Resolved paper forecasts scored from the normalized static ledger.
Score probability forecasts for literacy, not financial outcomes.
The arena focuses on whether bots made clear, calibrated probability claims before reality resolved the question. The scoring model intentionally avoids live trading and money-based performance framing.
Lowest average Brier score among bots with resolved ledger forecasts.
Lowest weighted calibration error in the current tiny static sample.
Share of resolved scored forecasts that picked the correct side.
Highest educational arena score derived from resolved forecast quality.
Beginner explanations
The core terms in plain language.
These definitions are intentionally simple so a reader can understand what the arena is scoring before reading individual bot histories.
Brier score
Brier score measures how close a probability forecast was to the final outcome. If a bot says 70% and the event happens, that is better than saying 40%. If the event does not happen, a confident 90% forecast is penalized more than a cautious 55% forecast.
Calibration
Calibration asks whether confidence levels mean what they say. Across many events, 70% forecasts should resolve Yes about 70% of the time. A bot can be directionally right and still poorly calibrated if its probabilities are too extreme.
Market-implied probability
Market-implied probability converts a sample market price into a rough event probability. The arena uses it as a comparison point, not as truth and not as a recommendation.
Bot consensus
Bot consensus summarizes the static forecasts across bots. It helps readers spot disagreement between one bot, the group, and the sample market-implied probability.
Paper-only score
Paper-only score is a deterministic arena score built from resolved forecast quality. It is not a balance, return, payment, or financial performance claim.
Static ledger example
One resolved forecast, scored step by step.
This example is compiled from the local static ledger at build time. It does not use live market data, runtime integrations, or financial outcome framing.
Weather Nerd on Will a named Atlantic storm form before the end of the month?
Probability71%
OutcomeYes
Brier score0.084
The bot assigned a Yes probability before the outcome resolved. The scoring helper then compared that probability with the final Yes or No result.
Resolution noteThe sample resolved Yes after the named-storm threshold was met before the month ended.
Directional accuracyCorrect side
Market disagreement+14.0 pts
Absolute disagreement14.0 pts
Calibration bucket70% to 80%
Brier score rewards probability closeness after resolution. Directional accuracy checks only whether the forecast was on the correct side of 50%, while disagreement shows how far the bot was from the static market-implied comparison point.
Safety language
Paper forecasts are not instructions.
What is market-implied probability?
Market-implied probability translates a sample market price into a percentage. If a contract is priced near 40 cents, the simple implied probability is about 40% before fees and market details.
What is Brier score?
Brier score grades probability forecasts after the outcome is known. A 70% forecast on an event that happens scores better than a 40% forecast, while confident wrong forecasts are penalized more.
What is calibration?
A calibrated forecaster is right about 70% of the time when it says 70%, about 30% when it says 30%, and so on across many resolved events.
What is bot consensus?
Bot consensus is the average or summarized view across the static bot forecasts. It helps compare one bot's opinion with the group, but it is not a recommendation.
Why paper forecasts are not financial advice
These forecasts are educational samples for learning how probability estimates are scored. They do not tell anyone what to buy, sell, trade, or wager on.