\begin{table}[htbp]\centering\small
\caption{Main (held-out): QLIKE relative to Random walk (trailing RMS)}
\label{tab:main_cmp_qlike_heldout}
\begin{adjustbox}{max width=\textwidth}
\begin{tabular}{lcccc}
\toprule
Model & $H=5$ & $H=15$ & $H=30$ & $H=60$ \\
\midrule
Seasonal volatility & -10197881952792.29\% (-23.6) & -9820053262498.37\% (-22.7) & -8492739370379.95\% (-12.6) & -6828601646415.30\% (-6.5) \\
EWMA-$t$ & -13985074515913.91\% (-37.0) & -12209923172830.35\% (-27.0) & -9897403040026.40\% (-16.8) & -6799201127724.36\% (-8.9) \\
HAR-RV & -11561898079453.18\% (-33.6) & -15457887605396.15\% (-37.3) & -14272360974230.15\% (-26.1) & -13273707831644.94\% (-19.4) \\
GARCH(1,1)-$t$ & -7388166947421.55\% (-16.5) & -6696713758741.78\% (-16.0) & -4690422515322.66\% (-10.4) & -865268517734.10\% (-1.5) \\
Ridge & +386478985928.25\% (0.5) & -14190659869919.63\% (-23.7) & -14022698526759.01\% (-20.3) & -13322533504300.98\% (-16.0) \\
Elastic net & +374095534458.66\% (0.5) & -14152759381067.28\% (-23.2) & -13988559408128.51\% (-19.9) & -13300257261994.19\% (-15.8) \\
LightGBM & +1297050835657109.75\% (2.7) & -13398286632246.19\% (-13.6) & -14995763686121.92\% (-27.9) & -14475278961493.17\% (-23.8) \\
MLP (1m) & -17034644963943.96\% (-19.2) & -15946370199621.97\% (-19.5) & -14535364269361.72\% (-13.7) & -13005658787725.62\% (-8.5) \\
TCN (1m) & -18636537181776.08\% (-55.4) & -16971749949659.03\% (-41.8) & -15714216179967.48\% (-33.8) & -14554204854126.64\% (-26.7) \\
MRPT-S (1m+5m+30m) & -19163873503843.14\% (-52.8) & -17894759635105.88\% (-52.3) & -16820059068029.27\% (-44.7) & -15964761744030.21\% (-35.5) \\
MRPT-S (1m only) & -18764124938752.46\% (-51.9) & -17287778159720.96\% (-38.4) & -16054207150684.29\% (-31.3) & -14954052048596.54\% (-25.1) \\
MRPT-B (1m+5m+30m) & -18760431723724.96\% (-54.8) & -17487271173252.59\% (-44.0) & -16348978484358.31\% (-35.0) & -15378874335569.79\% (-27.6) \\
\bottomrule
\end{tabular}
\end{adjustbox}
\begin{minipage}{0.95\textwidth}\scriptsize Relative change in mean QLIKE versus the reference (negative = better); Diebold--Mariano $t$-statistic with HAC variance in parentheses.\end{minipage}
\end{table}