Where the Signals Point Now

In one minute:

A story only becomes useful when it turns into a probability that someone can check later.

Mid-October's signals point the same way: energy-led inflation, high long-term yields, and an Ebola outbreak that is slowing but not over.

We registered twelve forecasts on 16 October. The Brier score will show whether our numbers were honest.

1 October, and a fridge that keeps humming

Imagine a clinic in Nairobi on the first day of October. The vaccine fridge hums in the corner. A nurse writes the stock count on a clipboard and wonders, quietly, what the next delivery will cost. Nothing in the room has changed yet. Two weeks later, the same room has numbers attached to it.

A good forecaster reads the world the way a farmer reads weather. No single cloud decides the day. What matters is when several pressure systems start moving together.

What was visible

The first system was oil. Brent settled at 105.69 on 13 October, after an intraday high of 106.75 (Investing.com). Two days later it had eased to 103.89 (Trading Economics). The Gulf conflict had not settled: the US naval blockade remained in force, and on 12 October Iran's security chief named the US-UAE oil shuttle out of Hormuz as a target (Al Jazeera).

The second system was food and fuel. The FAO Food Price Index reached 136.0 for September, its highest since November 2022 (FAO). In Kenya, inflation rose to 6.8% in September from 6.6% in August (KNBS). Pump prices followed. From 15 October, super petrol in Nairobi was set at KSh 226.96 (The Star).

The third system was money. The US 10-year Treasury touched 5.35% intraday on 7 October, its highest since 2002 (CNBC), and closed at 5.24% on 9 October (FRED). In September the Federal Reserve had raised its target range to 3.75%–4.00% (CNBC).

The fourth system was health. Kenya confirmed its first imported Bundibugyo ebolavirus case on 6 October. The patient had reached Nairobi on 3 October and died on 5 October (AABB). The Democratic Republic of Congo had 9,144 cases as of 13 October (ECDC).

The fifth system was climate. NOAA-based reporting gives an 83% chance that the October–December El Niño reaches at least +2.5°C. That figure comes from a secondary source and should be checked against NOAA's own discussion (Watchers.news). Kenya's Met Department expects enhanced short rains over about 80% of the country (The Star).

Each of these is weather on its own. Together they form a pattern.

What happened

Table sources: FAO, CBK, CNBC, ECDC, Kenyans.co.ke.

The Ebola curve teaches the most about method. The doubling time lengthened from about 24 days in the late June to mid-July stretch, to about 32 days across the period to 23 September, and to roughly 98 days in the most recent stretch. That is slowing, not stopping. Among confirmed cases, about 48% have died, 3,799 of 7,890 (Wikipedia).

We count confirmed cases only. A second Kenyan case reported on 15 October has no WHO or Ministry of Health confirmation, so this chapter does not count it.

The math: from story to probability

A story says "this will last." A forecast says "70%." Calibration is the discipline that links the two. Each forecast is scored with the Brier score:

Brier = mean((p − outcome)²)

Here p is the probability you gave, and the outcome is 1 if the event happens and 0 if it does not.

Step one: state the forecast as a number. Say p = 0.7.

Step two: if the event happens, score (0.7 − 1)² = 0.09.

Step three: if it does not happen, score 0.7² = 0.49.

Step four: compare with a lazy baseline. A forecaster who always says 50% scores 0.25, whatever happens: (0.5 − 1)² = 0.25 and 0.5² = 0.25.

Read the result carefully. The 70% forecaster beats the baseline when the event happens and loses to it when the event misses. One forecast tells you very little. Over many forecasts, the average tells you whether you are calibrated. Calibration means that of all the events you called 70%, about 70% actually happened. A score clearly below 0.25 over a set of forecasts is a claim of skill. A score above it is not.

The score also punishes overconfidence. A forecaster who says 0.95 and misses scores 0.9025. Saying less than you feel, and being right about it, earns more than saying more and being wrong.

The Pawanax lens

For a clinic, the first lesson is about the cold chain and cash. Fuel prices move the cost of every delivery, and inflation moves the cost of every stock order. When 77.6% of tax revenue goes to debt service, as it did in the first quarter of the 2026/27 fiscal year, every KSh 100 of tax leaves KSh 22.40 for everything else. Clinics feel that squeeze before the national statistics show it.

For families, the signal is in the food line. Food prices rose 9.5% year on year in September (KNBS), and a household budget has little room to absorb a second shock. A health system that can see a referral coming, and a stock count that reacts to the price of fuel, is worth more when incomes tighten. At Pawanax, that is the kind of tool we are trying to build.

The next 90 days

We registered our forecasts on 16 October. Three of them speak directly to this chapter:

• F5: 55%. Kenya's December 2026 CPI, year on year, prints at 7.0% or higher. Resolves 5 January 2027. September was 6.8%, so this needs a further rise.
• F6: 80%. The WHO Bundibugyo Ebola emergency of international concern is still in force on 31 January 2027. It was declared on 16 May (Wikipedia).
• F7: 20%. Kenya confirms at least one locally acquired, non-imported Bundibugyo case by 31 January 2027.

Each one resolves YES or NO. We will score each forecast as (p − outcome)² and publish the results, including the ones that score badly. The full set of twelve closes in 2027, and the average will be compared with the 0.25 baseline. If we are not better than a coin, the numbers will say so.

Part 10 of The Signal Ledger. Read the series: [The Signal Ledger](/blog/signal-ledger-2026-read-backwards).

Cover image: Maasai Mara landscape beneath clouds and sunbeams.jpg — Daniel Case · CC BY-SA 4.0. Photographed 20 August 2025 in the Maasai Mara, Kenya. Illustrative landscape for a forward-looking outlook, not a 2026 weather observation or forecast. Resized for web display.