These are my efforts to clarify the rules. They are not the official rules or official responses.
Is the investment portfolio by which intended to be:
- One large portfolio comprising stocks and ETFS
- Five portfolios of at most 5 stocks, and another five portfolios of at most 5 ETFs each?
It's just one big portfolio with both stocks and ETFs.
It's just one big portfolio with both stocks and ETFs.
How are information ratios combined across the overlapping time-periods, and possible portfolios?
Denominator and numerator are totalled first, leading to a single information ratio.
Should the return formula be modified to account for shorts?
1 + \sum. w_i \left( \frac{S'}{S} - 1 \right)
We can't collect the -1 unless sum w_i=1 which is not true if there are shorts.
Formula is corrected and no longer assumes sum of weights is unity.
Why is the benchmark chosen to be the zero portfolio (rather than say, equal positive weights)? The information ratio is undefined at this point.
Portfolios must be defined away from the origin.
3a. If a company defaults, will the RET for the day be -1 and 0 thereafter?
3b. If a company defaults, will the rank be 1 (or 1.5 if two companies default)?
3c. If a company de-lists, will the rank be 1 (or 1.5 if two de-list, etc)?
3d. If a company is taken private, will the last recorded stock price persist into the future?
3e. If a company is aquired, will the last recorded stock price persist into the future?
As inferred from the questions.
Do the prices used include dividends?
Yes, and pretty clearly indicated by the wording "total return". Same for stock splits.
At present the rules state that submissions not summing to unity will be disqualified. However this is a set of measure zero. In theory it is still a proper scoring rule without the constraint so one might wonder if it is necessary at all.
What is the tolerance for sum of probabilities?
This is updated. Five decimal places.