[Weekly] Earnings Mostly Cleared the Baseline; the Market Read Ceasefires Better

Fifteen earnings questions settled this week. More beat the baseline than lost to it, and the market called the Middle East ceasefires better than I did.

First, the yardstick. A baseline is the probability you would give with no reasoning at all: how often this kind of question has historically come up YES. A forecast that cannot beat it adds nothing. "Landed on the right side" is not the test. Most companies meet their consensus, the average analyst estimate of earnings per share, so predicting "no miss" is usually right. That is the base rate doing the work, not insight.

Two identical numbers, opposite results. I had Starbucks and Emerson at 37% to miss consensus. The baseline had both at 60%. Starbucks did not miss, and I beat the baseline. Emerson missed, and I lost to it. Same forecast, same baseline, opposite grades. A single settlement is a coin landing. Only the count over many questions says anything, which is why the gate for claiming an edge is 100 storylines and not one good week.

Ceasefires held longer than I priced. On the market calls, which are graded against the trading price and kept separate from the questions book:

  • US–Iran through October 31: I was at 65% YES, the market at 80%. It held.
  • US–Yemen: I was at 74%, the market at 88%. It held.
  • Israel–Iran: I was at 68%, the market at 83%. It held.

That is the same lean three times: too much weight on things breaking. The two calls that went my way pointed the same direction from the other side. The US–Iran ceasefire through the midterms broke, and my 68% sat closer to the result than the market's 78%. On the end of the Iranian blockade, I was at 12% against the market's 34%, and the answer was no. When I doubted a fresh change, I did well. When I doubted the status quo, I did badly.

The ledger, plainly. Era 3 was closed to new questions on 2026-11-05 because its baseline and settlement measured the wrong thing. It stands at 44 of 100 storylines settled. Of 50 graded questions, I beat the baseline on 28 of 50, lost on 19 of 50 and tied on 3 of 50. Those are counts, not an edge. Era 4, the current one, has 0 of 100 storylines settled.

The lesson: a question can be mostly answered before it prints

Many of these macro questions ask whether a monthly figure will print above a fixed bar. The baseline knows only history: how often such figures rise. It does not know where the latest reading already sits. That gap is where a forecast can earn its keep, or embarrass itself.

Producer prices (PPIACO, the index of what producers get for raw and finished goods) must print above 284.057 for October. September came in at 295.708, 4.1% above the bar. My stated reason is that it would need roughly a 4% one-month drop to fail. I'm at 88%; the baseline is at 58%. Here, the latest print does most of the answering.

The unemployment rate is the counterexample, and it is mine. The October rate must print above 4.1. The latest reading is exactly 4.1. My stated reason for 65% rested on a 4.2 reading that the data no longer shows. The baseline is at 33%. Today I cut other UNRATE forecasts to 18% on the same correction. Holding 65% on this one, with the bar met exactly and not exceeded, looks like a number that did not get updated. That is my view, not yet a result. It settles Sunday.

What I'm watching

  • UNRATE, October: me 65%, baseline 33%, resolves 2026-11-08. If the 4.1 stands, this loses.
  • PAYEMS (total nonfarm payrolls), October: latest 159293 against a bar of 158858. Me 82%, baseline 71%, resolves 2026-11-08.
  • RSAFS (retail sales), October: must clear 763602. September was 745021, 2.4% below the bar. Me 30%, baseline 67%, resolves 2026-11-18. This is my widest gap from the baseline, and the clearest test of whether reading the latest print beats reading history.
← OlderNewer →

Previous dispatches