Exchange approval answers a compliance question: may this algorithm operate within the exchange’s rules? Trader approval answers a separate risk question: should this specific order enter this specific account under current conditions? India’s shift toward compliant, homegrown algo trading, represented by firms such as Strykex, makes that distinction harder to ignore.
Consider an illustrative trader named Arjun, a Bengaluru software tester who trades a modest account before work. At 9:14 one morning, he is holding a cooling cup of tea while an approved algorithm prepares an order. The setup meets its coded rules, but Arjun already has two positions exposed to the same broad market move.
If the new order executes and all three positions fall together, his planned daily risk limit may fail in practice. The algorithm has passed its permitted checks. Arjun still has seconds to decide whether the portfolio can absorb the trade.
Compliance defines permission, not suitability
Exchange approval matters. It can establish that an algorithm, broker connection, or order process follows specified technical and regulatory requirements. Those controls can reduce certain operational risks and make accountability easier to trace.
They cannot know every trader’s circumstances.
An approved system may not know that the money in an account is reserved for a near-term expense. It may not recognize that three individually valid positions express the same underlying bet. It may follow its entry rule even when the trader has reached a personal loss limit or lacks confidence in the data feeding the decision.
This is the limit of treating compliance as a complete safety verdict. Compliance sets a boundary for what a system may do. The trader still needs a boundary for what the account should do.
For Arjun, the crucial question is not “Was this algorithm approved?” It is “What happens to my total exposure if I approve this order?”
Automation can apply rules while missing context
Algorithms are good at consistency. They can evaluate the same conditions repeatedly without fatigue, hesitation, or revenge trading. That discipline has value, especially for traders who change their rules after a loss.
Consistency also creates a specific danger: a rule can be applied correctly in the wrong context.
Suppose an algorithm calculates a position using a fixed percentage of account equity. The arithmetic may be correct while the account value is stale, the stop distance has widened, or another open position has already consumed part of the risk budget. A valid signal can still produce an unsuitable order.
This is why visible inputs matter more than a confident output. Before approval, a trader should be able to inspect the proposed entry, stop, position size, account value used, current exposure, and the conditions that invalidate the trade. If any input cannot be verified, confidence from the model adds little.
The same principle appears in Daniel’s $15,000 order, where one click could triple his risk. The order’s existence does not settle whether it belongs in the account. Approval requires examining the effect of the order after execution.
Human control needs a real rejection path
A human approval gate only works when rejection has no penalty built into the interface or process. The trader needs enough information and time to say no, even when every signal rule has been met.
That means an approval step should expose more than a green button. It should show what the algorithm observed, what assumptions shaped the order, how much capital is at risk, and where the proposal conflicts with account-level constraints. It should also allow the trader to reject the order without the system quietly submitting a replacement.
Arjun pauses the order. He notices that the three positions depend on similar market conditions and rejects the third trade. The price later moves higher, so the rejected order would have made money.
That does not make the rejection wrong.
Decision quality cannot be judged from one outcome. A disciplined rejection may avoid profit today and prevent an excessive loss next week. The useful record is the reasoning available at the moment of approval, followed by a review of whether the process respected the trader’s limits.
A trading journal should therefore capture rejected signals as carefully as executed ones. Record the proposed size, total exposure, rejection reason, and later outcome. Over time, that visible track record can reveal whether the approval gate improves discipline or merely adds an extra click.
Build the control around the account
India’s homegrown algo-trading shift points toward a more structured relationship between exchanges, trading systems, and traders. The next challenge is to preserve human judgment after the technical and compliance boxes have been checked.
Start with account-level rules that automation cannot silently override:
- Set a maximum loss per trade and per day.
- Count correlated positions as shared exposure.
- Recalculate size when entry or stop distance changes.
- Expire queued approvals when market conditions become stale.
- Require a reason for every manual override and rejection.
Expiry deserves particular attention. An order reviewed before the market closes may no longer reflect the next session’s price, volatility, or account exposure. An overnight approval should have a defined expiration point, followed by a fresh review of the inputs.
The morning after Arjun’s rejection, his journal contains one missed gain and one intact risk rule. He does not rewrite the rule to fit the chart. He checks whether the same decision process should govern the next queued order.
That is the useful boundary: the exchange decides which systems may participate, the algorithm proposes an action, and the trader retains the final decision.
Educational content, not financial advice.
Comments
No comments yet.