Combinatorial auctions are auctions of multiple heterogeneous objects that allow bids on subsets of the objects, giving bidders the flexibility to express if the objects in a set are more valuable together than separate. This added flexibility makes it possible for the bidders to express a variety of preferences, but also complicates the problem they need to solve to find their bidding strategies. I study the problem a bidder has to solve in first-price sealed-bid combinatorial auction of two objects. I find that bidders should avoid bidding for overlapping sets, if their bids can be greater than the best competitive bids; however, this theoretical prediction fails to hold in controlled laboratory experiments.