Platforms that match contestants by rating also decide whether contestants see their opponents' ratings. We study this choice after assortative matching. Contestants know their own Gaussian ratings under both policies, and the platform maximizes aggregate effort holding matching fixed. The policy gap between concealment and disclosure splits into a Jensen term, governed by marginal-cost curvature, and a strategic loss. Under concealment, effort depends on one's own rating: by mean reversion, a contestant closer to the mean rating expects a closer match and works harder, and the resulting responses widen performance gaps within matches. We obtain three main results. First, near perfect matching the Jensen term is second order and the strategic loss fourth order, so positive marginal-cost curvature favors concealment, while quadratic costs favor disclosure even though they would be neutral without own-rating information. Second, with quadratic costs, disclosure raises effort at every matching correlation when costs are sufficiently high or ratings sufficiently uninformative. Third, for nearly quadratic costs with positive marginal-cost curvature, concealment is preferred within a band around perfect matching whose width is proportional to the square root of the ratio of marginal-cost curvature to the curvature of marginal winning returns at a tie. The band exists because the strategic loss builds up through a chain of small effects and fades faster than the Jensen term as matching becomes precise. Numerical equilibria show that the reversal can occur at moderate correlations and without an effort bound.