Why I Went Back to Watching Every Single Trade

For a while, the algo ran the way most people assume an algo should run: unattended. It sat on a platform that chewed through a huge stack of 1-minute data across NIFTY, BankNifty, and Sensex, found its setups, fired its entries, and closed its exits — and I checked in the way you check in on something you trust. End of day. P&L, done, move on.

This July, I stopped doing that.

The decision

Nothing broke. That's the part worth saying clearly, because it would be a much easier story if I could tell you the system misfired and I had to step in and rescue it. That's not what happened. The system was doing exactly what it was built to do. What changed was me — I decided I wanted to actually watch it. Every entry. Every exit. Every single session, in real time, for the whole month.

Partly that was curiosity. When you build something that trades on its own, there's a specific kind of distance that creeps in — you know the logic on paper, you've backtested it, you trust the numbers, but you stop having a felt sense of what it's actually doing minute to minute. You start relating to it as a P&L line instead of a trading decision. I wanted that felt sense back.

And partly it was accountability. It's easy to let an automated system be a black box you defer to — "the algo did it" is a comfortable sentence. But the algo doesn't own the outcome. I do. If it's going to trade real capital, I wanted to be able to say, honestly, that I watched every decision it made and would have made the same call myself, not that I was hoping the backtest held up while I wasn't looking.

What a month of actually watching looks like

Twenty-three trading days. Every entry, every exit, sat through in real time — no glancing at the P&L at 3:30 and calling it a day.

18 wins, 5 losses. A 78% hit rate across the month, on a strategy running 1 lot each in NIFTY, BankNifty, and Sensex options.

+₹37,252 net, on roughly ₹4.2 lakh of capital required to hold those three positions — call it a touch under 9% for the month.

Those numbers aren't the point of this piece, though. The algo scorecard is its own running log, elsewhere. The point here is what watching, rather than just checking in on, actually taught me.

What I actually learned by watching

The losses are where the discipline lives, not the wins. A winning trade doesn't test anything — the system said go, it went, it worked, nothing to evaluate. The five losing days are where it mattered whether I actually understood why the system was doing what it was doing, because that's the only way to tell the difference between "this is a losing trade the strategy is supposed to take sometimes" and "something about market structure has actually changed and the strategy needs to be rethought." You can't tell those apart from a P&L number alone. You can only tell them apart by watching the order flow, the level being defended or lost, the volume behind the move — the same structural reads that show up in the daily recap here, just applied in real time instead of after the close.

Automation is a tool, not a substitute for understanding the tool. I didn't stop trusting the system this month. I trust it more, actually, because I watched it earn that trust in real conditions rather than just in a backtest. But "I built a system that works" and "I understand why it works well enough to know when it'll stop working" are two different claims, and only one of them survives contact with a genuinely bad week. Watching every trade is how you keep making the second claim honestly.

Consistency isn't a personality trait, it's a practice. Nobody watches 23 straight sessions because they feel like it on day 23. You watch because you decided to before the month started, and then you just... show up. That's the whole secret, and it's a boring one, which is probably why it works.

What happens next

This isn't a one-month experiment I'm closing out. I'm planning to keep watching every entry and exit live for at least the next three months — possibly longer, and I'm not putting a hard deadline on it, because the moment you set an artificial finish line is the moment you start watching to get through it rather than watching to actually learn something.

I want to be upfront about why, because it would be easy to assume the plan is "watch it long enough to prove it works, then package it up and sell it." That's not the intent. The algo isn't something I'm building toward commercializing for its own sake, and if that ever happens, it's a long way off and it won't happen until I've watched it through enough different kinds of months — not just good ones — to actually stand behind it. The real reason I'm continuing this is simpler and less exciting: it's the discipline itself I'm after. Watching every trade, every day, for months on end, is its own practice — the same way showing up to read the tape every session is the practice behind the daily recap. The scorecard is a side effect of that practice, not the goal of it.

Drop me a note if you want to talk about any of this — I'm always up for a conversation over coffee, while the algo does its (now closely watched) thing.