Weekly research digest
| AI Quant portfolio | AI Quant | S&P 500 |
|---|---|---|
| Week to 23 Oct | -0.1% | -0.4% |
| Since inception | +11.9% | +13.9% |
| Since inception, at market β | +44.0% | +13.9% |
| Sharpe | 2.69 | 1.62 |
| Beta | 0.27 | 1.00 |
At market β restates the record at the market's beta — the same strategies held at 3.70× the exposure: the return is multiplied by that factor, the drawdown re-measured on the scaled line, and Sharpe is unchanged by it. Arithmetic on the record, not a second track record — no leverage was used and no financing cost is modelled.
Each month, the strategy trades a capped, long-only minimum-variance portfolio targeting the cross-sectional mean estimated return. It also solves a higher-target benchmark on the same data for diagnostics, but does not trade that benchmark.
we backtested it — Sharpe 0.07 · quant relevance 0.98
The paper derives causal position-sizing filters for return forecasts under inventory risk and transient trading costs. Its results are theoretical, not a reported backtest. A daily US-equity implementation is feasible, but its assumed price-impact kernel cannot be validated from bar data alone.
we backtested it — Sharpe 0.48 · quant relevance 0.98
The study directly tests long-only ETF allocation using a tail-sensitive Rachev-ratio objective with historical, TimeGAN-generated, and hybrid scenarios. Generated scenarios distort tails and dependence, and paired inference does not establish a reliable investment-performance advantage. The method is buildable, but available prices require a later evaluation start after accumulating the 756-day training window; the original 2007–2025 dataset cannot be reproduced in full.
we backtested it — Sharpe 0.18 · quant relevance 0.97
The paper derives dynamic allocations that balance expected logarithmic growth against Range Value-at-Risk, producing state-dependent terminal payoffs and replicating strategies. It is directly relevant to portfolio risk management, but its model-based numerical illustrations are not evidence of realized trading performance. A daily implementation is backtestable, although it cannot reproduce exact continuous-time replication.
quant relevance 1.00
The study constructs daily cross-sectional equity portfolios using continuous news-sentiment scores extracted from language-model label probabilities. It reports improved risk-adjusted returns relative to discrete sentiment labels and FinBERT benchmarks. The mechanism is implementable with available US equity news and prices, although testing our news feed would not reproduce the original dataset exactly.
quant relevance 1.00
LineageEvo uses lineage-conditioned LLM mutation and crossover, together with lineage-balanced search allocation, to discover predictive equity factors. It reports improvements in factor predictiveness and portfolio performance on CSI 500 and S&P 500 stocks. A US-only implementation is buildable, but the CSI 500 experiments and exact historical S&P constituent universe cannot be replicated with the available data.
62 new papers written up in this period. Open the Quant Paper Radar →
Backtests are hypothetical research results; forward-portfolio figures are a live record. Neither is a guarantee of future performance. Research and education only: not individualised investment advice, not an offer to sell or a solicitation of an offer to buy any security.