Determing Risky expectations grid for Spark

[sbenthall]

See #162 ...

For Spark SHARK, we will need a good grid over risky expectations. These will only come into play when the expectations regime is STRANGE.

One way we could do this is set the STRANGE threshold low, and cover the full range -- this should be much like the WHITE SHARK case.

[alanlujan91]

Is it possible to get a hark-less run and cover that range? then see if hark presence changes the range much.

[sbenthall]

RiskyAvg range: from 1 to .. (what's implied by the usual dividend) ... maxRiskyAvg.

Get maxRiskyAvg by:

• Generate returns based on the USUAL process. (How much?)
• Get small chunks of that data. (How small?)
• Compute the mean rate of return from those chunks
• Use the max as the maxRiskyAvg.

Or:

• Look at the usual distribution
• Find the value of the tail at some high confidence level.

RiskyStd range: from 0 to ... (the usual dividend $\sigma$) ...

(something similar to set the max).

There are many more clever ways to do the calibration but we probably don't have time to implement them in SPARK.