1. The week in one page

Last Sunday I told you two things. First, that the 26-week average closing price would fall again unless the week printed above $1,690,655. It printed $2,096,232, the fifth-highest single week in 35 years of daily records, and the 26-week average rose $15,511 to $1,606,256. That is the second-highest reading in 1,820 weeks and $1,958 below the record set July 20, 2025. I was wrong about the direction, and I will show you why in section 2: the week's closings were carried by a handful of very large sales on a very thin day count.

Second, I wrote that if pendings printed under 22 a day this week, the honest reading would be that demand broke in October rather than wobbled. Pendings printed 19.7 a day. That is the second straight week under 20. In 35 years of October weeks, back-to-back Octobers under 20 pendings a day have happened in exactly three stretches before this one: 1991 through 1994, 2007 and 2008, and now. I set the line, the market came in under it, and I am not going to move the line now that it is inconvenient. Demand broke in October. Price has not noticed yet.

• Price: 26-week average $1,606,256, up from $1,590,745. Second-highest ever. Year-over-year gain widened to +2.4%. The 13-week average also rose, to $1,607,412.
• Demand:Supply Ratio (#P/#N.L.): a negative divergence, and the early-warning line just tripped. Raw 0.484, up a hair from 0.450 only because new listings fell to 40.7 a day. The 13-week average fell to 0.5657, which is below the 0.566 line that separates zone Q3 from zone Q2 on my 35-year map. The 26-week average ticked up 0.002 to 0.571 because an easy week (0.44) left the window. It is 0.005 above the line.
• Success to Failure Ratio ((#C+#P)/XK): both averages are now moving the wrong way, and the 13-week has crossed. Raw 2.29. Expired and cancelled listings stayed low at 14.1 a day, which is the October pattern. The problem is the numerator: 12.7 closings a day plus 19.7 pendings is the thinnest success count since mid-September. The 13-week average fell to 2.72, below the 2.78 zone line. The 26-week average fell a seventh straight week to 2.954.
• The counterweight is weaker than it was. The 4-week average pending price slipped to $1,791,744 from last week's all-time high of $1,848,263. Sales still closed at 98.9% of list. The buyers who remain are paying up; there are simply fewer of them every week.
• My forecast for the 26-week average closing price: −3% to +1% over 13 weeks (trimmed from −2% to +2%), −1% to +3% over 26 weeks (unchanged), and +2% to +6% over 52 weeks (unchanged). Conviction: moderate on the 13-week call, lower on the longer two.
• What to expect next Sunday: the week leaving the window is $1,887,818. Only 5 of the last 52 weeks have printed above that, so the 26-week average most likely falls, and the record most likely stays out of reach for now. The week after that the exit is $1,313,307, which is easy to beat. Expect a down Sunday, then an up Sunday. Neither is news.

2. Scorecard: what I wrote last Sunday, and what happened

On the miss. I used the last 52 weeks as my base rate for a $1.69 million print, and 7 of 52 is 13%. What I under-weighted is that a thin closing week is exactly the kind of week that produces an outlier average. Twelve-point-seven closings a day is the 8th percentile of all weeks since 1991. On a day count that low, three or four trophy closings move the week's average by hundreds of thousands of dollars. My daily records for this week show what carried it: a Strand duplex at $10,950,000, a Lower Paseo La Cresta estate at $8,000,000, a Strand triplex at $8,500,000, and a 4th Street Manhattan Beach sale at $8,650,000, among others. Strip the outliers and the week looks like the two before it. I am not stripping them, because that would be grading on a curve, but the lesson goes in the method: when closings are thin, the direction of the weekly print is closer to a coin flip than the 52-week base rate suggests.

3. The chart: price and its two leading indicators, last three years

The top panel is what everyone looks at. The bottom two panels are what I look at. Price is at the record line. Both indicators are falling toward, or already through, the zone lines that have separated flat markets from rising ones since 1991. The heavy lines are 26-week averages; the light lines are 13-week averages, which lead the heavy ones by construction. Both light lines are now below their zone lines. Both heavy lines are within a few weeks of following.

4. The Demand:Supply Ratio (#P/#N.L.): the divergence, and what I mean by "broke"

The Demand:Supply Ratio is pendings divided by new listings. It answers a simple question: for every 100 homes that come to market in a week, how many homes find a buyer? This week the answer was 48. Last week it was 45. The ten-year norm for the third week of October is 71, and last year this same week printed 63.

I want to be precise about what "broke" means, because the word carries weight. It does not mean prices are falling. It means the count of buyers going into escrow has dropped to a level that, relative to its own 52-week average, is in the bottom 14% of all two-week stretches since 1991, and the lowest since January. Pendings over the last four weeks averaged 22.6 a day against 25.0 a year ago. The number of new listings has not changed; 42.5 a day on a four-week basis is normal for October. This is a demand event, not a supply event. Sellers did not flood the market. Buyers stepped back.

Here is the window arithmetic on the 26-week average, which is what I use for the zone test. It sits at 0.571, five one-thousandths above the Q2/Q3 line of 0.566. The weeks leaving the average over the next eight Sundays read 0.59, 0.66, 0.60, 0.52, 0.57, 0.56, 0.49 and 0.53. Every one of the first three is above this week's print. If the next four weeks print at this week's 0.484, the 26-week average runs 0.567, 0.561, 0.556, 0.555. It crosses into zone Q2 on November 8, three Sundays from now. To hold the line through January 10, the next twelve weekly prints need to average 0.558. The last two averaged 0.467.

What the history says about weeks like this

I went back and found every two-week stretch since 1991 in which pendings, measured against their own trailing 52-week average, were as weak as they are right now. After removing overlapping weeks there are 34 earlier episodes. Thirteen weeks later the 26-week average closing price was lower 62% of the time, median −0.8%. Twenty-six weeks later it was higher 68% of the time, median +1.8%. Fifty-two weeks later it was higher 67% of the time, median +6.1%. The near-term tilt is slightly negative. The one-year tilt is still positive. That shape is why my 13-week range moved down and my 52-week range did not.

The three episodes that most resemble today on the calendar, meaning two straight October weeks under 20 pendings a day, are 1992 through 1994, 2007 and 2008. Two of those three were the worst stretches in this data set. I am not predicting either. The 1991-1994 market had 150 new listings a day against 40 today, and 2008 had a credit system failing underneath it. But I would be hiding the ball if I did not tell you that the calendar analog is ugly, and that the only thing keeping it from being the base case is that the price indicators have not confirmed it.

5. The Success to Failure Ratio ((#C+#P)/XK): the trend is the problem

The Success to Failure Ratio is closings plus pendings, divided by expired and cancelled listings. It measures how many listings succeed for every one that fails. This week: 12.7 closings plus 19.7 pendings, divided by 14.1 failures, is 2.29. Last week it was 2.45. Last year this week it was 2.72. The ten-year norm for the third week of October is 4.09.

Failures are not the issue. At 14.1 a day, expired and cancelled listings are back at the October seasonal low after the quarter-end spike, which is the pattern in 36 of 36 Octobers since 1991. The issue is that the success side is shrinking. Closings fell to 12.7 a day and pendings to 19.7. The 26-week average fell to 2.954, a seventh straight decline. The 13-week average fell to 2.72 and is now below the 2.78 line that separates zone Q3 from zone Q2.

Window arithmetic: the weeks leaving the 26-week average over the next eight Sundays read 2.96, 2.74, 2.93, 3.83, 3.10, 4.05, 2.97 and 3.42, and they average 3.25. At this week's 2.29 print, the 26-week average crosses 2.78 into zone Q2 in five weeks, on November 22. To stay in zone Q3 through January 10, weekly prints need to average 2.72. The last five weeks averaged 2.53.

A seven-week decline in the 26-week average is not rare: it has happened 38 times before, and the median such run lasts 11 weeks. So the trend alone is not a signal. What makes it matter is where it is headed, which is the same zone the Demand:Supply Ratio is headed, in roughly the same month.

6. Convergences and divergences: are the signals statistically valid?

Three tests, same as every week, run on all 1,845 weeks. I report them straight.

1. Level test (does the zone an indicator sits in predict forward price?) Yes, and this is the strongest result I have. The Demand:Supply Ratio's 26-week average level correlates with the forward 13-, 26- and 52-week change in the 26-week average closing price at r = 0.40, 0.47 and 0.58. After correcting for overlapping weeks (the Bartlett adjustment, which cuts the effective sample to between 26 and 64 independent observations), the t-statistics are 3.4, 3.5 and 3.5. All three clear the 1% bar. The Success to Failure Ratio's level is weaker but also valid: r = 0.32, 0.32, 0.51, t = 2.7, 2.2, 3.0.
2. Momentum test (does a 13-week change in an indicator predict forward price?) Mostly no. Over the full 35 years the Demand:Supply Ratio's 13-week change has no predictive value at any horizon (t under 1.2). The Success to Failure Ratio's 13-week change predicts the 13-week forward price change at t = 2.0 over the full sample and t = 2.8 since 2012, which is a real but modest result. That momentum is negative right now (the 26-week average is down 10% in 13 weeks), and it is the main reason I trimmed the top of my 13-week range.
3. Divergence test (what happens when price is near a record while an indicator is low and falling?) I found six earlier clusters where the 26-week average closing price was within 0.5% of its running record while the Demand:Supply Ratio's 26-week average was at or below 0.60 and the Success to Failure Ratio's 26-week average had fallen five or more straight weeks. Four of the six saw the price average lower 13 weeks later; five of six lower 26 weeks later; the median 26-week change was −2.8%. The sample is six. I would not bet the house on six. But it is the one test that points down at every horizon, and today's reading matches its conditions.

The joint cell matters more than either table alone. When both 26-week averages sit in zone Q3, as they do today, the 13-week forward median is +2.0% and the 26-week is +3.3%. When both sit in zone Q2, where both 13-week averages already are and where both 26-week averages are headed by late November, the medians drop to +0.7% and +1.1%, with 62% to 63% odds of being higher. That is a flat market, not a falling one. The difference between those two cells is the difference between my forecast two weeks ago and my forecast today.

7. What I would do if I were selling

• Price to the closings, not to the headline. Next Sunday's report will almost certainly show the 26-week average falling, and the Sunday after will show it rising. Neither move is about your home. Price to the last three comparable closings within a half mile, adjusted for condition, and ignore both headlines.
• The window before Thanksgiving is three weeks and it is the best one you will get until March. On current prints both of my indicators cross into the flat zone between November 8 and November 22. The buyers active today are serious (98.9% of list) and they know inventory thins in December. If you are ready, list now. If you have been on the market more than 30 days without an offer, the next price adjustment should be a real one, not a $25,000 gesture.
• Do not anchor to the $2.1 million week. That number was made by a few eight-figure sales on the Strand and in Palos Verdes Estates on a day count in the 8th percentile. If your agent cites it as evidence the market is up, ask how many closings were in the sample.
• Waiting for spring buys you less than you think. My 26-week range is −1% to +3%, so the price difference between now and April is inside the noise. And over the last ten years spring pendings have run only about 6% a day above October's, while spring new listings run well above. You would be trading a thin field of buyers for a slightly thicker one and a lot more competing sellers. If you are ready, the arithmetic favors now.

8. What I would do if I were buying

• You have more leverage than you have had since January, and about three weeks to use it before sellers pull listings for the holidays. For every 100 new listings only 48 are finding a buyer. Sellers who have been on the market 45 days or more know that, even if their list price does not show it yet.
• Write the offer at the comparable closings and let the seller's expectations come to you. The 4-week pending average fell $56,519 this week. The premium that pending buyers were paying over closing buyers has narrowed from 15% to 2% in one week. The air is coming out of the top of the market first.
• Do not wait for a crash. I know the 2007-2008 analog is in this report. I put it there because it is true, not because I expect it to repeat. The zone my indicators are entering has produced a higher price a year later roughly two times out of three, with a median gain of 2% to 6%. The expected cost of waiting a year for a 3% drop that happens one time in three is higher than the expected saving.
• Target the thin segments. Closings fell to 12.7 a day. The segments where that shows up first are the ones with the fewest buyers to begin with: large-lot Peninsula homes above $4 million, older condo stock in Torrance and the Harbor area, and anything that has already had one price cut. That is where a full-price offer with clean terms gets a yes today.

9. Devil's advocate: how I could be wrong

I write this section against myself every week. This week it is longer than usual, because my own scorecard shows I called the direction of price wrong seven days ago.

• "Demand broke" is my rule, not a law of nature. I set 22 pendings a day as the line last Sunday without a historical test behind it. The market printed 19.7. A fairer analyst might note that the two weeks in question include Columbus Day, that pending counts are noisy on a one-week basis, and that a single 25-pending week next Sunday would put the four-week average back where it was in September. If that happens I will have declared a break that was a two-week air pocket, and the headline of this report will look foolish.
• The thin-closing argument cuts both ways. I just told you that a $2.1 million week on 12.7 closings a day is a small-sample artifact. By the same logic, 19.7 pendings a day on a holiday week is a small sample too. I cannot call one number noise and the other number signal because one suits my thesis. The defensible version is: both are thin, and I will not know which was real until mid-November.
• The 1991-1994 and 2007-2008 calendar analog is more scary than useful. Those markets had three to four times today's listing volume, so 20 pendings a day meant something very different. If I normalize for volume the analog mostly disappears. I included it because leaving it out would be selective, but a reader who dismisses it is on reasonable ground.
• The six-case divergence analog is six cases. I said that in section 6 and I mean it. With six observations the 95% confidence interval on "five of six were lower" runs from roughly 36% to 99%. It is consistent with a coin flip. If I were reviewing someone else's report I would tell them not to lead with it.
• The momentum result I leaned on to trim the 13-week range is the weakest of my three tests. t = 2.0 over the full sample is right at the edge, and I know from experience that results at the edge are the ones that fail out of sample.
• I may be fighting the last war. Two weeks ago I was wrong about the calendar helping pendings. One week ago I was wrong about price falling. There is a version of this report where I am now over-correcting toward caution precisely because I have been burned twice, and the right call is the one I made on October 4: nothing much changes for two quarters. If prices are higher by 2% in January and both indicators turned up in November, that version was right.
• Rates. Nothing in this spreadsheet knows what mortgage rates will do. A half-point move in either direction over the next 13 weeks would dominate every indicator in this report.

Where that leaves me: I put the bear case (a 3% to 5% decline in the 26-week average by mid-January) at roughly one chance in four, up from one in five last week. I put the flat case at better than even. I put the record being taken out before year-end at roughly one in five, and only because another thin-closing trophy week can do it by arithmetic. I would rather tell you the odds moved than pretend they did not.

10. Data notes

• Source: weekly averages of daily CRMLS activity for the Palos Verdes Peninsula and South Bay, 1,845 consecutive weeks from May 24, 1991 through October 18, 2026. Verified in two independent passes (openpyxl and pandas) with row count, year distribution, last-row values and both ratio recomputations matching to machine precision. No duplicate dates, no missing values.
• Demand:Supply Ratio = # of Pendings ÷ # of New Listings (spreadsheet column #P/#N.L.). Success to Failure Ratio = (# of Closings + # of Pendings) ÷ # of XK's, where XK means expired or cancelled (spreadsheet column (#C+#P)/XK).
• Zone boundaries are the 20th, 40th, 60th and 80th percentiles of each indicator's 26-week average across all 1,820 available readings: 0.410 / 0.566 / 0.686 / 0.815 for the Demand:Supply Ratio and 1.784 / 2.780 / 3.470 / 4.251 for the Success to Failure Ratio. Last week I quoted the upper Success to Failure boundary as 3.471; this week's recomputation on one more week of data gives 3.470. The standing announcement threshold is whatever that boundary is on the Sunday it is crossed.
• Forward returns measure the change in the 26-week average closing price, not the weekly print. Correlations use Bartlett-corrected effective sample sizes to account for the overlap in rolling windows; quoted t-statistics are on the corrected sample.
• The "low normalized pendings" analog compares the two-week average pending count to its trailing 52-week average and takes every episode at or below this week's reading (0.790), de-clustered at 12 weeks. The divergence analog uses a 0.5% band around the running record and is de-clustered at 8 weeks.
• Individual sale references in section 2 come from my daily CRMLS status logs for October 12 through October 18, 2026 and are cited to explain the week's average, not as comparables.
• This is market analysis, not an appraisal or a guarantee. Past patterns in these indicators do not guarantee future prices.