記事
AIエージェントの研究、市場予測のバックテストと監査、プラットフォームのアーキテクチャ。
証明可能性
AI can generate a confident prediction in seconds — but can it prove the prediction came first? Here is how blockchain timestamping turns unfalsifiable AI output into an auditable record.
読了時間 4分6セクション
スポーツ
Can an AI model actually beat the bookmakers and betting exchanges? Here is what the vig, the closing line, and market efficiency really mean for machine forecasting — and what a genuine edge would require.
読了時間 4分5セクション
予測市場
Polls, models, and prediction markets all try to call elections. Here is how AI forecasting handles politics — base rates, small samples, calibration — and why honesty beats confidence.
読了時間 3分5セクション
モデルの落とし穴
Survivorship bias is the error of studying only the winners because the losers have quietly disappeared from the data. Here is how it distorts financial results and how to guard against it.
読了時間 4分5セクション
モデルの落とし穴
Backtesting runs a strategy against historical data to see how it would have performed. Here is what it really measures, why a great backtest is so easy to fake, and what an honest one looks like.
読了時間 4分5セクション
確率で考える
Ignoring how often something usually happens is one of the most common forecasting mistakes. What the base rate fallacy is, the classic example, and how good forecasters avoid it.
読了時間 3分5セクション
予測の採点
The Brier score grades a probability for a yes/no event. The CRPS does the same job for a full distribution over a number. Here is how it works and why it is a proper scoring rule.
読了時間 3分5セクション
証明可能性
We put our forecasting where it can be independently scored: NeuPortal has entered public AI forecasting arenas, shipped an honest benchmark, and measured a real improvement.
読了時間 3分5セクション
予測市場
Prediction markets turn scattered beliefs into a probability estimate. Here's the mechanism that makes them one of forecasting's most powerful tools.
読了時間 7分7セクション
予測市場
Why the average of many independent guesses often beats the sharpest lone expert — the statistics behind the wisdom of crowds and why prediction markets exploit it.
読了時間 4分5セクション
確率で考える
A forecast becomes useful only when you compare its probability to a price. Here is how expected value works, why positive EV is never a promise, and the calibration it depends on.
読了時間 4分5セクション
モデルの落とし穴
A model that looks spectacular on historical data is the easiest thing to build and the hardest to trust. What overfitting is, why it fools you, and the only test that cannot be gamed.
読了時間 4分5セクション
