New quantitative finance research with out-of-sample results: trading, portfolio construction, risk and market microstructure.
This issue leans toward market microstructure: one paper checks Polymarket trade signs against on-chain settlement, another finds EURO STOXX 50 futures order-book states form a continuum rather than discrete regimes, and a third shrinks transformers for order-book forecasting. Also in the mix are trading and portfolio papers, a credit-card LGD comparison, and two items flagged for lacking market data or out-of-sample tests (items 3 and 6). The Polymarket sign-validation paper and the order-book continuum paper are the most worth reading.
Validates Polymarket trade-print sides against the taker leg of on-chain settlement: 92.1% to 100.0% of prints match on twelve selected days, with exact side agreement on 24.5 million pairs. Balanced accuracy is 0.942 for Lee-Ready, 0.768 for the tick test and 0.670 for bulk classification, and Lee-Ready raises effective spread by 0.429 cents per share.
Analyses EURO STOXX 50 futures order-book data over 987 days (2022 to 2025) and finds a continuum of market states rather than discrete regimes. One latent coordinate captures 83.66% to 85.27% of variation and tracks VSTOXX. Walk-forward tests show out-of-sample strength at the macro tier only, with pooled R2_OOS peaking at 0.7483 near ten minutes.
TradeGrad uses LLM textual gradients with a Cross-Period Robust Objective to optimise trading strategies. It reports the best in-sample and out-of-sample results across four settings in Chinese A-share and U.S. equity markets; the Chinese cross-sectional strategy shows 27.99% annualised return, 12.19% maximum drawdown and Sharpe 1.63.
Streams that already watch this field. Follow one as it is — it costs nothing extra.
Fintech startups raising money — seed to growth rounds, with the amount, the investors and what the company builds.
Rate decisions, policy statements and leadership news from the Fed, the ECB, the Bank of England and the Bank of Japan — from the banks themselves.
Pre-seed, seed and angel rounds of up to $2M, worldwide and across sectors: who raised, how much, who led, and what they are building — the deals that never make the big headlines.