Closes #27. Extends the existing alpha-101 app to also implement WorldQuant Alpha #2: -1 * correlation(rank(delta(log(volume), 2)), rank((close - open) / open), 6) Shared pipeline changes (affecting Alpha #1 too, verified non-breaking): - random_market_data gains a `volume` field (uniform [1, 100] per tick). - market_data + mv_market_data plumb volume through. - v_bars now exposes `open` (earliest price in bucket), `close` (latest price), and `volume` (sum of tick volumes). Alpha #1's downstream chain works unchanged. New views for Alpha #2: - v_features_2: per-stock per-bucket intraday_ret = (close - open) / open, log_vol_delta_2 = log(volume_t) - log(volume_{t-2}), and bucket-to-bucket returns (plumbed for the backtest). - v_ranks_2: per-bucket cross-sectional rank of BOTH features (using the same mean-zero rank pattern as Alpha #1, computed once per bucket via group_array + array_sort + array_first_index for each feature). - v_alpha_2: per-stock 6-bucket rolling Pearson correlation between the two rank series, negated. Manual covariance / variance computation via array_reduce('avg', array_map(...)) over lags(_, 0, 5). - v_backtest_2: same shape as v_backtest, branched on `strategy` config (linear | sign). Two new dashboards: - "Alpha #2 Live" (volume per bucket, leaderboard, alpha over time) - "Alpha #2 Backtest" (t-stat tile, summary, per-stock, portfolio PnL) End-to-end live-verified: install of the rebuilt .tpapp produces both alphas; v_alpha_2 values bounded in [-1, 1]; v_backtest_2 emits sensible pnl on each (time, stock) row. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

alpha-101 : more alphas #27

Description

Metadata

Metadata

Assignees

Labels

Type

Fields

Projects

Milestone

Relationships

Development

Issue actions