摘要：Cases are won and lost in discovery, yet discovery draws little academic attention. Most scholarship focuses on how much discovery to allow, not on how courts decide discovery disputes-which, unlike trials, occur in most cases. The growth of computer data-e-mails, lingering deleted files, and so forth-increased discovery cost, but the new e-discovery rules just reiterate existing cost-benefit proportionality limits that draw broad consensus among litigation scholars anti economists. But proportionality rules are impossible to apply effectively; they fail to curb discovery excess yet disallow discovery that meritorious cases need. This Article notes proportionality's flaws but rejects the consensus blaming bad rulemaking or judging. Rather, proportionality requires impossible comparisons between discovery value and cost before parties gather the evidence. Like other arguments that procedural rulings should depend on case merits, this Article notes how discovery has more probative value in close cases-yet a case's merits are unclear during discovery because the court cannot yet examine all the evidence. In game theory terms, parties with discovery disputes cannot convey case merit credibly; courts have too little information, so low-merit parties can claim high merit, and courts are compelled to act as if all cases of a similar type warrant similar discovery. In this "pooling equilibrium," ruling the same on all cases in the "pool," regardless of merit, is courts' best strategy but a suboptimal one, yielding too much discovery in low-merit cases, too little in higher-merit ones. Thus, the quest for better discovery has disappointed not because of bad rules or decisions, but because courts and parties are stuck in a pooling equilibrium with information-timing circularity: optimal evidence gathering requires merits analysis, which in turn requires evidence gathering. As a solution, courts could defer close decisions on possibly useful but costly evidence until meritorious cases separate from the pool, turning pooling into separating equilibria. Summary judgment can be this separation: cases going to trial after summary judgment not only have higher average merit than the pool of all filed cases, but are disproportionately likely to be the sort of close calls in which juries struggle to reach verdicts. No one yet has proposed post-summary judgment discovery to redress the costly discovery dilemma because summary judgment typically occurs only after all discovery, but high-cost evidence can be an exception to that usual sequence: cases surviving summary judgment are close calls warranting more fact gathering, so some costly discovery regularly denied should be allowed
