Two days with the same gamma — the two measurements that separated a trend from a range

After the close on September 22, 2026, this is the question I wrote down:

"Yesterday and today, the gamma readings were nearly the same. Yesterday the index rose more than 100 points. Today it chopped inside 20. What was different?"

Both sessions were in strong positive gamma — the regime where dealer hedging tends to dampen moves. By the textbook, both should have been quiet. One was. The other climbed like a staircase.

This post is a measured record of that difference. The short version: what separated the two days was not the gamma level. It was two other things.

1. How persistent the call-side buying flow was
2. How thick the put gamma sat right next to the current price

⚠️ The sample is two days. This is not a rule. It is a record that pins down the definitions, so the next time a day like this comes along I can measure it with the same ruler.

1. First — how the two days ended

	Mon 9/21	Tue 9/22
Prior close	7,650.50	7,764.70
Open	~7,697 (gap up)	7,771.49
Low	7,691.69	7,756.26
High	7,778.99	7,782.19
Close	7,764.70	7,764.77
Day range (high − low)	87.3pt	25.9pt
Change vs. prior close	+114.2pt	+0.07pt

On 9/21 the index rose from the open into the afternoon without making a meaningfully lower low. Between 06:30 and 12:30, 6 of the 25 fifteen-minute bars closed down, and every one of those had a body under 3 points.

9/22 was the next day. It traded inside 25.9 points all session and closed 0.07 points from the prior close.

2. The level did not separate them

Here is net GEX (the sum of dealer gamma exposure, in the app's own units) as calculated by the app I use, from 06:30 to 12:54 PT. I left out the last few minutes before the close, when same-day options settle and the reading jumps around minute to minute.

	9/21 (trend)	9/22 (range)
Net GEX high	+47,706	+44,070
Net GEX average	+39,922	+28,905
Gamma regime all day	Positive	Positive

The highs are within 8% of each other. The average was actually 28% lower on 9/22, the day that ended as a range. If "more gamma means quieter" were the whole story, 9/21 should have been the calmer day.

This did not surprise me. When I regressed daily range on net GEX in the app, the r² came out around 0.19 (my records, 2026-07). In other words, the GEX level explains roughly a fifth of why ranges differ. My bar for adopting a variable is r² of 0.3 or higher, so I had already decided not to use the GEX level as a stand-alone signal. These two days were one more case consistent with that decision.

So I measured something other than the level.

3. First difference — the persistence of call buying

From Unusual Whales' Market Tide I took market-wide net call premium (premium traded at the ask minus premium traded at the bid; "net calls" below) in 5-minute bars. The window on both days was the regular session, 06:30–13:00 Pacific — 79 bars counting both ends.

I cared less about how big the net-call number got than about whether it held where it got to, so I measured two things:

• New-high share — the fraction of 5-minute bars that set a new intraday high for net calls
• Max drawdown from peak — the largest pullback from the running high, divided by the day's high

	9/21 (trend)	9/22 (range)
Net calls, start → end	+$53M → +$617M	+$12M → +$245M
Day's high	+$639M	+$247M
New-high share	43/79 = 54%	17/79 = 22%
Max drawdown from peak	$60M (9% of peak)	$108M (44% of peak)
Bars lower than the prior bar	23/78 (29%)	37/78 (47%)

On 9/21, net calls almost never gave back ground once they had taken it. More than half the bars were new highs, and the deepest pullback was 9% of the peak.

Net calls on 9/22 were positive all day too — this was not a day without call buying. But they pulled back often (47% of bars were lower than the bar before) and at one point gave back 44% of the peak.

🔑 Both days leaned toward call buying. What differed was persistence, more than size.

During the session I described this as "the net-call line kept making higher highs and never rolled over." The two numbers above are that impression turned into something I can measure again.

4. Second difference — how thick the put gamma was next to price

The second difference was the larger one.

I compared put-side dollar gamma by strike, as the app computes it, at the same clock times on both days. Adding up every put strike within 30 points of the current price gives the "±30pt put total."

Time (PT)		9/21 (trend)	9/22 (range)	Ratio
09:00	SPX	7,737.22	7,761.29
	Heaviest put strike	7,725 · −$412M	7,755 · −$3,491M	8.5×
	±30pt put total	−$2,080M	−$17,548M	8.4×
11:00	SPX	7,759.76	7,771.36
	Heaviest put strike	7,750 · −$424M	7,775 · −$2,539M	6.0×
	±30pt put total	−$1,125M	−$14,360M	12.8×

On 9/21 the space below price was nearly empty. Even the heaviest put strike was only around −$0.4B.

On 9/22, between $2.3B and $3.5B of put gamma sat right on either side of price, around 7,765, 7,770 and 7,775. Those strikes showed up as thick bars on my screen all session.

Here is how I read it. When that much open interest sits at strikes right next to price, dealer deltas change fast as price moves between them, so hedging flows are large. If dealers are long those options, they sell into strength and buy into weakness, and price stays boxed in. If they are short them, the same flows push price further. My app's sign convention treats puts as dealer-short, so this data cannot tell me which it was.

⚠️ What I confirmed stops at two things: the thickness differed by 8 to 13 times, and on 9/22 price stayed near those strikes all day. On 9/21 the area around price was empty, and there was nothing there to hold it.

5. Net puts — same direction both days

I also looked at net put premium ("net puts").

(For reference: when net puts rise, put buyers have the upper hand, and the bet on downside is growing. When net puts fall, the side closing out put positions has the upper hand, and the downside push is weakening.)

	9/21 (trend)	9/22 (range)
Net puts, start → end	+$9M → −$195M	−$14M → −$96M
Day's low	−$205M	−$109M
Bars that set a new low	37	19

Puts were on the selling side both days. The size differed by about two times, but the direction was the same, so net-put direction alone does not tell the two days apart.

I have been collecting observations that when net put premium keeps falling instead of rising, the push to break lower fades. 9/22 fit that. In the morning price dropped to its low of 7,756.26, within 1.3 points of that morning's put wall at 7,755 — but net puts did not build, and price came back up.

What these two days added was the next sentence:

Having no push lower did not mean price would go higher.

On 9/22 the downside was blocked and the upside never got going. The result was a close 0.07 points from the day before.

6. The definitions I have fixed for next time

Setting thresholds from two days would mean fitting the numbers to the outcome. So I fixed what I measure and how, and left the cutoffs blank.

	Definition (will not change later)
Window	06:30–13:00 PT, 5-minute bars (79 bars counting both ends; the post-13:00 settlement bars are excluded)
A-1 New-high share	Net-call bars that set a new intraday high (excluding the first bar) ÷ total bars
A-2 Max drawdown from peak	Largest (running high − current value) ÷ the day's high
B Put thickness	Sum of put dollar gamma within ±30pt of price, at 09:00 and 11:00
Outcome	The day's high−low range ÷ that day's expected move (from the prior day's VIX1D close) — not yet calculated in the table below

There are two hypotheses:

• A — On trend days, net calls pull back less from their peak and set new highs more often
• B — On trend days, put thickness around price is thin (empty). On range days it is thick (boxing price in)

I drop a hypothesis if I see this:

• Two or more days where the net-call drawdown was small but the day ended as a range
• Two or more days where put thickness was thin but the day ended as a range

The sample so far is these two rows:

Date	Outcome	Net-call drawdown	±30pt put thickness (11:00)
2026-09-21	Trend · range 87.3pt	9%	−$1.1B
2026-09-22	Range · 25.9pt	44%	−$14.4B

The next time the index moves around 100 points in a day, I add a row. The cutoffs get set after at least 10 days are in.

📎 The five things I watch to spot a trend day during the session are in section 7 of On a trend day, I put credit spreads on the side price has already left. Measure A here is a way to score item 1 in that table (fuel on the breakout side) by how persistent it was. B is new — that table doesn't have it.

7. What I don't know yet

• The sample is two days. They are also consecutive, and 9/22 carries a second condition — "the day after a big rally, with no major news." I cannot yet say whether that condition or the put thickness produced the result
• I don't know which of A and B shows up first. B already differed by 8 times at 09:00. A is only final at the end of the day. To be usable intraday, I need to check whether "A through the morning" separates the days too
• This data cannot say whether the flow led price or followed it. I only confirmed that they moved together at the same times
• Net calls and net puts are market-wide. Isolating SPX options alone could give a different result

These two days showed, in numbers, that the same gamma level can end in opposite outcomes. The rest goes into the table as the sample grows.

---

This post is a record of my own observations and measurements. It is not investment advice and not a recommendation to trade any security or strategy. The sample here is two days. The prices and readings quoted are from September 21–22, 2026. Net calls and net puts are from Unusual Whales Market Tide; net GEX and put dollar gamma are my app's own calculations from Unusual Whales raw data. All investment decisions and their outcomes are your own.