✓ BetCheckEXPERIMENTAL · PLAY MONEY

TRY THE IDEA. SEE THE COST.

Would your BTC
picks pay off?

Pick a model and replay real historical Kalshi Bitcoin markets. No account, API key or waiting required.

Historical practice only. New experimental models have mixed results: FixedAnchor makes no bets at standard costs, DirectionalAnchor loses on one bet, and a timing stress test shows a small, unproven profit. Positive returns before costs do not mean a profitable strategy. BTC live automation is not enabled.

Loading model results…

Choose your BTC model

New experiments: June 28–July 20, 2026 · 527 contracts / 23 days. Legacy: September 2026 · 2,830 contracts. Different cohorts must not be ranked against each other. All dates previously inspected; $1,000 initial paper balance.

How this works & what the scores mean

EXPERIMENTAL: FixedAnchor corrects market odds using minute order flow and exchange basis. DirectionalAnchor Joint adds signed price and volatility context. Delayed MicroFlow applies a 60-second delay to microdata as a timing stress test; its apparent profit is not reliable evidence of an edge. These three load frozen models from four chronological training folds. They share one deterministic trading policy, not separately trained trading models.

BTC Signal and BTC Regime learn from January–April 2025 Binance prices. We test on September Kalshi outcomes using Coinbase price inputs. BTC Volatility uses recent volatility directly. These price sources differ from Kalshi’s settlement index.

Each prediction asks whether the price will finish at or above the contract target near the end of a 15-minute market. It does not predict whether the price touched a target at any point.

Lower Brier and log loss are better. Compare each against the market at the same execution timestamp. Legacy models underperform market prices. New experiments offer small, uncertain forecast differences; better scores do not guarantee profit.

The replay refreshes forecast inputs at the execution quote, 5 seconds after the recorded decision. Fees use an assumed 0.07 coefficient. Fills have no verified historical depth. Trades risk up to 1% of balance, with 3% shared open exposure per BTC day, 20% overall, and at most 100 contracts per trade.