DEV Community

Rahul Goel
Rahul Goel

Posted on

How to Brief an Algo Trading Desk: The One-Page Document That Makes the First Call Useful

Most first calls with an algo desk or provider go the same way. The trader says he wants "a good algo for option selling". The desk asks a string of questions. He answers some of them from memory, guesses at others, and says he will get back on the rest. An hour later both sides know a little more and nothing has been decided.

That hour can be cut to fifteen minutes. The fix is a single page, written before the call, that answers the questions the desk is going to ask anyway. It also helps you, because writing it forces you to settle things you may have been avoiding, like how much you can lose in a day without changing your behaviour.

This article sets out what belongs on that page, line by line, with the reasoning for each. It uses real data from a study of NIFTY and SENSEX weekly expiry sessions to show how some of the numbers should be framed, and it ends with a sample brief you can copy as a layout. It applies whether you are speaking to a software vendor, a managed-account provider or a proprietary desk.

Why most first calls waste an hour
A desk cannot give you a useful answer to a vague question. "What returns can I expect?" has no answer until the desk knows your capital, your tolerance for a bad week, and what you will do when the strategy is losing. So the first part of any serious conversation is about you, not about the product.

The usual failure is not that the trader hides information. It is that he has never written it down, so each answer is a fresh guess. On Monday his daily loss limit is Rs 5,000. By Thursday it is Rs 15,000, because the first number was an estimate and the second is what he feels after a bad session. A desk that hears both numbers cannot tell which one to design for.

Writing the brief fixes this. It does three things.

It forces consistency. Once a number is on the page, you either stand by it or change it deliberately. It shortens the call, because the desk can read the page in two minutes and spend the call on the interesting parts. And it protects you, because a desk that proposes something outside your page has to explain why. If the proposal fits your page, you have a clear starting point. If it does not, you have an early warning.

The one page, at a glance
Here is the layout. Each row is expanded in later sections. The right-hand column shows what a filled-in line could look like. All values are illustrative and invented for the layout; they are not recommendations.

Where to send the brief, and who should read it
A brief is only useful if it reaches someone who can act on it. That sounds obvious, but a lot of traders send it to a general sales address and get back a brochure. The questions to settle first are who reads it and whether they trade or sell.

Look for a way to reach the desk directly. A good sign is a contact route that names the type of enquiry, shows a location and a phone or email, and does not force you through a chat widget first. A bad sign is a form that asks only for your phone number and nothing about what you want. If a firm is set up as a proprietary or quantitative desk, the person who answers should be able to talk about execution and risk, not just plans.

If you are looking for a desk based in the capital region and want a direct route, you can contact the EliteAlgo team in New Delhi through the firm's own contact page, and paste the one-pager into your message. Whichever desk you write to, do two checks before you send anything personal. Confirm the legal entity you are writing to, and confirm what it does and does not do. EliteAlgo, for instance, describes itself as a New Delhi proprietary and quantitative trading firm that provides analytical software. It is not an investment adviser. Reading that kind of description first tells you whether your brief is going to the right kind of firm.

Two more practical points. Send the brief as a PDF or plain text in the body of the email, and keep it free of identifiers you do not need to share. Your name and a contact number are enough. Do not send account numbers, API keys, PAN or screenshots of your holdings. A desk does not need those for a first conversation, and one that asks for them early is a flag.

Second, ask for a reply in writing to the key rows. A phone call is fine for discussion, but anything that affects your money should end up as a written line from the desk: what they would do, with what limits, and what they would not do.

Capital: three numbers, not one
The first row looks easy. "I have Rs 10 lakh." But a single capital number hides the three things a desk actually needs.

The first is the amount dedicated to the strategy. This is the money you will let the system use. It is not your net worth and not what is sitting in your broker account. Keep it separate in your own books, ideally in its own account, so a bad month cannot reach money you need.

The second is the margin that the strategy needs. For short option positions the broker blocks margin, and it changes with volatility and with the rules in force. Use your broker's current margin calculator for the structure you have in mind, and ask the desk to do the same. Do not rely on a number remembered from last year. The point of writing it down is to see how many lots your capital can carry with room to spare, since a position that uses nearly all your margin has no room to survive a spike.

The third is the floor. This is the account value at which you stop and review, before the strategy continues. It is the line in the sand. Most people skip it, then drift past what they can afford. A floor turns a vague fear into a rule.

Instruments and structures: be specific, and say what is out
"Option selling" is a category, not a specification. A desk needs to know which index, which expiry, and which structure.

On the index, state NIFTY, SENSEX or both. These are different contracts with different lot sizes, strike spacing and liquidity. NIFTY's lot size is 65. For SENSEX and for anything else, check the exchange's current contract specifications rather than trusting memory. Say which weekly expiry you want to trade. Weekly expiries exist for NIFTY on the NSE and SENSEX on the BSE. Bank Nifty weekly expiries were discontinued after the regulator's one-weekly-expiry-per-exchange rule, so do not ask a desk for them. A desk that says it will trade them is behind the times, or is not being straight with you.

On structure, name what you want in: short straddle, short strangle, iron condor, or a hedged version. And be just as clear about what is out. "No naked positions through the close" is a one-line instruction that changes the whole design. So is "no averaging down on a losing leg". These are the lines that save you money, because they stop a clever idea from becoming a dangerous one.

Last, say how you think about the premium. The size of the premium you can collect depends mostly on India VIX. In a study of 163 NIFTY and SENSEX weekly expiry sessions between January 2025 and July 2026, the 09:20 ATM straddle premium averaged 0.38% of spot when VIX was under 12, and 0.83% when it was over 20, with a steady rise across the bands in between. That is a doubling of the premium budget from the calmest to the most stressed. If your brief says "I want to collect about this much per day", the desk will rightly point out that the budget moves with VIX. A better line is "I want the structure sized so that a stop costs no more than X, in whatever VIX band applies."

Risk limits: write them as rupees, then as rules
This is the row that matters most, and the one most often left vague. "Low risk" is not a limit. A limit is a number in rupees, with a time frame, and a stated action when it is reached.

Write three limits. A daily loss limit: when the day's loss reaches this, everything is closed and no new entries are made until tomorrow. A weekly loss limit: when reached, the system is paused until you and the desk have talked. And a per-position stop: the rule that closes a single position, expressed in a form the desk can code, such as a rise in combined premium by a stated multiple, or a spot move by a stated percentage.

Here is an illustrative piece of working to show how these connect. Suppose you dedicate Rs 10 lakh and decide that one bad day should cost no more than 1%, which is Rs 10,000. You plan to sell a NIFTY straddle with a combined credit of 100 points and to exit if the combined premium rises 30%, to 130. That is a 30-point loss per unit at the stop. With a lot size of 65, one lot loses 30 x 65 = Rs 1,950 at the stop. Rs 10,000 divided by Rs 1,950 is about 5.1, so on paper, five lots fit inside the daily limit.

Now stress that working. It assumes the stop fills exactly at 130. In a fast market it may fill at 140 or worse, which changes the loss to 40 x 65 = Rs 2,600 per lot, and five lots lose Rs 13,000. It also ignores costs. So a careful trader would cut the size to four lots, or widen the daily limit on purpose. The point is not the exact answer. It is that a written daily limit, a stop rule and a lot size can be checked against each other in a minute, and a desk that sees them can tell you where they do not fit.

Add one more line: what happens to the limit on a gap or a halt. A stop is not a guarantee. If trading is stopped or the market jumps through your level, the loss can be larger than planned. Say so on the page, and say what you expect the desk to do about it, whether that is smaller size, a hedge, or avoiding certain days.

Finally, say what you measure success against. The same study looked at a short ATM straddle entered at 09:20 and found that the median give-back from peak profit to the close was 42.8%, and that on 48 of 159 usable days (30%) a position that had been in profit at some point ended the day at a loss. That means "end-of-day profit" and "peak profit" are very different yardsticks. If you want the desk to manage for end-of-day results, say so. If you want it to take profit on a trigger before the close, say so. A desk cannot read your mind on this, and the choice changes the whole design.

Broker, permissions and hours: say who holds the keys
The fifth row is where many briefs are thinnest and where a short paragraph does the most good.

Name your broker and your account type. A desk needs to know whether its tools connect to that broker at all. Not every platform supports every broker, and a good desk will say so early rather than discover it after you have signed.

State where orders are to be placed. The cleanest arrangement is your own broker account, in your own client code, so every order shows in your own app and you can always see the truth. If the desk proposes anything else, such as a pooled arrangement or an account in its name, you should hear that in the first call and ask about the legal status of the entity before going further. The regulator's website, sebi.gov.in, is the place to check what a given entity is registered for.

Hours, and what happens outside them
Markets are open for a fixed window, but your system and your attention have their own hours. State both.

Give the window in which entries are allowed. Many expiry-day sellers enter early, and the study above anchored its straddle at 09:20 for a reason: the early session is when the premium is richest and when most of the day's decay and re-pricing are still to come. The same study found that the average ATM straddle premium fell from 100% of its 09:16 level to about 88.6% by 11:16, then moved sideways between roughly 88% and 93% for the rest of the afternoon, because the strike is fixed while spot keeps walking. So if you plan entries late in the day, tell the desk, since the premium behaves differently then.

Give the time by which everything must be flat. That could be 15:15, 15:20, or a time before the close that suits your own nerves. And state what should happen to any position still open at that time: closed at market, or closed by a limit with a fallback.

Then cover the less obvious hours. When will you be reachable? If the desk needs a decision at 11:30 on an expiry day, will you pick up? If not, who can decide in your place? The answer "I will be in a meeting from 11 to 1 and cannot be reached" is perfectly fine, as long as the desk knows it and the brief says what the system does in that case, such as pause or reduce.

What to ask back, and what a good reply looks like
A brief works in both directions. Once the desk has your page, you have the right to a clear reply. Here is what to look for.

A good desk replies in its own words to each row. It tells you which parts of your page it can meet exactly, which parts need changing and why, and which parts it will not do. It answers the risk row with numbers. It is open about costs, including brokerage, exchange charges and taxes, and tells you how they affect a result that looks fine before costs.

It tells you what history it relied on. If it shows a backtest, ask for the period, the number of sessions, the entry and exit rules, the slippage assumed, and how it handled days when data was missing. A result without those details is a picture, not evidence. Sample size matters too. In the study referred to here, there were 163 sessions across two indexes, about 80 for each. That is enough to see broad patterns and not enough to say anything firm about a narrow slice such as one VIX band with eleven days in it.

It says what could go wrong, in specific terms. Not "markets carry risk", but "on a gap through the stop, the loss can exceed the planned amount, and here is how large it was on the worst day in our test". In the study the distance between the expiry close and the max-pain strike had a worst case of 415.4 points, or 0.81% of spot, against a median of 0.083%. The tail is far from the middle, and a good desk knows its own tail.

It does not promise a return. Not a fixed monthly number, not "safe", not "guaranteed". Any of these is a reason to stop.

A few patterns should slow you down. A desk that wants your login credentials rather than a scoped key. A desk that cannot say which legal entity it is. A desk that dismisses your daily loss limit as "too tight" without showing the maths. A desk that pushes you to start before you have seen the written reply. A desk that rewrites your brief into its own template and drops the rows you cared about.

None of these proves bad faith on its own. Together, or when the desk is evasive when you ask about them, they tell you enough.

A sample one-page brief you can copy as a layout
This sample is invented and illustrative. The numbers are not advice for anyone. Change every line to match your own situation.

Purpose. Looking for a systematic approach to weekly-expiry index option selling, with me approving changes.

Capital. Rs 10 lakh dedicated. Margin usage not to exceed 60% at entry. Floor: Rs 8 lakh, at which all activity pauses for review.

Instruments. NIFTY and SENSEX weekly index options. Short premium with defined stops. No overnight positions. No averaging down.

Risk limits. Daily loss limit Rs 10,000. Weekly Rs 25,000. Per-position stop as a premium multiple, to be proposed by the desk and approved by me in writing.

Broker. My own account with a mainstream broker. Scoped API access. I do the daily login. Stops rest at the broker where possible.

Hours. Entries between 09:20 and 10:30. Flat by 15:15. I am reachable by phone except 11:00 to 13:00.

Decision rights. Desk proposes, I approve. I can pause the system at any time. Manual overrides by the desk must be logged and flagged.

Reporting. Order-level log each evening. One-page summary each week.

My experience. Two years of manual expiry-day selling. Past losses came mostly from moving stops.

Exit. I revoke access at the broker. The desk deletes stored credentials and sends me the full log.

That is under 200 words, and it carries more information than most first calls. A desk that has this page can answer in specifics. You can compare two desks on the same page. And in six months, you can read it again and see whether you or the desk drifted.

Frequently Asked Questions
What should I prepare before contacting an algo trading company?

Write one page covering your dedicated capital, the instruments and structures you want, your daily and weekly loss limits in rupees, your broker and how access will be granted, your trading hours, and how you would exit the arrangement. It takes under an hour and saves the first call from becoming a questionnaire.

How much money do I need to start algo trading on NIFTY options?

It depends on the structure, the broker's current margin rules and how much buffer you want. Use your broker's margin calculator for the exact position, then add a buffer so you are not near the limit. Decide your floor and your daily loss limit first, and let those set the size.

Can I give an algo provider my broker login?

It is better not to. A scoped API key issued by the broker is narrower and can be revoked on its own, while a login and authenticator seed is full access. Ask the provider what credential it needs and why, and keep the daily login in your own hands if you can.

What loss limit should I tell an algo desk?

The one you can live with without changing your behaviour. Many traders choose a figure as a small percentage of dedicated capital, then test it against the stop rule and lot size, as in the worked example above. Gaps and fast markets can make a loss larger than the limit, so leave room.

How do I know if an algo desk is legitimate?

Ask for its exact legal entity name, what it is and is not registered for, and where you can verify that. Check the regulator's website. Be wary of any desk that promises returns, asks for your password, or cannot explain where your money and orders go.

Do I need a backtest before I speak to a desk?

No. You need your own limits first. A backtest is something you ask the desk to explain, with the period, the sample size, the entry and exit rules and the slippage assumed. A result with no such detail is not worth much.

Conclusion
A good first call starts before the call. One page, with capital, instruments, risk limits, broker, hours, decision rights and exit, lets a desk answer in specifics and lets you compare it fairly against the next one. The rows that matter most are the ones people skip: the floor, the loss limits in rupees, who holds the keys, and what happens outside market hours.

Write the page, keep it short, and keep it honest about your own habits. Then use it to test what comes back. If a desk's reply fits your page and explains its tail risks plainly, you have a basis to talk further. If it does not, you have lost fifteen minutes instead of an hour.

Algo trading carries a real risk of loss, and expiry-day option selling can lose more than the premium collected. EliteAlgo is analytical software and not an investment adviser, and nothing in this article is investment advice. All figures in worked examples and in the sample brief are illustrative.

Top comments (0)