Scoring the Forecasters

The short version: anyone can sound right about the economy. The memorable calls get replayed and the misses quietly disappear. The 65 Signals accuracy board does the unglamorous part: it writes down each directional call a tracked macro voice makes, waits, and checks it against what the market actually did.

What is on the board today

As of the October 19, 2026 scoring run:

• 4,751 directional macro calls from 93 tracked voices have been considered.
• 3,323 of those calls have reached at least one checkpoint and been scored against prices.
• 32 voices have enough scored outcomes to be ranked. Together they account for 4,072 calls and 6,328 scored call-checkpoint outcomes.

The board, October 19, 2026 scoring run

4,751directional calls considered

93tracked voices

3,323calls scored at a checkpoint

32voices with enough outcomes to rank

These numbers grow as new calls are recorded and older ones reach their checkpoints, so the live board is the place for current figures.

Why grade on outcomes at all

Macro commentary is usually judged by how convincing it sounds: a clear story, a confident tone, a famous past call. None of those is evidence that the next call will be right. Three things make outcome-based scoring worth doing:

• Memory is selective. One dramatic hit can carry a reputation for years, while a long run of misses fades. A ledger remembers every call equally.
• Base rates matter. Markets rise more often than they fall, so "up" is right more often than not. A voice is only adding information if it beats what a simple default would have scored.
• Timing is part of the claim. "Eventually" is not testable. Fixed checkpoints turn a view into something that can be checked.

This is the same idea behind the rest of 65 Signals: compare the narrative with the data and let the data decide.

How a call is scored

One call, four checkpoints

0 dcall 30 d×1 60 d×1.5 90 d×2 180 d×2.5 call

At each checkpoint: the price move since the call

above +2%Right for an "up" call, wrong for a "down" call ±2%Inconclusive: counts as neither below −2%Right for a "down" call, wrong for an "up" call

• A call is a direction on an instrument at a date. For example, a claim that a liquid, tradeable market will rise or fall, made on a known day.
• Four checkpoints. Each call is checked against the instrument's price 30, 60, 90 and 180 days later.
• A dead zone for noise. If the price has moved less than 2% either way at a checkpoint, the result is inconclusive. It counts as neither right nor wrong.
• Right ÷ (right + wrong). Each checkpoint's hit rate leaves the inconclusive results out.
• Longer horizons weigh more. The overall score weights the 30, 60, 90 and 180-day hit rates 1, 1.5, 2 and 2.5. A durable call counts for more than a quick bounce.
• A minimum before ranking. A voice needs at least 10 scored outcomes before it is ranked, and each cell on the board needs at least 10 resolved calls before it shows a rate. Every rate is shown with its sample size.
• A consistency check. The board's out-of-sample column asks whether a voice's later calls track its earlier ones. That separates a steady record from a lucky streak.

What the score does not tell you

• Direction only. It does not measure how big a move was, or whether a call was early and then right.
• The checkpoints are choices. A call that is wrong at 30 days and right at 180 shows both results; neither one is the whole story.
• Extraction is imperfect. Calls come from public commentary. The methodology notes on the board record changes, such as tighter eligibility rules that dropped undateable and non-predictive claims.
• A high score is a record, not a forecast. Past performance doesn't guarantee future results. A voice at the top of the board today can be wrong about the next call, and the board will record that too.

How to read it

Use the board to calibrate, not to follow. When a well-known voice makes a call, the board tells you how that voice's past calls actually turned out, at which horizons, and on how many observations. Pair that with the macro phase, which reads the credit market directly, and you have both the story and a check on the storyteller.

See the accuracy board on 65 Signals →

Sources

• Counts and scoring parameters: the 65 Signals accuracy backtest, run October 19, 2026 (checkpoints 30/60/90/180 days; 2% noise threshold; horizon weights 1/1.5/2/2.5; 10-outcome ranking minimum).

This article is for educational and informational purposes only. It is not investment advice, and nothing in it endorses or recommends any commentator or position. Past performance doesn't guarantee future results.