spb/wp12_uqo

1\begin{tabular}{lccc ccc c}2\toprule3 & \multicolumn{3}{c}{Net of 5 bps} & \multicolumn{3}{c}{Sharpe at cost (bps)} & \\4\cmidrule(lr){2-4}\cmidrule(lr){5-7}5Model & Sharpe & \makecell{Ann.\ ret.\\(\%)} & Turnover & 0 & 10 & 20 & \makecell{Utility gain\\vs HAR (bps)} \\6\midrule7\multicolumn{8}{l}{\itshape A. GARCH family}\\8\quad GARCH & 0.35 & 4.7 & 0.033 & 0.39 & 0.32 & 0.24 & 17 \\9\quad GJR & 0.35 & 4.7 & 0.031 & 0.39 & 0.32 & 0.25 & 10 \\10\quad EGARCH & 0.34 & 4.6 & 0.041 & 0.39 & 0.30 & 0.21 & 3 \\11\quad RealGARCH & 0.33 & 4.5 & 0.062 & 0.40 & 0.26 & 0.13 & -13 \\12\multicolumn{8}{l}{\itshape B. HAR family}\\13\quad HAR & 0.34 & 4.3 & 0.054 & 0.40 & 0.27 & 0.15 & --- \\14\quad HAR-CJ & 0.34 & 4.4 & 0.058 & 0.40 & 0.27 & 0.14 & -1 \\15\quad SHAR & 0.31 & 4.0 & 0.078 & 0.39 & 0.22 & 0.05 & -42 \\16\quad HARQ & 0.31 & 4.5 & 0.087 & 0.40 & 0.21 & 0.03 & -79 \\17\quad LogHAR & 0.32 & 4.3 & 0.073 & 0.40 & 0.24 & 0.07 & -31 \\18\multicolumn{8}{l}{\itshape D. Machine learning --- extended information set}\\19\quad Ridge-X & 0.30 & 4.1 & 0.085 & 0.40 & 0.21 & 0.01 & -47 \\20\quad LASSO-X & 0.31 & 4.2 & 0.082 & 0.40 & 0.21 & 0.03 & -41 \\21\quad RF-X & 0.31 & 4.3 & 0.078 & 0.39 & 0.22 & 0.05 & -52 \\22\quad XGBoost-X & 0.28 & 4.1 & 0.106 & 0.39 & 0.17 & -0.06 & -90 \\23\quad LightGBM-X & 0.26 & 3.9 & 0.118 & 0.39 & 0.13 & -0.12 & -121 \\24\multicolumn{8}{l}{\itshape E. Deep learning and pooled}\\25\quad LSTM & \textbf{0.38} & 4.9 & 0.012 & 0.39 & 0.36 & 0.33 & 33 \\26\quad Transformer & 0.35 & 4.7 & 0.004 & 0.35 & 0.34 & 0.33 & -19 \\27\quad Pooled-LGBM & 0.29 & 4.2 & 0.105 & 0.41 & 0.18 & -0.05 & -69 \\28\multicolumn{8}{l}{\itshape F. Forecast combinations}\\29\quad Comb-Mean & 0.34 & 4.5 & 0.052 & 0.40 & 0.28 & 0.16 & 10 \\30\quad Comb-InvMSE & 0.33 & 4.3 & 0.056 & 0.39 & 0.26 & 0.13 & -5 \\31\bottomrule32\end{tabular}33