If I were bidding at such a sealed-bid auction, I would try to estimate (a) how much revenue I could get from the item if I got it, and (b) what is the minimum profit that I would be happy with on my investment. Then I would bid for the difference (a) minus (b).

For example, I am currently pessimistic about the chances of the price ever getting much higher than now, and expect a downtrend in the short term.  Therefore, if I were to get those bitcoins I would try to sell them right away for the best price, as fast as I could.  If the auction were today, I would expect to get maybe 550 USD/BTC average from selling 3000 coins on the exchanges over the next week.  Thus, I would bid for 500 USD/BTC to have an expected profit of 150'000$ with an investment of 1'500'000$.

By the same reasoning, someone with a more optimistic outlook may bid higher than market.  If he expects BTC to be 10'000 USD next year, then he can bid at 5'000 USD/BTC and still expect to make a nice profit.  But if such person existed, he would be buying all coins in the market now, at any price.  Obviously the optimists are all out of money now.

I suppose that one can build more sophisticated probabilistic models to find one's optimum bid, given one's expectations about future bitcoin prices.  However, one must assume a very broad probability distribution for the other people's bids, so I doubt that one can do much better than the simplistic method above.