Max pain and the butterfly on expiration day — I close while price passes the center

As an expiration approaches, I watch two things on the same screen. Where max pain is pointing, and where the body of my butterfly sits.

On August expiration those two came apart in a way I did not expect. Max pain landed within 4 points of the close, and that is exactly why my butterfly went to zero — my center sat 110 points below it.

A month later, during September expiration week, I held three butterflies with essentially the same body. I closed one of them the day before expiry. That one came back at a little over three times the debit. The two I did not close expired empty.

Same structure, same body, same week. What separated them was not the forecast. It was whether I closed.

📌 This is a dated observation. SPX · August 21, 2026 (August expiration) and September 16–18, 2026 (September expiration week). Seven butterfly positions across two expirations. The sample is small. I am not turning this into a rule — I am writing down what I decided to do.

0. The conclusion first

This runs long, so here is the short version.

① Max pain being right and my butterfly winning are two different events. On August expiration, max pain called the close to within 4 points. The butterfly I carried into settlement that day went to zero. Max pain answered where the market would finish. My butterfly was where I had bet it would finish. Those two were 110 points apart. Three other positions aimed at the same expiration did the opposite — they followed max pain, and there max pain itself was 218 points wrong.

② So I settled on this: while the center is being crossed, if there is value on the screen, I take it. Expiration settlement is not something to wait for — it is a line not to cross. Cross it and a butterfly becomes either exactly the body or zero. There is no in-between.

1. Thirty seconds on how a butterfly looks at expiry

Skip to section 2 if this is familiar.

A butterfly is three legs. Buy one lower strike, sell two at the middle, buy one higher strike.

7,600  buy 1 call    ← upper wing
7,500  sell 2 calls  ← center (body)
7,400  buy 1 call    ← lower wing

wings 100 pts wide · debit 5.30  →  $530 for one contract

The expiration payoff is a tent.

      value
   100  │           ▲          ← maximum only AT 7,500
        │          ╱ ╲
        │         ╱   ╲
     40 │        ╱     ╲
        │       ╱       ╲
      0 └──────╱─────────╲──────  index
          7,400  7,500  7,600
              ↑           ↑
        zero outside   zero outside

Three things to hold onto.


Max loss	the entire debit — $530 here. You cannot lose more than that
Max profit	wing width − debit = 94.70 ($9,470). But only if settlement prints exactly 7,500
Breakevens	lower wing + debit = 7,405.30 · upper wing − debit = 7,594.70

The capped loss is what makes this structure attractive, and it is also the trap. Since you can only lose the debit, "I'll just hold it to expiry" comes easily. What that choice actually does is the subject of this post.

I have already written about why I buy a butterfly early and how the curve sharpens:

2. August expiration — I followed max pain, and max pain was wrong

August 21, 2026. Monthly options expiration.

That morning there were two max pain numbers

Chain	Settles on	Max pain	What happened
SPX monthly	the AM settlement (SET)	7,470	—
SPXW weekly (the same-day chain)	the close (PM settlement)	7,670	close 7,674.30 → off by about 4 pts ⭐

📌 An "AM settlement" is not the open. It is SET — a separate number computed once from the opening prints of the 500 components, which is why it matches nothing on your screen. I wrote that one up separately: AM settlement lands on a third number.

Same day, same index, and the two numbers sat 200 points apart — one is the monthly chain, the other the same-day chain.

⚠️ I check which chain a max pain number came from before I quote it. I mixed these two up once and nearly drew the opposite conclusion. More on that in Max pain is a curve, not a number.

On the same-day chain: max pain 7,670, close 7,674.30. Four points. For something that is only supposed to act as a magnet, that is an impressive hit.

I was carrying four positions in two groups

Group	Butterfly	Opened	Debit	Body	Settles on
AM group (monthly)	put fly 7475 / 7525 / 7575 ×2	8/06	2.40	7,525	the AM settlement
	call fly 7430 / 7500 / 7570 ×1	8/07	3.00	7,500	the AM settlement
	put fly 7475 / 7525 / 7575 ×2	8/10	1.40	7,525	the AM settlement
Weekly group (PM)	put fly 7530 / 7560 / 7590 ×1	8/20	1.35	7,560	the close

The two groups settle against different things. The first three end at Friday's SET; the last one ends at Friday's close. Same "August expiration," but the max pain that applies to them is a different number — 7,470 for the monthly chain above, 7,670 for the weekly chain below.

When I opened them, the index was more than 200 points above the bodies

When	Index	Distance to body
8/07 (the call fly)	about 7,746	body 7,500 → −246 pts
8/11 close	7,728	all three −200 to −250 pts
8/13	about 7,799	−274 to −299 pts (further away)

I did not buy these with the index anywhere near the bodies. I bought them a long way out. That part is fine on its own — it is what makes the debit cheap. Section 8 calls it the knee.

The problem was my reason for expecting the index to come down there.

The AM three — I followed max pain, and max pain was 218 points wrong

  monthly max pain   7,470   ← what these three were aimed at
  body               7,500   ← 30 pts from it
  body               7,525   ← 55 pts from it
     ⋮
  upper wings  7,570 · 7,575 ← above this, value is zero
     ⋮
  actual SET        7,687.93 ← 113 pts above the higher wing

I picked the right chain. AM-settled options settle against the monthly chain, and that chain's max pain was 7,470. Putting the bodies there was at least internally consistent.

What was wrong was max pain itself. The actual SET came in at 7,687.93 — 218 points away.

My read on why it sat so low: the monthly chain carries open interest stacked at low strikes from months earlier, which drags the calculation down. ⚠️ I have not counted that separately, though. Section 6 counts how far max pain misses on the same-day chain; it does not count whether the monthly chain leans low. What I have here is one observation.

I closed these three on the morning of Thursday, August 20 — which, for AM-settled monthlies, is the last trading day anyway. You cannot touch them on Friday at all. So this was not a judgment call; it was the deadline. By then there was almost nothing left in them, and holding would have ended the same way: SET 7,687.93 came in 113 points above the upper wings.

🔗 Why an AM settlement is a different number from anything you see on Friday: AM settlement lands on a third number.

The weekly one — max pain was right to 4 points, and I had bet away from it

  close            7,674.30
  weekly max pain  7,670      ← off by 4 pts
     ⋮  80 pts
  upper wing       7,590      ← above this, value is zero
  body             7,560      ← 110 pts below max pain
  lower wing       7,530

Even the upper wing sat 80 points below max pain. The moment price pinned to the magnet, it was sitting entirely outside my tent. This was the only position that reached expiration, and it came to −$135 — the full debit.

And the numbers on it were already on the screen the day I opened it:

max profit 28.65 · max loss 1.35 · breakeven win rate  4.5%

A 4.5% breakeven win rate means I only had to be right about one time in twenty to get the debit back. On its own that is not a bad number — it is what a butterfly is shaped like.

The problem was the chance of that one time arriving. Reaching the upper wing at 7,590 meant 88 points down in a single session, with the magnet sitting 110 points the other way. The August 21 morning report put it plainly: "needs −88 pts to reach the upper wing. There is all day for it, but as things stand this is heavily unfavorable."

🔑 A breakeven win rate tells you how often a structure has to be right, not how often it will be. Read them as the same thing and every 4.5% structure looks good.

So August has two answers, not one

Group	Relation to max pain	Why it lost
the AM three	followed it (30–55 pts)	that chain's max pain was 218 pts wrong
the weekly one	bet away from it (110 pts below)	max pain was right to 4 pts

On that one expiration, one group lost by trusting max pain and the other lost by not trusting it.

But they share a root. All four were a view that the index would come down, expressed as a butterfly. My log from that stretch says it plainly:

"The 21AUG position is a call butterfly, but with the center at 7,500 it is a downside bet all the same."

A butterfly is a bet on where price finishes. All four of mine were bets on how far price would fall. Max pain gave that view a number to point at; it never tested the view.

Two lines I wrote in my log after that close:

"Everything I entered the day before died. No more putting positions on in advance." "The pinning effect on expiration day is frightening."

That second line is exactly the weekly position. Price pinned to max pain at 7,670 and finished there, and my tent was 80 points below it.

3. September expiration week — same 7,500 body, opposite outcomes

A month later. This time I was carrying three butterflies with nearly the same body, all expiring September 17.

#	Butterfly	Opened	Debit	Body	How it ended
A	call fly 7400 / 7500 / 7600 ×1	9/03	5.30 ($530)	7,500	closed 9/16, the day before expiry → +$1,195.24 ✅
B	put fly 7450 / 7500 / 7550 ×1	9/10	3.85 ($385)	7,500	held to expiry → −$389.88 (full debit)
C	put fly 7495 / 7515 / 7535 ×1	9/16	1.50 ($150)	7,515	held to expiry → −$154.88 (full debit)

A and B share the same body, 7,500. Same direction — both need the index to come down. Same expiry, September 17. One came back at a little over three times the debit. The other returned nothing.

September 16 — the day the center was crossed

There was an FOMC decision that Wednesday, and the index dropped quickly after the release.

all times PT · the FOMC statement hit at 11:00 PT (2:00 pm ET)

 11:02   breaks 7,600
 11:36   7,585
 12:00   7,554.32
 12:27   7,508.84   ◀ low of the day — 76 pts in 51 minutes
 13:00   7,551.81   close (+43 pts off the low)

That low, 7,508.84, is 8.84 points from the 7,500 body of butterflies A and B. The index actually walked onto a center I had marked out weeks earlier, and it did it on that one day.

And that day was the day before expiry.

I closed A while price was still coming in

A was bought at a 5.30 debit and closed that day for +$1,195.24 net.

Working backwards, the exit was around 17.3:

exit ≈ debit 5.30 + ($1,195.24 + fees) ÷ 100  ≈  17.3
→ $530 at risk came back as roughly $1,725 (about 3.25× the debit)

That is about 12% of the 94.70 maximum. It sounds small. Section 7 works out why it was enough.

B and C went into settlement

The next day, September 17, the index gapped up more than 80 points overnight, opened near 7,632, and settled on that day's close, 7,637.76.

  7,637.76  ← 9/17 settlement
     ⋮
  7,550     ← B's upper wing   ← 87 pts below settlement. Value zero
  7,535     ← C's upper wing   ← 103 pts below settlement. Value zero
  7,515     ← C's body
  7,500     ← B's body         ← 137.76 pts below settlement

Both settled outside their wings. B came to −$389.88 and C to −$154.88 — each one the full debit.

The one line from that week

Same day, same screen, three positions with the same body. One was closed on the way through the center. Two were sent into settlement. What separated the results was not the forecast. It was whether I closed.

One more thing worth noting. The monthly max pain for the September 18 expiration was 7,500 — the same number as B's body. But B expired a day earlier, on September 17.

🔑 The day max pain points at and the day my position settles may not be the same day. A right number on the wrong date is not a right answer.

4. What holding to expiry actually does

A friend of mine was carrying a butterfly around the same time. One thing was different: they waited for expiry. The position went to zero.

I do not read that as their mistake, because my own August log has exactly the same shape — I opened one the day before expiry, carried it into settlement, and it went to zero. They did in September what I had done in August.

Held into settlement, a butterfly has no in-between. Time value goes to zero at settlement, so all that is left is the tent itself.

If settlement prints	the butterfly is
near the body	worth a lot
near a breakeven	roughly flat
outside the wings	exactly zero

And outside the wings is far wider. In the example above, the range that retains any value is 7,405.30 to 7,594.70 — 189 points. Outside it runs without limit in both directions. Waiting for expiry means standing on the wide side with nothing prepared.

5. September 18, the quarterly expiration — the week ran between two max pain levels

September 18 was a quarterly expiration. I did not use a butterfly that day — there is no butterfly in that day's log. What was worth recording was on the max pain side.

Chain	Settles on	Max pain	What happened
SPX monthly	the AM settlement (SET)	7,500	9/16 low 7,508.84 — turned 8.84 pts above it
SPXW weekly	the close (PM)	7,650	9/18 settlement 7,650.50 — 0.5 pts ⭐

The monthly number behaved like the week's floor, the same-day number like its destination.

⚠️ This is one occurrence. I am not making a rule out of it. Whether the next expiration does the same thing, I do not know yet. So it goes in as an observation, not a rule.

6. So can max pain be trusted on expiration day? I counted 14 months.

Those three hits were impressive. But the three I find impressive are the three I remember. So I went back over 14 months of expirations.

The thing that matters here is a baseline. Without one, "off by 50 points" tells you nothing about whether that is good. So I used the laziest possible forecast as the baseline: the market just closes where it closed yesterday.

Expiration type	Sample	Max pain mean error	Baseline (yesterday's close)	Times max pain won
Monthly expiration (third Friday · same-day chain)	14	101.7 pts	47.3 pts	5 / 14
Month-end	14	149.8 pts	48.2 pts	4 / 14
0DTE (queried the night before)	82 days	94.7 pts	45.6 pts	25 / 82

How I read that:

Expiration-day max pain is clearly more accurate than month-end or 0DTE. Open interest is overwhelmingly concentrated that day. That matches my sense of it
But it is about twice as inaccurate as "yesterday's close." It won 5 times out of 14
So I do not use it to call the close — not on this sample

The last three were unusually close, though:

2026-07-17   off by  62 pts
2026-08-21   off by   4 pts
2026-09-18   off by 0.5 pts   (same-day chain)

That is a sample of three. Whether it is a trend or a coincidence, I do not know yet. I am still counting.

How I actually use it

So this is what I do with max pain.


🚫 Not this	"Max pain is 7,500, so price goes there" → I do not use it as a reason to enter
✅ This	"The magnet sits near 7,670 today, so a move up may run out of force around there" → a ruler for «how far»

I do not use it to pick a direction yet. I use it to measure distance along a direction I already picked. August expiration shows the difference: as a ruler it was accurate to 4 points, and it did nothing to save my position.

7. Risk and reward — the head is one point, the shoulder is a range

This is the most important arithmetic in the post.

On paper, butterfly A looks like this:

max profit  94.70  ($9,470)
max loss     5.30    ($530)
                       →  1 : 17.9

1 to 17.9. On that number alone it can look like there is no reason not to take it. But the number carries an unstated condition.

That 94.70 exists only if settlement prints «exactly 7,500».

Inside a 100-point span, the odds of settlement landing on one specific point are effectively nil. 1:17.9 is real and almost never reached.

Here is what the position is actually worth at each settlement price.

Settlement	Butterfly value	P/L on a 5.30 debit	Value ÷ debit
7,500 (exactly)	100.00	+94.70	18.9× ← the head
7,520 or 7,480	80.00	+74.70	15.1×
7,560 or 7,440	40.00	+34.70	7.5×
7,583	17.00	+11.70	3.2× ← near where I actually got out
7,594.70	5.30	0	breakeven
7,600 and above	0	−5.30	total loss ← 9/17 landed here

📌 That last column is the whole payout divided by the debit, so it runs one higher than the profit ratio above (17.9 vs 18.9). Same position, counted two ways. And the 7,583 row is expiration intrinsic value, while my actual exit was the day before with time value still in it. The two landing this close together is a coincidence.

Reading that table top to bottom:

The top two rows (15× and up) almost never happen. Those bands are extremely narrow
The bottom row (total loss) happens easily. That band has no outer edge
The middle band is where it actually lands

And there is one more difference that decides everything.

The head (max profit) exists only at the instant of settlement. The shoulder (3–7×) exists for the whole time the center is being crossed.

On September 16 the index took close to an hour to travel from 7,585 down to 7,508. For that entire hour, something around 3× was sitting on the screen. The chance of settling exactly at 7,500, by contrast, was a single moment the following afternoon — and it never came.

The 3.25× I took was 12% of the maximum. And it is still not the same thing as zero.

8. Knee and shoulder — the three lines I settled on

In Korea there is a stock-market saying: buy at the knee, sell at the shoulder — never at the foot, never at the head. I think it fits options, and butterflies in particular, more exactly — with stock you can keep holding past the shoulder, but an option ends when expiration arrives.

So I wrote it down as three lines.

🦵 The knee — I buy while the center is far away

Butterfly A was bought on September 3. At that moment the index was more than 200 points above the 7,500 body. It looked like a place price would never reach, and that is why the debit was only 5.30.

Buying once price is already near the center feels far more confident, but by then I am paying for all of it up front. I wrote that up separately — I open the butterfly before price arrives.

💪 The shoulder — while the center is being crossed, if there is value, I take it

Not "when it touches the center" but "while it is crossing." That difference matters.

On September 16, as the index went from 7,585 down to 7,508, I did not wait for it to touch 7,500. If I wait for the touch, by the time I look it may already have turned. That day's close was 7,551.81 — 43 points back off the low.

I closed on the way in, and that was 3.25×.

🧠 The head — I do not wait for settlement

Holding to settlement is a bet on a single point called "exactly the body." I no longer treat that as my default.

I kept one case for holding into expiry: when the position is already small enough that I am fine getting zero for it — a position whose value has mostly drained away, where closing changes nothing. Holding something that still has value in order to wait for more value is what I stopped doing.

⚠️ This is my rule, and the sample is small. Four positions through August expiration, three through September expiration week — two expirations in total. There is no guarantee the same situation produces the same result. What I have written above is everything I was looking at when I decided.

9. Three things that are easy to get wrong

① "Max pain was right, so I must have made money"

No. August expiration is the proof. Max pain was right to 4 points, and because my center sat 110 points below it, max pain being right was the reason I lost.

Max pain is computed from the whole market's open interest. My body is a number I picked weeks ago. Unless the two happen to coincide, one being right says nothing about the other.

② "Hold to expiry and you get the maximum"

Hold to expiry and you get the maximum or you get zero. And the zero side is overwhelmingly wider.

Look at the section 7 table again — 18.9× requires settlement on one exact point, while the zero band runs without limit in both directions.

③ "The max loss is only the debit, so just leave it"

That thought is what let three of my August flies drain to almost nothing before their last trading day, and it is what sent the fourth into settlement at zero.

A small max loss and a low chance of loss are not the same thing. A butterfly is a structure where the loss is small and you take that loss often. Once you think "it's only $150," there is no longer any reason to press the close button — and then the $150 really does go.

10. Summary

Max pain being right and my position winning are different events. August expiration: weekly max pain was off by only 4 points, and with my center 110 points below it that was precisely the full loss
I check which chain a max pain number came from. Same day: monthly 7,470, same-day 7,670 — and the monthly number those three were aimed at missed the actual SET by 218 points
I check that the day max pain points at is the day my position settles. In September the number was right and the date was off by one
Over 14 months, expiration-day max pain averaged 102 points of error against 47 for "yesterday's close." I do not use it to pick direction yet — only as a ruler for distance
Held into settlement, a butterfly is the body or zero. There is no in-between
A 1:17.9 payoff is real and almost never reached. The head is one point; the shoulder is a range
I settled on closing while the center is being crossed, as long as there is value on the screen. In September the one position I handled that way returned 3.25×; the two I did not went to zero
The sample is two expirations and seven butterflies. I will keep counting

※ This is a record of my own trades and how I reasoned through them. It is not a recommendation of any security, strike, or structure, and I am not an investment adviser. A single result here says nothing about future results. The sample is small and there is no guarantee that the same situation produces the same result. Options can lose the entire amount committed over a short period, and a butterfly losing its full debit at expiration is common. All investment decisions and their outcomes are your own.