MinTrackRecord: Minimum Track Record Length

Description

The Minimum Track Record Length responds to the following question: "How long should a track record be in order to have a p-level statistical confidence that its Sharpe ratio is above a given threshold?". Obviously, the main assumption is the returns will continue displaying the same statistical properties out-of-sample. For example, if the input contains fifty observations and the Minimum Track Record is forty, then for the next ten observations the relevant measures (sharpe ratio, skewness and kyrtosis) need to remain the same as the input so to achieve statistical significance after exactly ten time points.

Usage

MinTrackRecord(
  R = NULL,
  Rf = 0,
  refSR,
  p = 0.95,
  weights = NULL,
  n = NULL,
  sr = NULL,
  sk = NULL,
  kr = NULL,
  ignore_skewness = FALSE,
  ignore_kurtosis = TRUE
)

Arguments

an xts, vector, matrix, data frame, timeSeries or zoo object of the returns input

the risk free rate

refSR

a single value or a vector when R is multicolumn. It defines the reference Sharpe Ratio and should be in the same periodicity as the returns (non-annualized).

the confidence level

weights

(if R is multicolumn and the underlying assets form a portfolio) the portfolio weights

(if R is NULL) the track record length of the returns

(if R is NULL) the sharpe ratio of the returns

(if R is NULL) the skewness of the returns

(if R is NULL) the kurtosis of the returns

ignore_skewness

If TRUE, it ignores the effects of skewness in the calculations

ignore_kurtosis

If TRUE, it ignores the effects of kurtosis in the calculations

Value

A list containing the below

min_TRL: The minimum track record length value (periodicity follows R)
IS_SR_SIGNIFICANT: TRUE if the sharpe ratio is statistically significant, FALSE otherwise
num_of_extra_obs_needed: If the sharpe ratio is not statistically significant, how many more observations are needed so as to achieve this

References

Bailey, David H. and Lopez de Prado, Marcos, The Sharpe Ratio Efficient Frontier (July 1, 2012). Journal of Risk, Vol. 15, No. 2, Winter 2012/13

Examples

Run this code

# NOT RUN {
data(edhec)
MinTrackRecord(edhec[,1],refSR = 0.23) 
MinTrackRecord(refSR = 1/12^0.5,Rf = 0,p=0.95,sr = 2/12^0.5,sk=-0.72,kr=5.78,n=59)

### Higher moments are data intensive, kurtosis shouldn't be used for short timeseries
MinTrackRecord(edhec[,1:2],refSR = c(0.28,0.24), ignore_skewness = FALSE, ignore_kurtosis = FALSE)
MinTrackRecord(edhec[,1:2],refSR = c(0.28,0.24), ignore_skewness = FALSE, ignore_kurtosis = TRUE)
MinTrackRecord(edhec[,1:2],refSR = c(0.28,0.24), ignore_skewness = TRUE, ignore_kurtosis = TRUE)

MinTrackRecord(edhec[,1:2],refSR = 0.26,weights = c(0.5,0.5), 
               ignore_skewness = FALSE, ignore_kurtosis = FALSE)

# }