GPUSHyperscale Data, Inc.— quant rating & factor grades

How the rating works

• Four factor percentiles. Value, Growth, Quality and Momentum are each scored as a percentile (0–100) against Tickwind's tracked US universe — the share of covered stocks this one ranks at or above on that factor's sub-metrics: P/E, P/B, P/S for value (inverted, so cheaper ranks higher); revenue & earnings growth; ROE, ROIC, EBIT margin, Piotroski F for quality; and 1-year total return for momentum.
• Insufficient, never imputed. A sub-metric a stock lacks (or that too few peers report) is skipped — never filled with 0 or a median — and a factor with no computable sub-metric is left absent, never shown as 50.
• A factor needs a real population. A percentile is withheld unless at least 8 tracked stocks carry that sub-metric — ranking against a handful of peers is noise, not a statistic.
• Composite = the mean of the available factor percentiles. A 1-leg composite (only one factor computable) is a far weaker read than a 4-leg one, so the leg count is disclosed.
• Grade = 20-point bands of the composite percentile: A ≥ 80, B ≥ 60, C ≥ 40, D ≥ 20, else F. A plain relabeling of where the percentile falls, lower-inclusive (exactly 80 → A).
• Rating = the same bands, relabeled: Strong Buy ≥ 80, Buy ≥ 60, Hold ≥ 40, Sell ≥ 20, else Strong Sell (A ⇔ Strong Buy … F ⇔ Strong Sell).
• It is RELATIVE, by construction. The rating ranks this stock within the covered population, so the bottom fifth is always Sell-tier no matter its absolute fundamentals. It is a quantitative model over the disclosed population — not personalized investment advice.

More on GPUS