Phân tích chuyên sâu về AI trading, phân tích thị trường và tương lai của DeFi.
Browse collections →Nothing found. Try a different query.
Why matching volume distributions cannot validate synthetic LOB data: LOB-ID, exact counterexamples for depth and time, and a separate acceptance test for execution.
How pruning a scenario tree preserves average wealth while hiding drawdowns and Expected Shortfall. An exact Python example and checks for cost-aware rebalancing.
Does LLM debate add information? Recent research, an exact correlated-vote example, and a protocol comparing debate with independent forecasts under the same budget.
Test whether a strategy can survive its path: account-level collateral, maintenance margin, intraday prices and transfer timing, with a reproducible Python example.
How to validate generated option surfaces: penalties versus guarantees, total variance, executable butterflies, dynamic consistency, and a separate hedging acceptance test.
Shadow-PPOV предлагает ставить лимитную заявку вслед за наблюдаемой чужой: без модели стакана и прогноза fill, но с жесткими требованиями к order-ID, отменам и задержке.
L2 показывает объем на цене, но не FIFO-очередь. Работа о частичной идентификации исполнения измеряет, насколько правила распределения отмен меняют результат пассивного бэктеста.
Что меняется, когда финансовую foundation-модель дообучают данными из будущего: версии разных лет, point-in-time эталон, 18 из 20 ухудшений MSE и почему Sharpe 35 не лечится статистикой.
Почему спред - не вся экономика котировки бинарного контракта: риск расчета, перекос спроса, CLOB, AMM и границы, за которыми ликвидность перестает быть бесплатной услугой.
Почему финал события, решение оракула, запись payout в протоколе и погашение токена - разные часы, и как этот разрыв превращается в капиталовый и арбитражный риск.
Почему 7.6 млн комбинаторных объектов не равны 7.6 млн независимых рынков: три коллекции, 83 701 примитив, эффективное число около 720 и границы activity-метрик.
Практический разбор prop challenge: как читать лимиты, считать вероятность прохождения и выбирать размер позиции. Учебный пример, проверяемые расчеты и Python без зависимостей.
We ran a trend strategy across 15 crypto pairs, checked costs and access to future candles, and tested Jev separately. Where did profits appear, and why can't we promise them in live trading yet?
Return-direction classifiers are imbalanced problems, and 0.5 is the wrong decision threshold. Comparing scale_pos_weight, focal loss, and precision-constrained threshold optimization on crypto data — plus the engineering differences between XGBoost, LightGBM, and CatBoost.
DCC-GARCH tells you when crypto dependence tightens. Transfer entropy tells you which way it points. A directed information-flow measure, its null calibration, and an honest account of what it does and does not add over average pairwise correlation.
Reconstructing trade direction from OHLCV, bar history, or venues without an aggressor flag — the classic rules, the stale-quote problem, bulk volume classification, and how to measure them against crypto's free ground truth.
Sequence models fed tick data still assume regular spacing. Three ways to tell a Transformer when a tick actually happened — learnable-timescale continuous encoding, ODE-RNN latent state, and delta_t as a plain feature — and the ablation that decides between them.
This series has priced the selection route to a false edge — DSR prices the winner, PBO prices the search. Neither touches confounding: the strategy that made money because volatility doubled the week you deployed it. The Synthetic Control Method builds a weighted counterfactual from a donor pool of untouched instruments and gives you a falsification criterion and a placebo p-value.
How to evaluate a predictive distribution honestly — CRPS as a proper scoring rule, the PIT histogram as a calibration diagnostic, and DeepAR sampling in GluonTS.
