Why Good Forecasters Aren't Afraid of Being Wrong

What each one is actually for

Self-Compassion Practice

Self-compassion practice is built to interrupt harsh self-criticism at the moment it happens, replacing it with something closer to what you'd say to a friend who'd just made the same mistake. It has two working parts: mindful self-kindness, actively choosing warmth toward yourself instead of the automatic harsh response, and what researchers call common-humanity framing — recognizing that struggling, failing, and falling short are things every person goes through, not evidence that something is uniquely wrong with you. Practicing it might look like literally pausing after a mistake and saying, out loud or in your head, something like 'this is hard, and everyone finds this hard,' before letting yourself react any further.

Superforecasting

Calibration training is built to close the gap between how confident you feel and how often you're actually right. The method is specific: you make explicit numeric predictions — 'I'm 70% confident this will happen' — about real, checkable outcomes, then you track, over time, whether things you called 70% confident actually happened about 70% of the time. If your 90%-confidence calls are only right 60% of the time, you're overconfident, and the training corrects that with repeated feedback rather than with a lecture about humility.

What you get out of pairing them

If you already practice self-compassion but haven't found a concrete arena to put it to work in, calibration training gives you one with an actual scoreboard — a place where the habit of not globally judging yourself over a setback gets tested against something real and countable, not just a vague bad feeling. If you already do calibration work, or anything that requires tracking your own error rate honestly — trading, forecasting, even estimating how long a project will take — self-compassion gives you a way to keep looking at your own miss rate without the part of you that hates being wrong quietly sabotaging the record, rounding a miss up to a near-hit, or just tracking less carefully on the weeks you did worse.

What Self-Compassion Practice rests on

• Acceptance reduces struggle randomised trials
  Allowing an internal experience to be present without resisting or acting on it reduces the psychological struggle that resistance would otherwise intensify.
• Common humanity reduces isolation observational studies
  Recognizing that suffering and imperfection are part of shared human experience reduces the sense of being alone in one's pain.
• Self kindness downregulates threat response randomised trials
  Responding to one's own setbacks with kindness instead of harsh criticism reduces the physiological threat response.

What Superforecasting rests on

• Decomposition improves estimate accuracy observational studies
  Breaking a complex uncertain question into smaller more tractable sub-questions produces more accurate estimates than judging the whole question at once.
• Explicit probability estimates improve calibration randomised trials
  Forcing a numeric probability rather than a vague verbal hedge makes a prediction falsifiable and trainable.
• Outcome feedback corrects judgment bias randomised trials
  Regularly scoring predictions against actual outcomes and reviewing the misses corrects systematic biases in judgment over time.

Where they actually meet

Both are structured around the same moment: discovering you were wrong about something. Calibration training treats that discovery as data to record accurately. Self-compassion treats that same discovery as an event to metabolize without turning it into an attack on yourself. Neither one denies that you were wrong — the forecast missed, the mistake happened — they're both about what happens in your head in the five seconds right after you find that out, just aimed at different outputs.

The principle underneath both

You can only get an accurate signal about your own performance from a source you're not simultaneously trying to protect yourself from. If being wrong feels like a threat to your worth, some part of you will start managing the threat instead of reporting the result honestly — softening a bad forecast in the retelling, or turning a specific mistake into 'I always mess this up' and shutting down the examination early to stop the pain. Removing the threat is what lets the honest signal through in both directions.

What that makes the real goal

If an undefended relationship to being wrong is the actual requirement, then what both practices are working toward is the same capacity: staying willing to keep looking at your own errors — in a forecast, in a decision, in how you treated someone — accurately and repeatedly, instead of needing to protect yourself from seeing them by hedging, minimizing, or looking away.

Other practices that serve the same goal

Several fields converge on this from different angles. Blameless post-mortems, standard practice in software engineering after an outage, are built specifically to separate 'what went wrong' from 'who's at fault,' on the theory that people investigate more honestly when they're not defending themselves at the same time. Carol Dweck's growth-mindset research finds a related pattern with skill rather than judgment calls: people who treat a failure as information about something improvable, rather than proof of a fixed trait, persist longer and, in some studies, actually improve more. Red-teaming and devil's-advocate exercises, used in intelligence analysis and strategic planning, try to build the same undefended posture socially — making it normal, even expected, for a plan to be shown wrong before it ships. Prediction markets do something similar at the level of a whole group rather than one person: because a wrong bet costs a specific, known amount rather than a vague hit to reputation, participants tend to report their real belief instead of the belief that makes them look best, which is the same shift self-compassion is trying to produce in a single forecaster's private log.

A tradition that already does this

Tetlock's own research on what separates superforecasters from everyone else points at something adjacent to this without naming it: the best forecasters scored high on what he called actively open-minded thinking, and treated a bad prediction as useful information rather than a personal indictment. He didn't call this self-compassion, and it wasn't taught as a practice — but the undefended relationship to being wrong that self-compassion trains directly shows up, unlabeled, in the group that was already best at calibration.

Using Self-Compassion to Keep Your Calibration Record Honest

The moment calibration training usually breaks down is quiet and easy to miss: you look back at a prediction you were very confident about, see that it was wrong, and instead of recording it plainly, some part of you finds a reason it 'basically' would have been right, or you just track less precisely for the rest of that week. Borrow the actual mechanics of self-compassion practice here, not just the sentiment. Right after you log a miss, especially a confident, embarrassing one, say the common-humanity part out loud: 'everyone who forecasts seriously has a bad month; this is one data point, not a verdict.' Then do the self-kindness part concretely: instead of berating yourself, ask what you'd say to a friend who just showed you the same miss, and write that down instead. This isn't about feeling better for its own sake. It's about keeping the record accurate enough to actually learn from, on exactly the entries where the temptation to fudge it is highest.

Using Calibration Training to Make Self-Compassion Concrete

Self-compassion can stay abstract — a nice idea you agree with in general but don't have a specific moment to practice — until you give it something exact to respond to. Borrow calibration training's habit of making a specific, checkable claim and then checking it. Once a week, write down one self-critical prediction you're carrying — 'I'll never get better at this,' 'I always freeze up in meetings' — as an actual falsifiable claim, the way a forecaster would state it, and then go check the record: how many times this month did that actually happen, versus how many times it didn't? Most self-critical predictions turn out to be miscalibrated in the same direction bad forecasters usually are — wildly overconfident, based on a handful of memorable instances rather than the full pattern. Treating your own harshest beliefs about yourself as forecasts to be checked, rather than truths to be accepted, gives self-compassion an actual mechanism instead of just a mood.

The Spreadsheet Row You Don't Want to Fill In

You made a call three weeks ago — 85% confident a deal would close by end of month — and it's now clear it isn't happening. The spreadsheet where you track every prediction is open on your screen, cursor sitting in the outcome column, and there's a real pull to leave it blank for a few more days, or write 'pending' even though you already know. You notice the pull and name what it actually is: not caution, but the same flinch that makes a bad forecast feel like a bad forecaster. You write 'no' in the column. Your confident miss now sits next to four other predictions from the same batch, three of which landed close to their stated odds. Looking at the row of five instead of the one, the miss stops looking like a verdict on your judgment and starts looking like exactly what an 85% confidence level should produce sometimes — about one time in seven, if your calibration is actually working.

When this is worth reaching for

The log gets softer right after a bad stretch — the language hedges, a week goes unrecorded, and opening the spreadsheet starts to feel like something to put off. That drift, honesty declining exactly when the numbers turn bad, is the condition worth watching for. A calibration record that has stayed accurate and consistent through both good runs and bad ones has nothing to fix here.

Where this does not apply

If your forecasts are inaccurate because of a skills gap — you're not breaking questions into checkable sub-parts, or not updating on new information — self-compassion won't fix that; it addresses the willingness to look honestly, not the technique itself. And if your self-criticism is severe enough to be a clinical concern rather than a performance habit, that calls for a therapist, not a forecasting spreadsheet. It's also not the right combination if you don't actually make trackable predictions in your work or life at all — self-compassion has plenty of other applications, but calibration training specifically needs a running record of numeric calls to attach itself to.

What the evidence does and does not support

Calibration is one of the more rigorously tested claims in this whole area — Tetlock's tournament data and subsequent replications show it's trainable with structured feedback, and the mechanism (numeric predictions checked against outcomes) is about as close to a lab result as this corpus gets. Self-compassion's effects on reducing self-criticism and improving psychological wellbeing are also well-supported across multiple meta-analyses. What's never been tested is the specific claim made here — that adding a self-compassion practice to a calibration habit keeps people tracking their own errors more honestly over time. That's a reasonable prediction built from two separately solid literatures, not a finding from either one.