Paper trading
Paper trading fills orders in an internal simulator against live prices. Nothing leaves the platform, no broker connection is needed, and no capital is at risk.
It is the cheapest information you will ever buy about a strategy.
What it tests that a backtest cannot
This is the whole reason to do it.
| Paper tests | A backtest cannot |
|---|---|
| Current market conditions | It only ever saw the past |
| Whether your risk limits actually engage | Gates are per-deployment; the backtest does not run them |
| Whether the strategy fires at all now | A condition true historically may not be true in this regime |
| Whether trade frequency matches your expectation | Backtest frequency is a historical average |
| Your own reaction to the drawdown | A number on a page is not the same as watching it |
The nine execution gates run identically in paper and live. A paper run that trips your daily-loss cap has told you something true that no backtest could. That your limit is set where you thought, and that it fires when you expected.
A gate you have never seen trip is a gate you are trusting on faith. See How risk is enforced.
What it does not test
Be clear about the boundary, because paper is often over-trusted:
| Not tested | Why |
|---|---|
| Real fills | The simulator fills internally. No queue position, no partial fills, no venue latency |
| Real slippage in stress | Simulated slippage is a model, not a market |
| Broker rejection | No margin checks at a venue, no permission errors, no rate limits |
| Your reaction to real loss | Watching simulated money fall is not the same experience |
| Connectivity failure | No broker token to expire, no API key to be revoked |
Paper is a rehearsal, not a dress rehearsal. It confirms the strategy and your configuration behave as designed in current conditions. It does not confirm that live execution will match.
Paper is not a broker paper account
A distinction that has caused real confusion.
| Stretus paper | A broker's paper account | |
|---|---|---|
| Orders | Simulated internally. Nothing leaves the platform | Real orders that leave the platform |
| Settlement | None | None |
| Connection needed | No | Yes, separate credentials |
| Example | Any strategy, any market | Alpaca paper: keys begin PK, live begin AK, different hostnames |
Both are safe to experiment with, and they are not the same mechanism. See Connect a broker.
Deploying to paper
- Open a backtested strategy and click Deploy
- Choose Paper
- Set a capital allocation, use a realistic number
- Confirm
Three steps in the wizard: Save, Select mode, Deploy. The strategy starts evaluating on the next candle close.
Percentages only mean something against an absolute. A ₹10,000 paper allocation on an instrument trading at ₹2,800 buys three shares, and three-share arithmetic is not representative of anything you would run.
Paper with the amount you actually intend to allocate.
Isolation from live
Complete. Paper and live share a strategy definition and nothing else:
| Separate | Meaning |
|---|---|
| Capital | Paper uses a simulated balance; live uses real allocated capital |
| Positions | A position opened in one mode never migrates to the other |
| History | Separate trade and performance records |
| Risk metrics | Each mode's gates evaluate against that mode's own state |
Switching a deployed strategy between modes is possible from its detail page, with a confirmation. Open positions do not carry over. They run to their exit conditions in the mode that opened them. See Paper vs live.
Derivatives in paper
Derivatives are paper-only on this platform. The paper executor implements coordinated close for multi-leg groups, group exit planning, expiry handling and margin reservation.
Live derivative entry is refused. So paper is not a step toward live derivatives today. It is the whole available path. See Derivatives
How long to run it
| Timeframe | A reasonable minimum |
|---|---|
| 1m - 5m | 1-2 weeks |
| 15m - 1h | 2-4 weeks |
| 4h - 1d | 4-8 weeks |
The real criterion is trade count, not elapsed time. A daily strategy that took three trades in a month has not been tested; a 1-minute strategy that took 400 in a week has.
What to watch
| Watch | Question it answers |
|---|---|
| Trade count vs backtest frequency | Is the strategy firing at the expected rate now? |
| Exit-reason distribution | Are trades ending on your exit condition, or all on stops? |
| Any gate engaging | Are your limits real and set where you meant? |
| Drawdown in money | Is this the number you signed up for? |
| Longest losing streak | Does it feel like the backtest figure suggested? |
The exit-reason distribution is the most diagnostic. If everything is stopping out, the stop is doing the work and your exit condition is decoration.
When paper says no
Reasons to not proceed to live, each of which paper reveals cheaply:
- It does not trade. The condition is not true in this regime.
- It trades far more than the backtest suggested. Costs will be worse than modelled.
- A gate fires constantly. Either the limit is wrong or the strategy is worse than expected.
- The drawdown was uncomfortable. At simulated money. It will not be easier with real money.
- The exits are all stops. The exit logic is not what is ending trades.
Deploying a Grade A backtest straight to live is the single most common expensive mistake on any systematic platform. Paper costs time and nothing else.