In high risk projects both individual task costs and durations as well as task dependencies are hard to evaluate. Some kind of modeling is required.
One of the simplest modeling methods is Monte Carlo, which relies on repeated random sampling to obtain numerical results - e.g. for project cost and duration. The underlying concept is to use randomness to solve problems that might be deterministic in principle. Results are accurate enough when number of sampling rounds is sufficiently high.
In project planning this would mean calculating the project several times. For each round each project task gets random cost and duration values, which are sampled from task specific distributions. When number of simulation rounds is high enough can total project cost and duration distributions be considered accurate enough.
The attached picture shows how random costs from different rounds are distributed between minimum (in the example 100€) and maximum (in the example 1200€) values, when task cost follows so called beta-PERT distribution. Random costs fall typically close to the most probable cost in the distribution. Same logic applies to task duration.
The result after large number of repetitions will be the probability distributions for total project cost and duration. These distributions are usually shown with histograms and cumulative probabilty distribution curves. Histogram shows how many times total project cost or duration falls into certain interval and cumulative distribution curve shows the probability of staying below a certain total project cost or duration.
Attached picture shows how to view results for total project cost.
MonteCarloProject is a project management application with Monte Carlo simulation. It has free basic features and can be extended with simulation plan.