If you were watching NSE's end-of-day data on 10 July 2026, one particular row in the ETF data file would have stopped you cold: PSUBANK (Kotak PSU Bank ETF) closed 89.6% down on that day. It is highly unusual for a stock or an ETF to lose 90% of its value in a single session outside of a genuine catastrophe. Therefore, before flagging this move as a "top mover" anywhere on this site, we went and checked what actually happened.
What we found is that it was not a crash. It was almost certainly a unit split, and the raw NSE data just did not say so out loud.
What the numbers show
Here is a snapshot of PSUBANK's close price for the days before and after 10 July:
| Date | Close (Rs) | Day change |
|---|---|---|
| 8 Jul | 810.59 | -2.71% |
| 9 Jul | 822.75 | +1.50% |
| 10 Jul | 85.31 | -89.63% |
| 13 Jul | 84.65 | -0.77% |
| 14 Jul | 83.36 | -1.52% |
| 15 Jul | 83.90 | +0.65% |
| 20 Jul | 86.47 | +2.77% |
Two things jump out. First, the drop is observed in one session exactly — 9 July to 10 July — not a gradual one. Second, the price never recovered. Ten trading sessions later, it was still in the Rs 82–86 band. A genuine 90% crash from bad news is occasionally followed by some bounce while a data error self-corrects the next day. Neither happened here. The new price level simply became the new normal.
Dividing it out, Rs 822.75 (close on 9 July) into Rs 85.31 works out to almost exactly 10x — the signature of a 10-for-1 split. Each existing unit of the ETF was likely broken into roughly ten and hence, the same amount of money now buys ten times as many units at a tenth of the price. Total value held by an investor would not have changed. Only the per-unit price and unit count would.
Two more data points backed this up.
- Trading activity re-rated permanently higher. Before 10 July, PSUBANK typically traded 8,000–130,000 units a day. By 15–20 July it was regularly trading 1.25 million to 2.2 million units a day. This is consistent with the logic that far more (and cheaper) units are now in circulation - exactly what you'd expect after a split.
- It was completely isolated. We checked every other ETF on that day's file — 325 other symbols — and not one showed anything close to this kind of move. That rules out a one-off data glitch or a parsing bug affecting the whole day's file. Whatever happened, it happened to PSUBANK specifically.
Why our own site briefly got it wrong
NSE's bhavcopy files do carry a "corporate action indicator" column to flag these kinds of events — a split, bonus issue, or face-value change — for a handful of instrument types. As it turns out, ETFs are not one of them and NSE does not populate that flag for ETF filings the way it does for corporate bonds and SME stocks. Hence, there is no clean signal in the data to say "ignore this session's % change, it is a corporate action."
Our pipeline computes daily and weekly % change purely from consecutive closing prices, which is the right approach 99.9% of the time. But it means a split like this one briefly showed up as a headline "Top Loser" on both the daily ETF card and the weekly rollup, sitting next to genuine market movers. That is misleading, even if every other number on the page (turnover, trade counts, 52-week range) stayed completely correct throughout.
What we changed
Rather than trying to guess "was this a split?" from price data alone — which is possible but fragile — we made a narrower, more defensible fix: implausible single-day and single-week moves (anything beyond ±50%) are now excluded from "top mover" style headlines across both the daily and weekly cards. A move that large is essentially never a real trading outcome. Therefore it is honest to leave that spot for the next real mover than to publish a number that would mislead anyone who does not dig into the split math above.
To be clear about what this does not touch - PSUBANK's turnover and trade count for the day are still counted correctly in every aggregate total, and it still shows up everywhere in the raw data and downloads exactly as NSE reported it. Only the "here are today's biggest movers" spotlight skips numbers that could not possibly be real.
The more complete fix — actually capturing NSE's corporate-action flag for ETFs the way we already do for bonds and SME stocks — is on our list. But even with that in place, a human sanity check like this one is a useful habit. When a single number looks impossible, it usually is, and the interesting question is why.