I'm afraid your going to have to roll your own algorithm. This forum discussion includes an excellent post on Deal or No Deal, including a few actual examples from the American version of the show. If I were approaching this problem, I would start with the following calculations of the remaining cases:
1. Mean (aka the expected value), median, and some sort of calculation that groups suitcases into groups of values (because there should be a difference between three suitcases of $1, $2, and $1M versus $1, $500K, and $500K, even though their expected values are equal. Use these values to compute some sort of "TRUE VALUE" amount.
2. The purpose of the show is drama, so the goal is to keep a contestant from actually taking a deal until well into the show. Thus, during the first few rounds you need to make low-ball offers so that the contestant declines the deal. As the game progresses, your offer should approach, but never equal the TRUE VALUE amount.

As to other tidbits. There is probably some psychology at play based on the behavior of the contestant. For example, a boring contestant may get better offers so that they get off the stage quicker. A risk-adverse contestant may get worse offers because the banker can get away with low-balling them. These are all things that you probably can't program into your game. Also, this TRUE VALUE should rarely, if ever, equal the expected value because expected value is only designed for repetitive events, and Deal or No Deal is definitely not repetitive... contestants get only one shot, which tends to make them naturally risk adverse (the same principle applies to other big money games, such as Who Wants to be a Millionaire, where it took over a decade before somebody actually answered the million dollar question incorrectly).