Why a private prediction journal
may sharpen your calibration
Making a prediction is an act of intellectual honesty: you commit to a belief before the outcome is known. But the moment that prediction is visible to others, something shifts — you are no longer just reasoning, you are also managing how you appear.
This page makes a single, limited argument: for a specific purpose — understanding your own reasoning — a private prediction log may reduce a particular kind of distortion. It is not an argument that private forecasting is universally better. Public and collaborative forecasting platforms serve genuinely different and valuable goals.
The social layer changes what you forecast
When predictions are visible to others, two pressures emerge. The first is what social scientists call social desirability bias — the tendency to respond in ways that seem acceptable or admirable rather than simply true. Applied to forecasting: you might shade your confidence to avoid looking reckless, or skip a prediction altogether because being wrong publicly would be embarrassing.
The second is subtler: audience-aware predictions can drift toward or away from consensus. Some forecasters become more cautious to blend in; others become more contrarian to signal originality. Neither tendency has much to do with what you actually believe.
Neither of these effects makes you dishonest. They are natural consequences of social context. But they do mean that what you write down in a public forum may not be the cleanest signal of your raw private beliefs — and calibration is ultimately a question about the accuracy of that signal. For a fuller explanation of what calibration means and how it is measured, see calibration and the Brier score explained.
What changes when no one is watching
In a private log, the audience is only you. The prediction you record is closer to the prediction you actually believe — there is less friction between your real estimate and the one you write down.
This matters for calibration: the question of whether your high-confidence calls land right more often than your low-confidence ones. Calibration is a property of a signal. If social pressure systematically skews the confidence you record downward from what you privately believe, your calibration data describes a socially adjusted version of your reasoning, not the raw one. The record becomes harder to learn from, because the distortion is invisible.
A private journal keeps the signal closer to the source. That is not the same as making you a better forecaster. It means the record you accumulate is more directly diagnostic of your actual reasoning patterns, rather than a mixture of your reasoning and your audience-awareness.
Public forecasting has real and distinct strengths
It would be a mistake to read this as a case against public forecasting. Platforms like Metaculus, Fatebook, and PredictionBook offer genuine value that private journaling does not.
Community aggregation is one of the more robust observations in judgment research: pooled forecasts from many well-calibrated individuals tend to outperform individual forecasts, including expert ones. Metaculus's community questions are a direct application of this. If you want to benchmark your judgement against others, or contribute to a shared epistemic project, public forecasting is the right tool.
Social accountability also has real motivational benefits. Knowing others will see whether you were right can sharpen the discipline of writing down specific, falsifiable predictions with concrete deadlines — something that is genuinely harder to sustain in private. For teams, Fatebook's shared question format serves exactly this function. PredictionBook sustained an active community of forecasters for over a decade before it was retired and made read-only in late 2023.
These are real strengths. A private journal does not have them, and claiming otherwise would be misleading.
Two different tools for two different purposes
The honest summary is this: private and public prediction tracking are not better and worse versions of the same thing. They are optimised for different purposes.
Public forecasting is well-suited to collective accuracy, accountability, and benchmarking. The social dimension is a feature — it provides motivation, cross-checking, and aggregated wisdom.
Private forecasting is well-suited to individual self-knowledge — understanding your own reasoning with the least social friction between your real beliefs and what you record. The audience of one removes a layer of distortion that is hard to account for after the fact.
If you want to contribute to a community prediction market, compare yourself against others, or stay accountable to a team, a platform with social features is likely the better fit. The comparison of Hunch, PredictionBook, and Fatebook gives a direct feature-by-feature breakdown. If your goal is to understand your own reasoning — where you tend to be overconfident, which domains your instincts serve you well — a private log may give you a cleaner signal to work from.
About Hunch
Hunch is built specifically for the private case. It runs entirely in your browser — your predictions never leave your device, there is no account, and no one else can see your record. This is a deliberate design choice, not a limitation: the goal is to keep the log as private as possible, so the data you accumulate is yours alone and reflects your reasoning as directly as possible.
If you would like to start tracking predictions privately, you can do so immediately — no sign-up required.