Automation removes the trader's hand from each order, which is its strength and its danger. A rule that is slightly wrong is executed perfectly, every day, until someone notices. A checklist is the cheapest protection against that: it turns the knowledge of what usually goes wrong into a list you tick before money is at risk. This page gives the full checklist in ten sections, shows it filled in for a real strategy, maps each item to the risk it protects against, and sets out the daily and monthly routines that keep a live strategy healthy. Print it, copy it into your notes, or keep it open beside the builder the first few times.
The checklist at a glance
| Section | The question it answers | Must be complete before going live? |
|---|---|---|
| 1. Rules | Does the strategy say exactly what to do in every case? | Yes |
| 2. Tests | Do we have evidence the rules work, after costs? | Yes |
| 3. Risk limits | What stops a bad day from becoming a bad month? | Yes |
| 4. Account | Can the account actually place these orders? | Yes |
| 5. Execution | Will the orders be the ones intended? | Yes |
| 6. Stopping | Can I stop everything quickly? | Yes |
| 7. Records | Will I know what happened and why? | Yes |
| 8. Daily routine | Who checks what, and when? | Before the first week ends |
| 9. Review | When and how will the rules change? | Before the first month ends |
| 10. Promises | Has anyone promised a result? | Yes: the answer must be no |
A short history of the checklist
Checklists became standard in aviation after a crash. In 1935, a new Boeing bomber prototype crashed on a test flight because an experienced crew forgot to release a control lock; the aircraft was declared "too much airplane for one man to fly". Instead of more training, test pilots wrote a short checklist of steps for take-off, landing and taxiing, and the aircraft went on to fly safely in large numbers.1 The lesson spread to medicine, construction and finance: when a task has many steps and mistakes are costly, experts still miss steps, and a short list catches them. Atul Gawande's *The Checklist Manifesto* (2009) set out how simple lists cut errors in surgery and other complex work.1
Trading automation has the same shape: many steps, each easy, any one fatal if skipped. Regulators have added items of their own. In India, SEBI's circular of 4 February 2025 on retail algorithmic trading, in force for every broker from 1 April 2026, made the broker responsible for algo orders, tied API access to a registered static IP and set a 10-orders-a-second threshold for unregistered algos.2 A circular of 2 September 2022 told brokers not to refer to the past or expected performance of any algo.3 Those rules belong on the list too.
1. The rules are complete
- Instrument, product and expiry rule are specified.
- Strike selection is a rule (ATM, a number of strikes away, premium or delta), not a strike number that will go stale.
- Entry conditions name the timeframe and use closed candles.
- Entry has a time window, the days it trades, and a maximum number of entries per day.
- Position size is fixed, or calculated from a rupee risk and the stop distance, using the current lot size.
- Every entry has a stop loss, with its basis stated: underlying price, premium or combined profit and loss.
- Every entry has a target or a trailing rule, and a time exit before the broker's square-off.
- Multi-leg strategies state what happens to the other legs when one exits or is rejected.
- Re-entry after a stop is allowed or not, explicitly, with a daily maximum.
- The product is right: intraday for same-day strategies; carry forward only where the strategy holds overnight.
2. The tests have passed
- Backtested over a long enough period to include a sharp fall or a volatile stretch, and across the expiry-day and lot-size changes of 2024 and 2025.
- Costs and slippage included, and the strategy still works with double the slippage.
- Enough trades to judge: dozens at least, ideally a hundred or more.
- Parameters are few, and neighbouring values give similar results.
- Checked on data the rules were not built on, by holding back the last part of the period or by walk-forward testing.
- Maximum drawdown is one you can sit through, and the capital covers it with room to spare.
- Forward-tested for two to four weeks, including an expiry day, with fills compared to the backtest's assumptions.
- The trade list has been read: a sample of trades did exactly what the rules say.
The backtesting vs forward testing guide explains what each test can and cannot show.
3. Risk limits are in place
- A daily loss limit for the account, set before the strategy goes live.
- A stop for the whole account across all strategies, at an amount decided in advance.
- A cap on orders a day, to catch a strategy that misbehaves.
- A maximum number of strategies running at once, and a maximum size per strategy.
- A rule for pausing the strategy after a drawdown of a set size, for example one and a half times the backtested maximum.
- Position size for the first month is one lot, whatever the backtest suggests.
- For your own code: a protective stop order at the exchange where the strategy allows it, so that protection exists even if your program disconnects.
4. The account is ready
- The F&O segment, or whichever segment the strategy trades, is active with the broker.
- Funds cover the peak margin the strategy needs on its worst day, including expiry-day additions, checked with an F&O margin calculator.
- The broker connection works: signed in from the platform, funds and positions visible.
- The static IPs used for API orders are registered with the broker.
- The second factor is an authenticator app, with the phone's clock set to automatic.
- No other strategy or manual trading uses the same contracts.
- For your own code: the order rate stays below 10 a second unless the algo is registered with the exchange through your broker.
5. Execution settings are checked
- Entry orders are limits with a buffer, or a "modify until filled" rule, rather than plain market orders in thin contracts.
- Exit orders cannot open a new position by mistake.
- Hedge legs are placed before short legs; on exit, short legs are bought back first.
- Quantities are in the right unit (units or lots) and are multiples of the current lot size.
- Orders stay below the freeze quantity, or are split.
- The time exit is before the broker's square-off time.
- The strategy is switched on once, and the order book was checked before switching it on.
6. You know how to stop it
- You have found the kill switch and know what it does: which segments it stops, until when, and whether it cancels open orders.
- You know the difference between pausing a strategy (no new entries), stopping it and squaring off, and stopping it while keeping positions you will manage yourself.
- You can sign in to your broker's own app and close positions by hand if the platform is unavailable.
- You know the broker's support number and the platform's.
7. Records are kept
- Every entry, exit, change and refusal is logged with a time.
- You keep a trade journal of every manual override, with the reason.
- Contract notes from the broker are saved monthly.
- The strategy's settings are saved, with a note of what changed and when.
8. The daily routine is set
- Before 9:15 am: sign in to the broker from the platform, confirm funds and the day's scheduled strategies.
- After each entry: check the order book and positions match what the strategy should have done.
- During the day: glance at alerts; do not adjust rules.
- After the close: review the day's trades, note any override, and confirm nothing is open that should not be.
9. The review is scheduled
- A monthly review date is in the calendar.
- At the review, live results are compared with the backtest and the forward test for the same weeks.
- Rules change only at the review, one change at a time, and each change is backtested before it goes live.
- There is a written condition for retiring the strategy, such as a drawdown beyond a set multiple of the backtested maximum.
10. Nothing has been promised
- Nobody has promised you a return from this strategy. SEBI bars brokers and their partners from referring to algo returns.3
- If the strategy came from a provider, it runs through your broker under the exchange's rules, and if its logic is hidden, the provider holds SEBI research analyst registration.2
What each item protects against
| Risk | Items that guard against it |
|---|---|
| A rule doing something you did not mean | 1, 2 (reading the trade list), 5 |
| A strategy that only worked in the past | 2 (unseen data, neighbouring values) |
| A bad day becoming a bad month | 3, 6, 9 (retirement condition) |
| Orders refused at the worst moment | 4 (margin, segment, static IP), 5 (lot size, freeze quantity) |
| Duplicate or runaway orders | 3 (order cap), 5 (switched on once), 6 |
| A platform or network outage | 3 (exchange stops for own code), 6 (manual exit) |
| Your own overrides | 7 (journal), 8 (no rule changes during the day), 9 |
| Being misled by a provider | 10 |
Our guide to algo trading risks explains each risk in detail.
Worked example: the checklist filled in
A trader plans to automate a Nifty expiry-day short straddle: sell the at-the-money call and put at 9:20 am, one lot of 65 each, a 25% stop on each leg, a combined target of 50% of the premium, and an exit at 3:10 pm.
| Item | Filled in |
|---|---|
| Instrument and strikes | Nifty, current weekly expiry, ATM by rule at 9:20 am |
| Size | 1 lot each leg for the first month |
| Stops | 25% of each leg's entry price; the other leg's stop moves to cost after one leg stops out |
| Target | Combined premium falls to 50% of entry |
| Time exit | 3:10 pm, before the broker's square-off |
| Re-entry | None |
| Backtest | Run over the period since the Tuesday expiry change, with costs and double slippage |
| Forward test | Three weeks, including three expiry days |
| Daily loss limit | ₹6,000 for the account |
| Margin check | Peak margin on expiry day, with the extra margin for short index options, covered with room to spare |
| Static IPs | Registered with the broker, shown as signed in |
| Kill switch | Located; set for F&O, it stops the strategy until the segment's next session |
| Review | First Saturday of each month |
Now put rupee figures on the worst realistic day. Suppose the legs are sold at ₹92 and ₹88. If the call stops at 125% of ₹92, about ₹115, it loses 23 × 65 = ₹1,495. If the put then reverses and stops at its moved stop, its entry price, it loses nothing more on that leg, and the day's loss is about ₹1,495 plus charges. If instead the market gaps through both stops, the loss could be several times that. The ₹6,000 daily limit allows for a bad gap without letting a disaster through, and the trader has decided in advance that losing ₹6,000 in a day will not change the plan.
The checklist did not make the strategy better. It made the trader certain of what the strategy will do, and of what happens when it goes wrong. That certainty is what lets a trader leave an automated strategy alone during the day, which is the whole point of automating it, instead of hovering over every tick and overriding it at the first scare.
Extra checks for indicator strategies
Strategies driven by indicators, such as moving-average crosses, RSI levels or Supertrend flips, add a few items to sections 1 and 2:
- The indicator settings are written down: the length, the price it uses (close, high, low) and the timeframe.
- Conditions are checked on closed candles, so a signal cannot appear and disappear while a candle is forming.
- The rule says what happens when two conditions disagree, such as a cross up while RSI is below 50.
- Each condition was checked by hand on the chart for a handful of past signals.
- The strategy has been forward-tested long enough to see at least a few signals; slow indicators on daily candles may fire only a few times a month.
- If the indicator is a custom formula, it has been checked against the chart's own built-in indicator where one exists.
- The data feed's gaps are handled: after a reconnect, a missed candle should not create a false cross.
Extra checks for your own code
If the strategy runs on your own program rather than a platform, add these to sections 4 to 7:
- Each order has a unique identifier, and after a timeout the program asks for the order's status before resending.
- The program reloads the broker's instrument list each day.
- It handles "too many requests" by slowing down, not by retrying at full speed.
- It stops placing orders when the session expires, and alerts you.
- Secrets are kept out of the code, out of logs and out of screenshots.
- It runs on a machine that does not sleep, with the static IP registered at the broker.
- Every order, response and decision is logged with a timestamp.
- There is a way to stop it at once, separate from the program itself, such as revoking the API key at the broker.
Checks before changing a live strategy
A change to a running strategy deserves its own short checklist, because most changes are made in a hurry:
- Write the reason. One sentence: what problem does this change fix?
- Change one thing. If two settings need to change, test them separately.
- Backtest the change against the current version over the same period, with the same costs.
- Forward test it alongside the live version for at least a week, unless it is a pure safety change such as a tighter loss limit.
- Switch the strategy off first, then edit. An edit and a switch-on at the same moment should never mix.
- Check open positions. If the strategy holds positions, decide whether to square off before changing legs, product or quantity.
- Log the change with the date, the old and new settings and the test result.
- Watch the first live day after the change closely.
Safety changes, such as a lower daily loss limit, can go live at once. Changes that make a strategy more aggressive, such as a wider stop or a larger size, should always wait for the next review.
Checks when several strategies share one account
Running three or four strategies on one account adds risks that no single strategy's checklist catches:
- Combined margin. Each strategy fits on its own; together, on expiry day, they may not. Add up the peak margin of every strategy that can be in a position at the same time.
- Combined loss. Three strategies with ₹3,000 stops can lose ₹9,000 on one bad morning. The account's daily loss limit should be set for the total you can accept, not for one strategy.
- Overlapping contracts. Two strategies that sell the same strike will each see the other's position as unexpected. Keep strategies on different contracts or different expiries.
- Exit-time crowding. Many strategies exiting at 3:10 pm send a burst of orders in one minute. Stagger exit times by a minute or two.
- Correlation. Two strategies that both sell premium on Nifty are really one larger bet. Count them as one when you think about risk.
The one-page version
For a quick pass before each switch-on, these twelve questions cover the essentials:
| # | Question | Tick |
|---|---|---|
| 1 | Does every entry have a stop, a target or trail, and a time exit? | |
| 2 | Is the strike chosen by a rule, with the current lot size? | |
| 3 | Has the strategy been backtested with costs and double slippage? | |
| 4 | Has it run in a forward test for two to four weeks? | |
| 5 | Is the daily loss limit for the account set? | |
| 6 | Is the size one lot for the first month? | |
| 7 | Is the broker signed in, with the static IPs registered? | |
| 8 | Does free margin cover the worst day, including expiry day? | |
| 9 | Are hedge legs placed before short legs? | |
| 10 | Do I know how to stop everything in one action? | |
| 11 | Is the monthly review in the calendar? | |
| 12 | Has anyone promised a return? (The answer must be no.) |
Scoring the checklist
Every item in sections 1 to 7 and 10 should be ticked before the first live order. An unticked item in section 1 or 3 means the strategy is not ready. An unticked item in section 2 means you do not yet know whether it works. Sections 8 and 9 keep it running once it is live, and should be in place within the first week and the first month.
Common reasons a checklist fails
- Ticking from memory. Open the settings and look; do not tick because you are sure you set it.
- Skipping it for small changes. "I only moved the stop" is how a missing time exit gets in. Any change means a fresh pass through sections 1, 2 and 5.
- One checklist for many strategies. Each strategy needs its own pass; a limit that suits one may not suit another.
- Ignoring section 9. Rules changed in the middle of a drawdown are the most expensive rules.
- No owner. If two people run the account, decide who does the daily routine.
A monthly review template
| Question | Answer |
|---|---|
| Did the strategy follow its rules on every trade? | Check a sample of trades against the settings |
| How did live results compare with the backtest and forward test for the same weeks? | Note the gap and its main causes: slippage, missed entries, overrides |
| What was the worst day, and was it within the plan? | Compare with the daily loss limit |
| Were there any manual overrides? | List them from the journal, with reasons |
| Is any change proposed? | One change, backtested, with a note of why |
| Is the retirement condition close? | Compare the drawdown with the written limit |
Fifteen minutes a month on this table does more for a strategy than any amount of watching during market hours. Keep the completed tables together; after a year they form the clearest record you will have of how the strategy really behaved.
Where automation checklists are going
The direction is towards checks built into the platform itself. Many items on this list, such as a missing stop, an expired contract, a signed-out broker or an exhausted daily limit, can be checked by software at the moment a strategy is switched on, and refused with a plain reason if they fail. AI assistants add another layer: asked to build or review a strategy, they can flag a missing stop or an unhedged sold option before anything runs. Neither replaces the trader's own pass through the list, because the most important items, such as whether the drawdown is bearable and whether the review will happen, are judgements, not settings.
How QuantPower covers the list
Many items are checked by the app when you switch an algo on. If a check fails, the algo stays off, you are told why, and no tokens are taken:4
| Checklist item | How QuantPower handles it |
|---|---|
| Live needs a connected broker | Without one, the app offers Forward test instead |
| Which broker | With two or more connected, you pick one; every leg goes there |
| Kill switch and daily loss limit | A switch-on that would enter while either blocks trading is refused |
| Broker signed in | Refused if the entry is now and the broker is signed out; otherwise the card warns you to sign in before the entry |
| Expired contracts | A fixed leg whose contract has expired cannot be switched on |
| Safety caps | Up to 20 algos running at once; up to 6 legs in a running algo |
| Tokens | The wallet must cover the charge |
| Changed since you looked | If the algo was edited after your screen loaded, you are asked to check it again |
Other items are settings you choose: per-leg stops, targets and trailing; whole-algo exit time, rupee target and stop, and lock-and-trail; buy legs first; entry order types including "modify until filled"; and the account limits in the M2M tab, which are Smart Exit, the daily loss limit, a cap on orders a day and the kill switch. Each algo card shows its legs, fills, profit and loss and anything it is waiting on, and the activity log records every entry, exit, change and refusal in one plain line each. Quanty can run a risk check on a strategy you describe, flagging a missing stop loss or a sold option without a hedge.
Section 2 is yours to run with QuantPower's backtests (2 tokens each, 10 free a week) and Forward test (100 tokens per algo per 28 days), and sections 7 to 9 are yours to keep. The paper vs live trading guide covers the step from Forward test to live.
When something does not behave as the checklist says it should, our guide to troubleshooting algo trading lists the usual causes and fixes.
Start with 500 free tokens
Connect your broker, build a strategy and test it on paper before a single live order goes out.
Frequently asked questions
How long does it take to get a strategy ready for automation?
Writing the rules takes an hour; testing them properly takes a few weeks, most of it forward testing. Rushing the testing is the expensive shortcut.
What is the minimum capital to automate a strategy?
Enough to cover the peak margin of one lot on the worst day, plus a cushion of at least one and a half times the backtested maximum drawdown. For a one-lot Nifty option-selling strategy, that is several lakh rupees.
Do I need to watch an automated strategy all day?
No, but you should check after each entry and after the close. Account-level limits and alerts cover the hours between.
What should I do if a live trade does not match the rules?
Stop the strategy, close the position by hand if needed, and find the cause before restarting. Common causes are a default setting, a unit error or a rejected leg.
How often should I review an automated strategy?
Daily for orders and positions, monthly for performance against the backtest. Change rules only at the monthly review, not after a bad day.
Which checklist items does a platform check for me?
On QuantPower, the switch-on checks cover the broker connection and sign-in, the kill switch and daily loss limit, expired contracts, the safety caps and tokens. The rules, the tests, the size and the review remain yours.
Should I go through the checklist again for every strategy?
Yes. Each strategy has its own rules, tests, size and limits, and a setting that suits one may not suit another. The account-level items in sections 3, 4 and 6 can be shared, but check them again whenever a new strategy is added, because the combined margin and the combined worst day change with every strategy you run.
Related guides
- Algo trading for retail traders: the full path from idea to live strategy
- MTM stop loss: limits for the whole account
References
- Atul Gawande, *The Checklist Manifesto: How to Get Things Right* (Metropolitan Books, 2009), including the history of the pilot's checklist after the 1935 Boeing Model 299 crash
- SEBI, circular on safer participation of retail investors in algorithmic trading, 4 February 2025, in force for every broker from 1 April 2026 (sebi.gov.in)
- SEBI, circular barring brokers from referring to past or expected performance of an algorithm, 2 September 2022 (sebi.gov.in)
- QuantPower: switch-on checks, safety caps, account limits and prices, from the app's code and price list, checked 12 November 2026