What is paper trading?

The output of a paper-trading phase is not the simulated balance. It is the journal. The balance is a by-product with a known upward bias, for reasons the next two sections make precise; the journal is unbiased evidence about the only questions a simulator can truly answer: can you define a rule, execute it without deviation, and review it honestly. A simulation run without controls produces neither, which is why most paper trading is a pleasant way to learn nothing.

Five controls separate a simulation from a game. First, trade the instrument you intend to trade live: an index future, a liquid large cap and a small cap are different markets, with different spreads, depth and gap behaviour, and skill in one does not transfer automatically. Second, use the size you would really use, derived from the capital you will actually deploy. A fantasy account of one crore teaches sizing reflexes that collapse on two lakh; set the paper capital equal to the real capital and size every trade from a fixed risk budget. Third, use the same order types you will use live: decide whether the entry is a limit or a market order, and whether the stop is a stop-loss market or stop-loss limit instruction, because that choice changes live outcomes and should be rehearsed, not improvised. Fourth, write the entry, stop, target and a one-sentence reason before the trade, never after; a record reconstructed afterwards is fiction with a memory of being right. Fifth, run long enough to cross regimes: a rule simulated only in a trending month is untested, so carry it through a choppy stretch, an expiry week and at least one gap open before concluding anything.

The journal that captures all of this needs a fixed schema, because ad hoc notes decay into commentary within a fortnight. Each field below exists to close a specific escape hatch, and the discipline of filling them in at the moment of decision rather than at the end of the week is the whole exercise: a simulation is only as good as the record it leaves behind, which is why it is worth running the finished log through a trade journal grader before you draw any conclusion from it.

A back-test asks whether a rule would have worked on data that has already happened, and it has its own long catalogue of ways to mislead, which the guide to back-testing integrity takes apart properly. Paper trading asks a narrower and, in one specific way, a better question: forward, in real time, with no hindsight available and no ability to peek at tomorrow's bar, can this rule actually be executed. That is a genuine advance over a back-test, and it is the reason the simulation stage earns its place. The trouble starts the moment you read the simulated balance as an estimate of what the method earns, because a simulator does not fill your orders. It grants them.

The cleanest way to see the size of that grant is to hold everything else still. Take one set of forty intraday signals on a stock near ₹500, two hundred shares a trade, so ₹1,00,000 of notional each time. Fix the entries, the exits, the sequence and the outcomes. Now run that identical set twice. The first run is the simulator's: every order fills at the price you asked for, instantly, in full, and nothing is charged. The second run books the same forty decisions at a fill the market would plausibly have given you, which means one tick of slippage on each side and the full statutory toll on every round trip. Not one thing about the strategy changes between the runs. Only the fills do.

The method underneath has a real if modest edge: twenty-one winners, nineteen losers, and an average of about ₹96 of gross profit per trade. On the simulator's fills that compounds into a record any beginner would be pleased to show someone, finishing ₹3,827 ahead, and the worst moment along the way is a dip of ₹184, which barely registers as a bad week. On realistic fills the same forty trades finish ₹280 behind, and the drawdown they pass through on the way down is ₹2,121, more than eleven times deeper than anything the paper record ever showed. Those are not two strategies. That is one strategy, priced twice.

Notice what this does not say. It does not say the method is bad. An average of ₹96 a trade against a ₹102.68 round-trip cost describes a method that is very nearly good enough, and the same signal set clears the toll comfortably on larger notional per trade, or with a wider average win, or at a lower turnover. That is exactly why the fill illusion is dangerous rather than merely inaccurate. It does not conjure a winning system out of a losing one. It moves a marginal system quietly across the line, and marginal systems are precisely what beginners build, because a beginner's edges are small by definition. The simulator flatters most where flattery is most expensive.

The slippage line in that gap deserves naming, because most simulations get its mechanism backwards. Exchange matching runs on price-time priority, so a limit order to buy at ₹500 joins the back of the queue of every order already resting at ₹500, and the price touching your level means only that the front of that queue traded. Touched is not filled. Worse, the error is not symmetric. When price kisses your level and turns, only the front of the queue fills and you miss the trade that would have worked; when price slices clean through the level, the entire queue fills, you included, so you reliably own the trades that are already going against you. That is adverse selection, and a simulator that fills you at the touch every time is handing you the single most valuable privilege on an exchange: queue immunity. The resting quantity that decides whether you were near the front of that queue is visible before you trade, in the market depth window, and reading it is the closest a retail trader gets to pricing the fill in advance.

A simulator answers one question with high fidelity: what would my rules have decided. It answers a second question with almost none: what would the market have given me for those decisions. The distance between the two is not vague. It is five specific subsidies, each of which biases the paper result upward.

First, fills at the touch, which the previous section priced. The mechanism is queue immunity: the simulator hands you the front of a price-time queue you were never in, and it does so on exactly the trades where the queue would have punished you. One everyday consequence is worth adding here, because it is invisible in a paper log rather than merely mispriced. Partial fills do not usually exist on paper at all. Live, a limit order for 200 shares can come back as 40, leaving you with a position that is a fifth of the size your stop and target were calculated for, and a decision to make about the other 160 that no simulation has ever asked you to make.

Second, zero slippage. A stop-loss is a trigger, not a price: when the last traded price crosses the trigger, a stop-loss market order is released and takes the next available price. Triggers fire, by construction, at moments when the market is moving against you, so the next available price is systematically worse than the trigger, and in a gap it can be far worse. Simulators execute the trigger price itself. Every stop-out in a paper log is therefore recorded at the kindest price the live market could conceivably have offered, and stops are exactly where the money is decided.

Third, zero costs. The Indian cost stack on a real round trip has seven lines, and the rates on several of them have moved twice in two years, most recently in April 2026. The next section prices every line in rupees, for intraday and for delivery, at the rates in force as of 18 July 2026.

Fourth, infinite liquidity at the quote. The best bid and offer on the screen are quotations for a finite quantity, the top of the order book. An order larger than that quantity walks the book, consuming successive price levels, which is market impact: your own order moves the price against you. The simulator fills any quantity at the last traded price. On a front-month index future or a large cap, small retail size rarely notices; on a mid cap, a small cap or a far out-of-the-money option, spread and impact are the difference between a strategy and a story about a strategy.

Fifth, zero consequence, which gets its own section below, because it is the one gap that does not shrink when you trade liquid instruments in small size.

The gap in the figure above was one number, ₹102.68. This section takes it apart, because a cost you cannot itemise is a cost you will not model. Take the cleanest possible test: an intraday round trip on a stock near ₹500, buying 200 shares for ₹1,00,000 and selling them later the same day at exactly the same price. Nothing happened. The simulator reports zero. The market does not, and the arithmetic is worth doing line by line, at the rates in force as of 18 July 2026.

One detail in that stack is worth pulling out before the table, because it is the line most guides quote wrongly. The exchange transaction charge on NSE cash is often given as 0.00297 percent. That figure is both stale and incomplete: it was superseded on 1 March 2026, and even before then it captured only the transaction-charge component while silently omitting the investor-protection levy sitting beside it. The all-in figure is ₹307 per crore each side, or 0.00307 percent. Note also what GST does and does not touch: it applies at 18 percent to brokerage, the exchange charge and the SEBI turnover fee, and not to the securities transaction tax or stamp duty, because those two are your own statutory liability rather than a service billed to you.

Rupees eighty-two and sixty-eight paise on a flat trade, at a flat-fee broker, before a single tick of slippage. Add one tick on each side, five paise on 200 shares twice, and the true cost of the round trip is ₹102.68. The simulator's breakeven is the entry price; the market's breakeven is the entry price plus a tenth of a percent, which on this notional means the stock has to move ₹0.52 in your favour before you have made a rupee. Run a hundred paper trades and the log is silently ₹10,268 per lakh kinder than the identical trades would have been live. That is more than enough to turn a marginal method's paper record from encouraging into meaningless, and it is the entire content of the gap in the figure above.

Delivery is heavier, not lighter, which surprises people who assume that holding longer dilutes the friction. Securities transaction tax on delivery equity is 0.1 percent on both legs, so ₹200 on the same notional round trip against ₹25 intraday; stamp duty rises to 0.015 percent on the buy; and the sell side attracts a depository charge, a flat fee per scrip for each day you debit shares from the demat account, levied in rupees rather than percentages, so it bites hardest on exactly the small positions a beginner should be trading. Even at zero brokerage the statutory floor on that flat delivery round trip is about ₹222 before the depository's debit fee, nearly three times the intraday number. The debit itself rides on a standing demat debit authorisation or a per-sale confirmation code, one more piece of live plumbing no simulator ever rehearses.

Derivatives have moved twice in two years, and this is where stale numbers do the most harm, because option premiums are small relative to notional and the tax lands on the wrong side of that ratio. The Finance (No. 2) Act 2024 raised STT on futures sales to 0.02 percent and on option sales to 0.1 percent of premium with effect from 1 October 2024. The Finance Act 2026 then raised both again from 1 April 2026: futures sales to 0.05 percent, and option sales to 0.15 percent of premium, with an exercised option taxed at 0.15 percent of intrinsic value in the purchaser's hands. Delivery and intraday equity were untouched by either Act. Any guide still quoting 0.0625 percent on option premium, or even the 0.1 percent that held for eighteen months, is describing a superseded regime, and a simulator built on those rates understates the cost of an options strategy by a wide margin. Verify all of it at source before you rely on it; the rates in this guide are stated as of 18 July 2026.

Everything priced so far can be narrowed by modelling. Charge yourself the stack, dock yourself a tick, refuse to count fills at the touch, and the paper curve moves most of the way toward the honest one. The fifth gap cannot be modelled at all, because it does not live in the simulator. It lives in you, and no amount of arithmetic reaches it. The loss-aversion literature that began with Kahneman and Tversky's prospect theory in 1979 keeps returning the same shape of result: a loss weighs roughly twice as much as an equal gain. Trading real money therefore runs on a loop that simulated money never starts. Fear of crystallising a loss makes traders hold losers and widen stops; the relief of a gain makes them cut winners early to bank the feeling; and a fresh stop-out makes the next trade bigger and faster than the plan allows, which is revenge sizing. On paper, none of it fires. Nothing is at stake, so the paper log measures your rules under laboratory conditions and says nothing about your ability to obey them under load.

The figure from the first section makes the point physically. The paper record's worst moment was a dip of ₹184 on a ₹1,00,000 book, which is not an emotional event at all; it is a rounding error you would not remember by Friday. The identical trades, filled honestly, drew down ₹2,121, and a real trader sitting in that hole has to decide, repeatedly, whether to keep taking the next signal. That decision is the one the simulation never asked for, and it is the decision that separates the two outcomes far more reliably than any refinement to the entry rule. Nothing in the paper log tells you how you behave at the bottom of a drawdown eleven times deeper than anything you have practised in.

The same weightlessness inflates paper results from a second direction, and this one compounds with the first. Simulated traders take entries they would never fund with real money, because a toy account makes "why not" feel like a reason. They sit through drawdowns no funded account would tolerate, because unrealised paper losses do not bleed. They hold overnight without checking the margin, and they carry positions through events they would have flattened. Every one of those behaviours pads the record, and none of them will survive contact with a real balance. The failure mode of the paper-profitable trader is therefore not that the edge disappeared. It is that the person executing it changed, and the log never contained a single observation of that person. The mechanics of that change, from loss aversion through to the sizing spiral that follows a bad run, are the subject of the guide to trading psychology; here it is enough to accept that the simulator measures a version of you who does not exist once money is on the line.

The live base rate is worth stating plainly, once, as context rather than as a scare. SEBI's study of individual trading in the equity derivatives segment found that 93 percent of individual traders in equity derivatives made net losses over FY22 to FY24, with aggregate net losses exceeding ₹1.8 lakh crore. That is the population a simulator's graduate joins, and no paper record, however clean, exempts anyone from it. What a disciplined bridge to real size can do, and it is set out further down this page, is stop you from crossing that distance in a single leap and finding out about the feelings gap at full size.

The toll and the queue are the simulator's contribution to the lie, and they are at least honest about being structural. They apply to everyone, they are the same size for everyone, and the previous two sections show exactly how to model them away. There is a second contribution, it is usually larger, and it is entirely yours. A consequence-free environment does not merely fail to punish bad habits. It actively invites them, because every mechanism that normally disciplines a trader has been removed at the same moment: the money, the finite account, the broker's record you cannot edit, and the simple fact that a mistake stays made.

Three habits do almost all of the damage, and the useful thing about all three is that they can be priced. The first is the retry. In a simulator a stop-out costs nothing, so the trade can simply be taken again at the same level, and again, until the level finally works, at which point the log records a winner and quietly forgets the three attempts in front of it. Nobody experiences this as cheating. It feels like conviction. The second is the fudge: filling in the entry price after the bar has closed, at a level the chart has already proven was available. A tenth of a percent is nothing to argue about on any single trade and it is invisible in a screenshot, which is precisely why it survives. The third is size drift, taking a larger position on the trades that feel right, which weights the winners you remember and shrinks the losers you would rather not.

Price the first two on the same forty signals used at the top of this page and the result is uncomfortable. Re-entering only the six deepest stop-outs until each one produced a ₹400 win adds ₹5,197 to the record. Logging every one of the forty entries a tenth of a percent better than the signal actually gave adds ₹4,000 more. Neither edit requires dishonesty in any conscious sense; both are the natural behaviour of a person practising in an environment where nothing is at stake. Stacked on top of the simulator's own ₹4,107 of structural optimism, they carry a record that honestly loses ₹280 all the way to a record that shows ₹13,024 of profit.

The cure is not willpower, it is procedure, and it is short enough to keep beside the screen. The rules below are not a wish list; each one closes a specific hole that the previous paragraphs opened, and the third column names what happens to the record when the rule is skipped. If that sounds austere for something with no money at stake, it is worth remembering the point of the exercise: a simulation is run precisely so that its output can be trusted later, and an untrustworthy output is not a smaller version of a good one. It is worse than no simulation at all, because it produces confidence without evidence, and confidence is the thing you are about to bet real money on.

Simulation itself has impeccable institutional credentials. NSE and BSE run periodic mock trading sessions, typically on Saturdays and announced by circular, in which members' systems fire orders at the live matching engine, including a planned mid-session switchover to the disaster-recovery site. The exchanges are explicit that mock trades create no rights, no liabilities, and no margin or settlement obligations. Notice what is being tested: order routing, risk checks, capacity, failover. No institution reads meaning into its mock-session profit and loss, and that is the correct posture toward every simulation: validate the machinery, ignore the score.

Around retail simulation, the regulator has spent two years drawing a careful boundary. A SEBI circular of 24 May 2024 barred exchanges, depositories and intermediaries from sharing real-time price data with third parties, a measure aimed squarely at platforms running virtual trading contests and fantasy stock games on live prices, and it permitted lagged data for investor education only where participation carries no monetary incentive. A second circular, dated 29 January 2025, regulated the educator side: a person engaged solely in education could not use market price data from the preceding three months while naming securities in any way that indicated future prices or a recommendation. In May 2026, SEBI harmonised the two into a single rule: a uniform 30-day lag for both the sharing and the use of price data for educational purposes, effective 1 July 2026. As this guide is written, that 30-day figure is days old, and any article still quoting the three-month rule is describing a superseded regime. The regulator's own securities-markets education institute, notably, was granted a one-day lag exclusively for its simulation lab, which is about as close as a regulator comes to endorsing simulation as a teaching instrument.

None of this restricts you. The circulars bind institutions redistributing data and people publishing education; a private individual paper trading against live quotes on their own screen is outside their scope. What the boundary tells you is what the regulator thinks simulation is for: prize contests dressed as practice were shut off, and teaching on lagged data was explicitly preserved. Practice, in other words, is education, not competition.

This is the rung almost nobody writes down, and it is the one that matters most, because the last mile of learning cannot be simulated. Everything above establishes that a simulator can prove your rules are followable and cannot prove you will follow them. The only instrument that closes that gap is money, in an amount small enough that the tuition is affordable and large enough that the loop actually fires. Almost every guide says "start small" and stops there. A usable protocol has rungs, promotion rules and demotion triggers, and it changes exactly one variable at a time, which is the same experimental discipline the simulation began with.

Promotion out of paper is earned by process, never by the simulated return, because the five subsidies have already corrupted that number. A reasonable bar: a block of thirty or more journalled trades of one named setup, spanning at least two regimes, with rule breaks near zero. Then go live at the smallest viable size, defined by feel rather than formula: a full stop-out should be financially ignorable and emotionally noticeable. That second half is not a figure of speech and it is the whole reason the rung exists. A size so small that losing it registers as nothing has not introduced consequence at all, and you have simply moved the simulation to a live account. The right size is the one that still hurts a little. You are not there to earn; you are introducing one new variable, consequence, while holding the strategy constant. Keep the single setup. Expect the numbers to degrade relative to paper, because fills, costs and your own pulse are now real; the degradation is the data you came for. Review on a fixed cadence, weekly, against the same journal schema, comparing process metrics rather than money: stops honoured, entries matching the written setup, slippage and costs inside what you modelled.

Promotion up a size rung follows a clean block at the current size, one increment at a time, and each increment should be small, because every new size is a new emotional regime: the moment a red number equals a day's salary, it stops being abstract. Demotion is automatic and immediate: a pulled or widened stop, a revenge entry minutes after a stop-out, size creeping above plan, or gaps appearing in the journal each cost one rung, back to smaller size, or back to paper if the process itself has broken. A losing streak with rules followed is not a demotion trigger; at any realistic win rate, streaks are a statistical certainty, which is expectancy arithmetic, not a verdict. Demotion is cheap. Discovering at full size that your discipline was never tested is not.

Run properly, paper trading proves four things, and it is worth being precise about them because they are genuinely valuable and routinely undersold. It proves you can operate the machinery without error, which is not trivial: order types, stop instructions and expiry handling are where beginners lose money for reasons that have nothing to do with analysis. It proves your rules are specific enough to be followed by the person you actually are rather than the one you imagine at the weekend. It proves your setup occurs often enough to be worth trading at all, which quietly kills a large fraction of promising-looking ideas. And it proves you can keep honest records under no supervision, which predicts more about a trading career than any entry technique. What it cannot prove is that the edge survives the queue, the cost stack and market impact, and that you will follow the rules when the loss is real. Those two are the fills and the feelings, and they are exactly the two that decide the outcome.

Held to that standard, paper trading is neither the indispensable rite its defenders describe nor the waste of time its critics call it. It is an instrument with a known calibration error, and an instrument with a known error is perfectly usable by anyone who knows the error. Use it to prove that your rules can be followed. Never let it tell you they will be.

The deeper point is that a simulator only rehearses what you bring to it. The setup worth testing, the level that invalidates it, and the sizing logic that survives being wrong are all upstream of any practice environment, and that upstream work is exactly what the method we teach is built around. Rehearse like an engineer, graduate on evidence, and let the journal, not the simulated balance, make every promotion decision.

• SEBI price-data norms for virtual trading and education. Circular SEBI/HO/MRD/MRD-PoD-3/P/CIR/2024/56 (24 May 2024) barred sharing of real-time prices with third parties, targeting virtual trading contests and fantasy stock games, and allowed lagged data for education with no monetary incentive; circular SEBI/HO/MIRSD/MIRSD-PoD-1/P/CIR/2025/11 (29 January 2025) set a three-month usage lag for educators; a May 2026 circular replaced both with a uniform 30-day lag, effective 1 July 2026. sebi.gov.in
• Exchange mock trading sessions. NSE and BSE circulars announcing periodic mock sessions, including disaster-recovery switchover drills; the exchanges state that mock trades create no rights, liabilities, margin or pay-in and pay-out obligations.
• The cost stack, as of 18 July 2026. Securities transaction tax under section 98 of the Finance (No. 2) Act 2004 as amended: delivery equity 0.1 percent on both legs (unchanged since 2012), intraday equity 0.025 percent on the sell, and, following the Finance Act 2026 with effect from 1 April 2026, futures sales at 0.05 percent and option sales at 0.15 percent of premium. NSE cash exchange transaction charges of ₹307 per crore each side, or 0.00307 percent all in, effective 1 March 2026; the widely quoted 0.00297 percent is both superseded and incomplete, having omitted the investor-protection component. SEBI turnover fee ₹10 per crore each side. GST at 18 percent on brokerage, exchange charges and the SEBI fee, and not on STT or stamp duty. Verify current rates at source before relying on them. nseindia.com
• Uniform stamp duty. Indian Stamp Act 1899, Schedule I, Article 56A, as inserted by the Finance Act 2019 and effective 1 July 2020: buy-side duty of 0.015 percent on delivery transfer, 0.003 percent on non-delivery, 0.002 percent on futures and 0.003 percent on options, charged at the central rate uniformly across states and collected by the clearing corporation. No securities stamp-duty amendment has been made in the 2024, 2025 or 2026 Finance Acts. pib.gov.in
• The behavioural evidence. Kahneman and Tversky, Prospect Theory: An Analysis of Decision under Risk (Econometrica, 1979), and the loss-aversion literature it began, for the finding that a loss weighs roughly twice an equal gain. SEBI's study of individual trading in the equity derivatives segment supplies the live base rate quoted once in the body of this guide.
• The two computed figures. The equity curves and the waterfall are computed, not drawn: one seeded set of 40 illustrative intraday signals on ₹1,00,000 of notional is run through the simulator's fill and through a realistic fill, and the difference is the toll above plus one tick of slippage per side. The retry and entry-fudge uplifts are priced on the same set at a ₹400 win per retried stop-out and 0.10 percent of notional per entry. The rupee amounts are illustrative and chosen to make the arithmetic legible; they describe no real instrument, trader or result.