Auctions have been celebrated as the best tool for efficient allocation of resources. However, no auction is good for all circumstances and every case needs an individual treatment. In this paper, we study the 2019 Russian crab quota auction, which raised 142 BLN rubles and became the largest auction in Russian history. We describe the auction format and compare it to alternative formats both theoretically and empirically. The multi-item auction was separate (to compete for each of the 40 items participants needed to register to a separate unit-item auction and to pay a deposit) and sequential (bidding for an item began only when bidding for the previous item was finished). As alternatives, we consider common auction (each participant registers for the number of items he wishes to buy and pays a corresponding deposit) and parallel auction (bidding for all items take place simultaneously and end only when bidding stops for each item). As comparison criteria we use (1) efficiency – items are allocated to those who value them most, (2) auctioneer's revenue, and (3) manipulability – participants' ability to win by using sophisticated strategies. (1) and (2) are standard comparison criteria, while (3) is new and we propose it as a new tool for assessing the auction's robustness against the participants' undesirable strategic behavior such as lobbying, espionage and collusion. To do that we first define so-called straightforward strategies – bidding up to one's value (like truth-telling in direct auctions), and then estimate the regret of using the straightforward strategy as opposed to using an optimal (sophisticated) strategy. The higher the regret, the more manipulable is the auction. As a result, we recommend using the common parallel auction, which has a few advantages compared to separate sequential auction used in 2019: it is more efficient, less manipulable and more robust to mistakes in setting the reserve price.