Author
Trend and reversion are one process split by the sign of beta. Regress a series on its own lag and OLS understates beta's magnitude in both cases, because the lagged regressor shares error terms with the target. You end up believing in less trend and less reversion than the market carries.
Replace the volatility term with volume in most alphas and the backtest barely moves, because they ride the same information clock. Feeding a model both is double-counting one factor. Keep one scale, and add their ratio, Amihud illiquidity, as the residual that actually carries new information.
Pick the model by interaction strength: ridge when near-additive, XGBoost when interactions dominate. XGBoost also handles NaNs natively, avoiding imputation lookahead; else use an IC-weighted ensemble.
A decision tree mines conditional alphas by carving feature space into boxes. It picks each split to maximize a similarity gain and predicts the mean return per leaf. Stop early or it memorizes noise.
One tree is fragile: a small data change flips its top split. Bagging averages many trees to cut variance; boosting fits each new tree to the residual to cut bias. XGBoost and LightGBM are boosting.
Relabel every long as a short and a return model should flip its sign. Linear models get this free through their weights; a tree relearns each split's mirror, so engineer the symmetry back.
A large order moves price by construction, so it manufactures the move it seemed to predict. Size needs capital and a firm view, which reads like information on the tape, so the market prices large flow as informed because it cannot tell you apart. Read the permanent impact to separate knowing from
The seller is easy to picture, so everyone narrates the seller after a crash. Price is set at the margin: the question that forecasts anything is who the marginal buyer is, and where they step back.
Got forecasts on different clocks? Don't pick the biggest. Normalize each to edge-per-second on its markout curve (30 bps/60s = 0.5 vs 2 bps/1s = 2) and average the curves to blend fast and slow edge into one.
Volatility predicts the size of a move, not its direction, so it can't trade alone. Multiply it by momentum and it works, until extreme vol gains its own sign, down for equities and up for gold.
The Fourier transform fails on price but nails a robot, because a wash bot or metronomic TWAP is the clean periodic signal price never is. Bin trades, FFT the counts, and a sharp spike is automation. The catch: your own equal-interval TWAP makes that spike, so randomize it.
The cheapest way for an MFT shop to enter a small position is to act like a market maker: feed your quoting engine a phantom short so it skews and fills you long, collecting the spread instead of paying it.