public SimulationManager(DataModel dataModel)
{
// Set up simulations
// Stocks
stocksSimulation = new StocksSimulation();
stocksProfile = new RunProfile(InvestmentType.Stocks, dataModel);
// Bonds
bondsSimulation = new BondsSimulation();
bondsProfile = new RunProfile(InvestmentType.Bonds, dataModel);
// Savings
savingsSimulation = new SavingsSimulation();
savingsProfile = new RunProfile(InvestmentType.Savings, dataModel);
}
public Trial[] Run(RunProfile withProfile)
{
if (withProfile.StepDistribution.Type == Distribution.DiracDelta ||
withProfile.StepDistribution.Type == Distribution.Testable)
{
RunTrial(0, withProfile);
var trial = trials[0];
for (var i = 1; i < NUM_TRIALS; i++)
{
trials[i] = trial;
}
}
else
{
Parallel.For(0, NUM_TRIALS, i =>
{
RunTrial(i, withProfile);
});
}
return(trials);
}
protected override void RunTrial(int trialNumber, RunProfile profile)
{
Trial trial = new Trial();
var addAmt = profile.ContributionAmount;
var withdrawAmt = profile.WithdrawalAmount;
var trialLength = profile.TrialLength;
var contribLength = profile.ContributionLength;
var stepSampler = profile.StepDistribution;
var sigma = profile.Volatility;
var mu = profile.Drift;
double[] balances = new double[trialLength];
double[] unitValues = new double[trialLength];
double[] returnRates = new double[trialLength - 1];
unitValues[0] = 1.0;
// Populate trial data
double step = 0;
balances[0] = profile.InitialAmount + addAmt;
for (var i = 1; i < trialLength; i++)
{
// Discretized Levy process (random walk with steps distributed as stepDistribution)
step += stepSampler.NextDouble();
// Discretized geometric Brownian motion using the Milstein method
var unitValue = (1 + mu + sigma * step + 0.5 * sigma * sigma * (step * step - 1)) * unitValues[i - 1];
// Get return rate
var returnRate = unitValue / unitValues[i - 1];
var prevBalance = balances[i - 1];
var balance = prevBalance > 0 ?
(i < contribLength ?
// Contribution period
(prevBalance + addAmt) * returnRate :
// Withdrawal period
(prevBalance - withdrawAmt) * returnRate)
: 0;
if (i == contribLength - 1)
{
trial.Peak = balance;
}
balances[i] = balance > 0 ? balance : 0;
unitValues[i] = unitValue;
returnRates[i - 1] = Math.Log(returnRate);
if (balance == 0)
{
break;
}
}
trial.Final = balances[trialLength - 1];
trial.Balances = balances;
trial.ReturnRates = returnRates;
// Add data to return value
mutex.WaitOne();
trials[trialNumber] = trial;
mutex.ReleaseMutex();
}
protected abstract void RunTrial(int trialNumber, RunProfile profile);