Working notes from
a live desk.
Market structure, risk arithmetic, and post-mortems when our own systems fail. Written by the people who build and run the thing — not by a content team.
The drawdown you didn't budget for
Everyone quotes maximum drawdown as a depth. Depth is the easy half. The number that actually ends subscriptions is how long you spent underwater — and almost nobody publishes it.
Why your backtest fills and your broker doesn't
A limit order at the touch is not a fill. What the queue actually does to a strategy that assumed otherwise.
41 minutes of stale ticks on 14 May
A feed handler held a socket open after the exchange stopped sending. Nobody's risk limits fired, because nothing looked wrong.
Reading a pool record like an auditor
Nine places a published track record can flatter itself, and the field on our own record page that catches each one.
What SEBI's ODR rollout changes for advisory platforms
Online dispute resolution is now a real escalation rung. Here is what we had to rebuild to sit correctly beside it.
Kill switches that actually stop things
A brake that only stops new orders is not a brake. What ours cancels, what it cannot, and why we say so out loud.
The cost of a 1% slippage assumption
Run the same pool through four cost models. Three of them turn a winning record into a losing one.
Sharpe is not a safety rating
Two pools, identical Sharpe, wildly different ways of losing your money. Sortino and Calmar say what Sharpe won't.
We double-sent 74 orders. Here is the exact minute
An idempotency key that wasn't. Every affected account, what it cost, and what we paid back.
Replaying six years of ticks in ninety seconds
How the backtest engine is laid out, why it is columnar, and the one shortcut we refuse to take.
One note a week, on the Sunday before the open.
What changed on the platform, what broke, and one piece of market arithmetic worth knowing. No trading calls — we are not allowed to give them, and would not anyway.