Reading your reputation

Your profile shows several numbers, and they are not the same number dressed differently. This page teaches you to read your own file the way the leaderboard does — because the number that ranks you is not the one that feels most flattering, and knowing the difference is half of getting better.

Everything here builds on two earlier ideas: the drift-aware baseline (what each call is scored against) and confidence as a stake. If those are fresh, this page will click.

Skill vs accuracy — the two headline numbers

Two numbers describe your record, and they answer different questions.

• Skill is the hero number — your average, decayed skill score: how much you beat the baseline, across your resolved calls. This is the one the ranking uses. It can be negative. A permabull who's often "right" but never beats the baseline sits near zero.
• Accuracy (hit-rate) is the friendly number — the plain share of your resolved calls that came true. It's easy to read and nice to show, but it is never used for ranking, because "right" is cheap when the market was going your way anyway.

If you remember one thing: accuracy says how often you were right; skill says how often you were right about something the market didn't already expect. The leaderboard cares only about the second.

Your reputation is a career

There are no seasons and no resets — ever. Your reputation is built from your entire history of resolved predictions, because telling skill apart from luck takes a real sample. A handful of lucky hits won't put you on top; a long record of beating the baseline will. Two forces then shape the headline skill number:

Recency decay. Older results fade gradually, on a long half-life of about two years. A result from two years ago counts about half as much as a fresh one; a very old result counts little — but it never vanishes. So your standing reflects who you are now without throwing away the history that proves it.

Effective sample size. Because of decay, your calls don't all count equally, so the system tracks an effective number of calls — smaller than your raw count when your record leans on old results. This effective size is what feeds the caution below.

Why the ranking is deliberately conservative

Here's the part that surprises ambitious newcomers. The leaderboard does not rank your raw average skill. It ranks a lower bound of it — your skill minus a penalty that shrinks as your record grows.

The intuition: three lucky wins and a sky-high average shouldn't outrank a forecaster with a hundred calls and a proven edge. So a thin record is discounted — the ranking assumes you might have gotten lucky until you've shown enough calls to rule that out. Post more genuine calls and the penalty shrinks, letting your true skill surface.

This is the anti-gaming property in one line: volume can prove skill, but it can never fake it. Spraying predictions doesn't inflate your rank — the ranking is an average, not a tally, and every call (hit and miss) is permanent and counts. More calls just make your real level clearer.

The provisional label

Until you've resolved roughly ten predictions in a topic, your profile wears a provisional badge. It is a caution label, not a penalty — the conservative ranking already handles small samples mathematically; the badge just tells readers "not enough resolved calls yet to read this as a track record." It clears itself as your record fills in.

(Some finer-grained views need even more data before they'll commit to a verdict — for instance a per-time-frame breakdown greys out any bucket with fewer than five resolved calls rather than show a colour it can't stand behind.)

Standing, streaks, and the trajectory

A few more things you'll see:

• Standing places you within a topic's ranked field — your rank out of the field's size, shown as a positive-sum position (a percentile / top-slice) rather than a bare number, so a profile is a credential, not a scoreboard humiliation.
• Streaks — your current and best runs of correct calls — are a fun, human read of form. They are not the ranking; a streak of easy calls still barely moves your skill.
• The trajectory is your cumulative skill over time — an equity-curve style line. Its endpoint is a display total, useful for seeing the shape of your record, but again: the ranking is the decayed mean, not this running sum.

Read together, these tell the honest story: not who was loudest or luckiest, but who has been demonstrably right, over time, about things that weren't obvious.

Go deeper: How reputation works for the overview, or The scoring math for the exact formulas and constants.