-return index but not a total-return one, or that only wins before costs, has not genuinely outperformed the simple alternative.

Limitations

  • Benchmarks are a matter of judgement, and an unfair or self-serving choice can make a weak strategy look good or a good one look weak
  • Risk-adjusted metrics have their own blind spots, so a single high ratio over a short or single-regime sample can mislead
  • Attributing returns to a mechanism is inexact, and a genuine edge can be temporarily masked by noise in a small sample
  • Keep-or-kill criteria set in advance still require judgement at the margin, where honest reviewers can disagree
  • Even a well-reviewed, kept strategy can decay after deployment, so review is a checkpoint rather than a permanent verdict

Common mistakes

  • Judging a strategy by its raw return without comparing it to a fair, investable benchmark
  • Focusing on headline profit while ignoring drawdown, recovery time and risk-adjusted measures
  • Failing to attribute the return, so a profit from luck is mistaken for confirmation of the hypothesis
  • Reviewing on gross figures and discovering only later that costs and STT erase the edge
  • Keeping a strategy that relied on two or three outlier trades whose removal collapses the result
  • Rationalising a failed strategy into a keep instead of accepting the kill the evidence supports

Professional usage

Institutional review judges a strategy against a fair, pre-chosen benchmark on a risk-adjusted, net-of-cost basis, and insists that the return be attributable to the hypothesised mechanism rather than to a handful of lucky trades. Keep-or-kill criteria are written before the final numbers are seen, an independent reviewer who did not build the strategy is often involved, and killing a bad idea is treated as a valued outcome rather than a failure. Under genuine doubt the default is to kill, because the base rate of real edges is low and the cost of trading a fitted strategy is paid in real capital.

Key takeaways

  • Judge a strategy against a fair benchmark, not against zero
  • Use risk-adjusted, net-of-cost figures, not the headline return
  • Check that the return came from the mechanism the hypothesis predicted
  • Discount results that rely on a few outlier trades
  • Under doubt, the disciplined default is to kill

Frequently asked questions

What is performance review in the research process?
It is the stage where a validated strategy's results are judged honestly against a relevant benchmark and the original hypothesis, on a risk-adjusted, cost-aware basis, to reach a keep-or-kill decision. It exists because raw profit is a misleading judge, and a strategy earns its place only if it genuinely beats a fair alternative for a reason its thesis predicted.
Why compare a strategy to a benchmark?
Because a positive return is meaningless without knowing what you could have earned more simply. Comparing to an investable alternative, such as holding the Nifty or a risk-free rate, reveals whether the strategy's extra risk and effort were rewarded. A strategy that underperforms a simple index has not earned its place, however positive its own figure.
What is risk-adjusted performance?
It is performance measured relative to the risk taken to achieve it, using tools such as the Sharpe or Sortino ratio and drawdown measures, rather than the raw return alone. Two strategies with the same return are not equal if one endured far deeper drawdowns, so risk-adjusted figures are essential for an honest judgement of quality.
Why check whether the return matched the hypothesis?
Because a strategy can make money for reasons unrelated to its thesis, which means the edge was not really confirmed. If a mean-reversion hypothesis actually profited from one trending move, the mechanism was not validated and there is no reason to expect it to repeat. Attributing the return to its source separates a genuine edge from a lucky by-product.
What is a keep-or-kill decision?
It is the disciplined choice at the end of review: keep the strategy, kill it, or narrowly refine it. Keep requires beating the benchmark on a risk-adjusted net basis with a confirmed, well-distributed edge. Kill applies when it failed the benchmark, contradicted its thesis, or leaned on a few lucky trades, and kill should be the default under doubt.
Why should killing be the default under doubt?
Because the base rate of genuine, durable edges is low, so most candidate strategies are noise, and the cost of trading a fitted one is paid in real capital. When the evidence is ambiguous, the expected cost of keeping a bad strategy exceeds the opportunity cost of killing a marginal good one, which makes a bias toward killing the prudent policy.
How do costs affect performance review?
Decisively, because a gross edge that realistic brokerage, taxes such as STT, and slippage consume is not an edge at all. Review must use net figures throughout, and many strategies that look profitable on gross numbers fail once the full cost of trading at the intended size and frequency is included.
Why is reliance on a few outlier trades a warning sign?
Because if removing two or three exceptional trades collapses the entire edge, the result is fragile and unlikely to repeat. A durable strategy tends to earn from many independent trades, so a return concentrated in a handful of outliers should be reviewed with far more scepticism than a broad, evenly distributed set of gains.
What role does the benchmark choice play?
A large one, because an unfair or self-serving benchmark can make a weak strategy look strong or a strong one look weak. The benchmark should be a relevant, investable alternative chosen honestly and, where possible, before seeing the result, to prevent the common trick of selecting whichever comparison flatters the strategy most.
How does confirmation bias affect performance review?
By the review stage the researcher is usually attached to the strategy, so there is a strong pull to weight favourable evidence and explain away the rest. Countermeasures include setting keep-or-kill criteria in advance, involving a reviewer who did not build the strategy, and deliberately arguing the case for killing it before deciding to keep it.
Can a profitable strategy still be killed?
Yes, and often should be. A strategy that grew the account but underperformed its benchmark on a risk-adjusted basis, or profited for reasons unrelated to its hypothesis, or depended on a couple of outliers, has not demonstrated a real edge. The disciplined decision is to kill it despite the positive headline number.
What is return attribution?
Return attribution is the analysis of where a strategy's profit actually came from, by examining the trade distribution and when the gains occurred. It answers whether the return was produced by the hypothesised mechanism or by something incidental, which is what tells you whether the edge is genuine and repeatable or a coincidence of the sample.
How is performance review different from validation?
Validation runs the technical tests that check whether an edge survives on unseen data, while performance review interprets those results to decide whether the strategy is worth trading. Validation asks whether the edge is real; review asks whether, given its risk, costs, benchmark and thesis, it is worth keeping. They are consecutive, complementary stages.
Does a good review mean the strategy will keep working?
No. Review is a checkpoint based on the evidence available, not a permanent verdict, because markets change and edges decay after deployment. Even a strategy that beat its benchmark, confirmed its thesis and passed review must be monitored live and retired if its behaviour departs from the validated expectation.

Voice search & related questions

Natural-language questions people ask about Performance Review.

What is a performance review of a strategy?
It is deciding whether a strategy is actually worth trading after the tests are done, by comparing it to a fair benchmark, checking its risk and costs, and seeing whether it did what your idea predicted.
Why compare my strategy to a benchmark?
Because making money is not enough if you could have made more by simply holding an index. The benchmark tells you whether all the extra risk and effort were actually worth it.
Why look at risk and not just return?
Because two strategies with the same profit are not equal if one put you through much deeper losses. A big return that came with terrible drawdowns may be impossible to actually trade.
Should I ever kill a profitable strategy?
Yes. If it lost to a simple index, made its money from luck, or leaned on a couple of freak trades, the honest move is to kill it even though it showed a profit.
Why is killing the default when unsure?
Because most ideas are not real edges, and trading a bad one costs real money. When the evidence is unclear, walking away is usually the safer bet.
How do I avoid fooling myself in review?
Decide your pass mark before you see the final numbers, get someone who did not build it to look, and make yourself argue for killing it before you decide to keep it.

Sources & references

    Last reviewed 12 July 2026. Educational content only — not investment advice. Markets and rules change; verify current conventions with SEBI, NSE/BSE and your broker.

    ← Previous
    Validation Workflow
    Educational content only — not investment advice. Examples use illustrative numbers and simplified models. Backtested results are hypothetical and trading derivatives involves substantial risk. See our Risk Disclosure and SEBI Disclaimer.