---
title: "I Tested My Own Finding And It Did Not | The Tanmay Edge"
description: "sion and it comes apart when size turns up. That is a depth problem, not a design problem. METHOD NOTE. The reversal test no longer needs the delivery file."
source_url: https://www.rupeecase.com/podcast/episodes/2026-09-03
episode: 117
season: 2
date: 2026-09-03
duration_seconds: 520
audio: https://content.rss.com/episodes/336728/3116800/tk-s-pre-market-radar/2026_09_03_03_24_45_e68071b8-5608-4687-8628-64e3845d501e.mp3
grade: 4/5
generated: instiv2-tools/build-podcast.mjs (do not rebuild from the HTML)
---

# I tested my own finding and it did not repeat

The Tanmay Edge, Season 2, EP117, Thursday 3 Sep 2026, 8:40. The EP116 finding did not replicate on a control day: correlation minus 0.087, slope minus 0.234

## Scorecard

Grade 4/5 (80 percent): 4 hit, 0 partial, 0 mechanical, 1 miss. Graded the next trading day, 4 Sep 2026, against NSE settlement.

The call: The EP116 finding did not replicate on a control day correlation minus 0.087, slope minus 0.234 Against minus 0.712 and slope minus 1.037 on the MSCI rebalance pair 44 of the 210 auction stocks printed more than 2 percent from their own average on rebalance day On a normal day three of 210 did, and the non auction control was worse at 4.2 percent On an ordinary Wednesday the auction beat everything else in the market On the rebalance day it was four times worse Verdict the auction is not broken, it is thin, a depth problem not a design problem Nifty high 23,914 equalled its close, and the Sensex high 76,570 equalled its close Three indices printed the

Graded Fri 04 Sep against the Thu close 23,873. TK: the actual expiry got moved in the cash session, but the rest of the day worked out superbly. The fade call was the day: a gap up into 24,000 with max pain below it was flagged as a setup for a fade and not a follow through, and the tape faded. Do not chase the open held. The magnet is behind us framing was right, and the band contained the session. The miss was the settlement itself, moved again by the closing auction, which no intraday plan can price. Four on five.

## Chapters

### 1. Cold open: the replication failed, and that is the story (about 0:00, estimated)

Yesterday I told you the closing auction session mismarks stocks and the market hands it back the next day. The correlation was minus 0.712, the slope minus 1.037, 69 out of 70 handed it back on Tuesday. I also told you on air that it was one pair of days, and that Monday was the MSCI rebalance. So I ran the next pair, an ordinary Tuesday into Wednesday. Correlation minus 0.087, slope minus 0.234. Nothing. That could have gone in a drawer and none of us would have known.

### 2. Why it did not repeat: thin, not broken (about 0:45, estimated)

The reason it did not repeat is more useful than the finding was. On the rebalance day, 44 of the 210 stocks in the auction printed more than 2 percent away from their own average price. On a normal day, three did, out of 210. And in the control group, the stocks that are not in the auction at all, 4.2 percent of them missed by that much. So on an ordinary Wednesday, the auction was better than everything else in the market. On the rebalance day, it was four times worse. So the honest answer, the one that survives a control day, is this. The auction is not broken. It is still thin. It handles an ordinary session perfectly well and it comes apart the moment size turns up. That is the depth problem, not a design problem. And it is exactly what the regulator's own paper worried about. I would rather give you the version that survives the second test than one that sounded better on Wednesday.

### 3. And yet: three indices printed the high as the close (about 2:10, estimated)

Look at yesterday's close. The Nifty's high for the day was 23,914 and its close was 23,914. The Sensex high was 76,570 and the close was 76,570. Three indices printed the day high as the closing print, and the continuous market never once traded above it. On the auction chart, the indicative was near 23,930 at twenty past three, and 24,042 thirty seconds later. Then nine minutes of decay to 23,914. That is 128 points, the day after 117 points. So both things are true at once. The marks are small on a normal day, and the window is still wide.

### 4. The tape: 24,000 finally went (about 3:02, estimated)

The Nifty closed 23,914, down 141 points. The open was 23,858, a gap down of 198 points, and the low was 23,786. And 24,000 finally went. It had broken intraday and closed above it three sessions running; yesterday it gapped straight through it at the open and never came back. The Sensex closed 76,570, down 380 points. India VIX is still under 12, and I am not quoting the decimal because the two feeds disagree. What did the damage was not banks, for once. Auto fell 1.79 percent with one of the 14 up. IT fell 1.25 percent with zero of ten up. The Nifty Bank was minus 0.41 with 7 up and 7 down. Breadth was bad for a fourth session, 14 up against 36 down in the Nifty 50, and worse in the broader market.

### 5. The board: the put book moved 200 points in one session (about 4:10, estimated)

Calls 1,466 lakh, puts 1,069 lakh. Max pain 23,950, sitting 35 points above the spot. And the put book moved 200 points in one session. 1.84 lakh puts were added at 23,900 and below, and 39 lakh were pulled off from 24,000 and above. They rebuilt the support 200 points lower. Support is now at 23,800 and 23,900, then 23,600. Resistance is at 24,200, the heaviest at 1 crore calls, then 24,000 at 88 lakh. The gamma flip happens at 23,900 to 23,950 and the spot closed 14 points above it, so we are sitting exactly on it, with a band of 23,800 to 24,050. And the futures held what the cash did not: the September contract closed at 24,000 and the basis widened to plus 86 points. The selling was in cash, not in futures.

### 6. The thing that does not add up (about 5:18, estimated)

The pros flipped their call book back to net long, the fifth flip in a week, so stop reading it as a view. The foreigners added another 7,000 futures short and sold 51,000 more calls. The crowd covered 44,000 naked puts. Now the thing that does not add up. The foreigners bought 6,688 crore in cash, the biggest single day buy in months. The locals bought 2,813. So 9,501 crore of institutional buying, and the index fell 141 points. Somebody was selling into all of it, and this data does not name them. It only tells you the buyers were not in charge. And the same foreigners bought 2,600 crore of stock futures and added 41,679 stock futures longs while shorting the index harder. That is I like these companies, I do not like the broader market, executed at size.

### 7. The plan: the magnet is behind us (about 6:31, estimated)

GIFT Nifty is at 24,102, up 0.57 percent, pointing to an open near 24,030 to 24,050, a gap up of around 135 points. Asia has steadied, Korea is up 1.26, the Nasdaq closed up 0.45. Gold reversed hard to 4,417 from 4,300, the yen strengthened 1.29 percent, America is at 4.78 against a 52 week high of 4.82, India at 6.98 with the 30 year at 7.58. Stocks bouncing, the yen up more than a percent, the VIX at 17, and every major bond market sitting on its high yield. Those four do not usually belong in the same paragraph. Episode 116 gets a 2.5 out of 5: the conditional was right, below 23,950 it trended, but the open was 198 points lower than I called. So the plan. A gap up near 24,030 to 24,050, straight back into the zone the market spent all of yesterday underneath. The first question is whether it holds 24,000 on the way back up, and the board says the writers have stopped defending it. Resistance is 24,000, then 24,100, then 24,200. Max pain stands at 23,950, below the likely open, so the magnet is behind us, not in front of us. And the trap: a gap up into 24,000 with max pain below it and the put support 200 points lower is the setup for a fade, not a follow through. Do not chase the open. The Sensex expiry is today with max pain at the 76,500 to 76,600 level and the spot 70 points above it. Spreads only, and do not judge the day before 3:15.

## Key takeaways

- The EP116 finding did not replicate on a control day: correlation minus 0.087, slope minus 0.234
- Against minus 0.712 and slope minus 1.037 on the MSCI rebalance pair
- 44 of the 210 auction stocks printed more than 2 percent from their own average on rebalance day
- On a normal day three of 210 did, and the non-auction control was worse at 4.2 percent
- On an ordinary Wednesday the auction beat everything else in the market
- On the rebalance day it was four times worse
- Verdict: the auction is not broken, it is thin, a depth problem not a design problem
- Nifty high 23,914 equalled its close, and the Sensex high 76,570 equalled its close
- Three indices printed the day high as the closing print
- The indicative ran 23,930 to 24,042 in thirty seconds, then decayed to 23,914
- 128 points of auction swing, the day after 117 points
- Nifty closed 23,914, down 141, after a 198 point gap down open
- 24,000 finally went after three sessions of closing above it
- Auto fell 1.79 percent with one of 14 up, IT fell 1.25 with zero of ten up
- Breadth bad for a fourth session, 14 up against 36 down in the Nifty 50
- 1.84 lakh puts added at 23,900 and below, 39 lakh pulled off 24,000 and above
- The put book moved its support 200 points lower in one session
- Max pain 23,950, sitting 35 points above the spot
- September contract closed at 24,000 with the basis widened to plus 86
- Pros flipped the call book back to net long, a fifth flip in a week
- FIIs bought 6,688 crore cash, the biggest single day buy in months
- 9,501 crore of institutional buying and the index still fell 141 points
- FIIs added 41,679 stock futures longs while shorting the index harder
- Plan on record: magnet behind the open, a gap up into 24,000 is a fade setup not a follow through

## Quotes

> Correlation minus 0.087, slope minus 0.234. Nothing.
>
> The control day result, run against the same test that produced minus 0.712 on the rebalance pair.

> That could have gone in a drawer and none of us would have known.
>
> On choosing to lead with a failed replication of his own finding.

> The reason it did not repeat is more useful than the finding was.
>
> On what a control day actually bought: an explanation rather than a headline.

> The auction is not broken. It is still thin.
>
> The verdict that survives the second test, on a depth problem rather than a design problem.

> I would rather give you the version that survives the second test than one that sounded better on Wednesday.
>
> On revising a finding in public a day after making it.

> The marks are small on a normal day, and the window is still wide.
>
> On a 128 point auction swing arriving the day after a 117 point one.

> 9,501 crore of institutional buying, and the index fell 141 points.
>
> On the foreigners' biggest single day cash buy in months landing on a down day.

> Somebody was selling into all of it, and this data does not name them.
>
> Stating the limit of what the participant file can show.

## Transcript

A very good morning, guys. Yesterday I told you the closing auction session mismarks stocks and the market hands it back the next day. 037, 69 out of 70 handed it back on Tuesday. I also told you on air that it was one pair of days, and that Monday was the MSCI rebalance. So I ran the next pair, an ordinary Tuesday into Wednesday. 234. Nothing. This is The Tanmay Edge. You are listening to episode 117. I am Tanmay Kurtkoti. Let's go. That could have gone in a drawer and none of us would have known.

But the reason it did not repeat is more useful than the finding was. On the rebalance day, 44 of the 210 stocks in the auction printed more than 2 percent away from their own average price. On a normal day, three did, out of 210. 2 percent of them missed by that much. So on an ordinary Wednesday, the auction was better than everything else in the market. On the rebalance day, it was four times worse. So the honest answer, the one that survives a control day, is this.

The auction is not broken. It is still thin. It handles an ordinary session perfectly well and it comes apart the moment size turns up. That is the depth problem, not a design problem. And it is exactly what the regulator's own paper worried about. I would rather give you the version that survives the second test than one that sounded better on Wednesday. And yet, look at yesterday's close. The Nifty's high for the day was 23,914 and its close was 23,914. The Sensex high was 76,570 and the close was 76,570.

Three indices printed the day high as the closing print, and the continuous market never once traded above it. On the auction chart, the indicative was near 23,930 at twenty past three, and 24,042 thirty seconds later. Then nine minutes of decay to 23,914. That is 128 points, the day after 117 points. So both things are true at once. The marks are small on a normal day, and the window is still wide. The tape. The Nifty closed 23,914, down 141 points. The open was 23,858, a gap down of 198 points, and the low was 23,786.

And 24,000 finally went. It had broken intraday and closed above it three sessions running. Yesterday it gapped straight through it at the open and never came back. The Sensex closed 76,570, down 380 points. India VIX is still under 12, and I am not quoting the decimal because the two feeds disagree. What did the damage was not banks, for once. 79 percent yesterday with one of the 14 up. 25 percent with zero of ten up. 41 with 7 up and 7 down.

Breadth was bad for a fourth session, 14 up against 36 down in the Nifty 50, and worse in the broader market. Now the board, because that is the real news. Calls 1,466 lakh, puts 1,069 lakh. Max pain 23,950, sitting 35 points above the spot. And the put book moved 200 points in one session. 84 lakh puts were added at 23,900 and below, and 39 lakh were pulled off from 24,000 and above. They rebuilt the support 200 points lower. Support is now at 23,800 and 23,900, then 23,600.

Resistance is at 24,200, the heaviest at 1 crore calls, then 24,000 at 88 lakh. The gamma flip happens at 23,900 to 23,950 and the spot closed 14 points above it, so we are sitting exactly on it, with a band of 23,800 on the lower end to 24,050 on the upper end. And the futures held what the cash did not. The September contract closed at 24,000 and the basis widened to plus 86 points. The selling was in cash, not in futures. So the positioning.

The pros flipped their call book back to net long, the fifth flip in a week, so stop reading it as a view. The foreigners added another 7,000 futures short and sold 51,000 more calls. The crowd covered 44,000 naked puts. Now the thing that does not add up. The foreigners bought 6,688 crore in cash, the biggest single day buy in months. The locals bought 2,813. So 9,501 crore of institutional buying, and the index fell 141 points. Somebody was selling into all of it, and this data does not name them.

It only tells you the buyers were not in charge. And the same foreigners bought 2,600 crore of stock futures and added 41,679 stock futures longs while shorting the index harder. That is "I like these companies, I do not like the broader market", executed at size. 57 percent, pointing to an open near 24,030 to 24,050, a gap up of around 135 points. 45. 58. Stocks bouncing, the yen up more than a percent, the VIX at 17, and every major bond market sitting on its high yield.

Those four do not usually belong in the same paragraph. 59. The next rebalance is on the 14th. 5 out of 5. The conditional was right, below 23,950 it trended, but the open was 198 points lower than I called, and the whole day traded under that band. So the plan. A gap up near 24,030 to 24,050, straight back into the zone the market spent all of yesterday underneath. The first question is whether it holds 24,000 on the way back up, and the board says the writers have stopped defending it.

Resistance is at 24,000, then 24,100, then 24,200. Max pain stands at 23,950, which is below the likely open, so the magnet is behind us, not in front of us. Support is present at 23,900, 23,800 and 23,600. The band is 23,800 on the lower end to 24,050. Gamma flips at 23,950, and yesterday proved what that means: it opened under the flip and trended for six hours. And the trap. A gap up into 24,000 with max pain below it and the put support 200 points lower is the setup for a fade, not a follow through.

Do not chase the open. Wait for it, watch it, and then go into it. The Sensex expiry is today and max pain stands at the 76,500 to 76,600 level with the spot 70 points above it. Spreads only, and we could see 77,000 as the expiry. And do not judge the day before 3:15. Just trade spreads. Have a great expiry, guys. Yesterday I gave you a finding. Today I gave you the test that broke all of it, and the half that survived is the better half.

This was The Tanmay Edge. If you found value in this, please share it with one trader friend who needs it right now. com. And give it a rating, guys, if you want to add something to the podcast. Thank you, see you all once again tomorrow morning.

## More

- Podcast archive: https://www.rupeecase.com/podcast/
- September 2026 archive: https://www.rupeecase.com/podcast/2026-09/
- Season 2 index: https://www.rupeecase.com/podcast/season-2/
- Track record and method: https://www.rupeecase.com/track-record
- Previous: EP116, https://www.rupeecase.com/podcast/episodes/2026-09-02
- Next: EP118, https://www.rupeecase.com/podcast/episodes/2026-09-04

Market commentary for education only. Not investment advice. RupeeCase is not a SEBI registered investment adviser.