Where the numbers come from
Every number on the Decision Index is computed from the contract's own events: each entry, each settlement and each claim since ClefRounds was deployed. The scoring is the same integer math the contract pays with, so the index and your payouts always agree. The page refreshes every minute.
The head shows the totals: wallets, entries, settled rounds and ETH staked.
The ranking
Wallets are ranked by mean score: the average card score over their settled rounds, where 1.0000 is a perfect card. Ties go to the wallet with more scored rounds, then to the higher net result. You can sort by any column.
- Rounds: rounds entered, settled or not.
- Staked: the ETH put in across those rounds.
- Mean score: the average card score, with a thin bar to compare at a glance.
- Skill vs reference: how much better the wallet scored than the Clef reference on the same rounds. It appears only while the house entry is running, because it needs the reference to have played.
- Net: payouts after fee minus stakes, for settled rounds.
The 24 h, 7 d and All time switch keeps only rounds whose hour started within that rolling window. It moves with the clock rather than resetting at a fixed time.
Your stats
Connect a wallet and a Your stats panel opens above the table. Without one, a single line invites you to connect. Each figure follows the period switch:
- Rounds played: every round the wallet entered; a note shows how many are scored when some are still running.
- Average score: the wallet's mean card score, the number it is ranked by, next to its rank.
- Best round: the highest card score, with a link to that round's recap.
- Current streak: consecutive hourly rounds entered, counted back from the newest entry. It stays alive while that newest entry is the open round or the hour in progress; miss a round and it shows the last round you entered instead.
- Calibration error: how far your probabilities sit from how often things happened, explained below. Lower is better and 0 is perfect.
- Net: your result in ETH after fees.
The reliability diagram
Each settled entry contributes every option it priced: nine forecasts from a weekday card, six from a weekend card. The diagram groups them into ten bands by the probability given (0 to 10%, 10 to 20% and so on) and plots, for each band, the average probability against how often those options came true by the chain's answer. A split answer counts as a share, so a 75% Yes counts as three quarters of a hit.
A calibrated forecaster sits on the dashed diagonal. Points above it are underconfident (things happened more often than said), points below it overconfident. The calibration error is the average gap between the two, weighted by how many forecasts each band holds. With a wallet connected your own curve is drawn beside everyone's.
An even split everywhere gives a calibration error of 0 but a middling score: calibration and sharpness are different things, and the score rewards both.
Recent rounds
Below the table, Recent rounds shows the chain's answer for the latest settled hours as probability rows. A clear move shows 100% on one option; a close call shows the split. Each round number and each Round recap link opens that round's page.
Round recaps
Every round has a page at /round?id= plus its number. For a settled round it shows:
- the questions and the chain's answer to each, as probability rows;
- the measured move of each asset, from the 300 second average price at the start to the one at the end, in percent and in ticks;
- the number of cards and the pot;
- the top cards by score, with short wallet addresses, stake, payout and net.
An open round shows its questions, the pool so far and a countdown to the lock, with a link to the app. The hour in progress shows where each question leans with the prices so far. Share copies the round's link, and the arrows step to the round before or after.