How to deal with asset data having zeros in part when backtesting with 'runBacktest'?

조회 수: 1 (최근 30일)
When backtesting with the functions 'backtestStrategy', 'backtestEngine', and 'runBacktest', I have an issue with an asset data including zeros.
The following code is an example, which equally weights 4 stocks with rebalancing the portfolio every 20 days. The following code works as expeted.
%% create a pseudo stock price table.
d = datetime('2020-01-01'):caldays(1):datetime('2020-12-31'); % create a date array
price = rand(numel(d), 4); % create a pseudo stock prices
% price(:,4) = 0;
priceTT = array2timetable(price, 'RowTimes', d', ...
'VariableNames', {'Stock_A', 'Stock_B', 'Stock_C', 'Stock_d'});
%% create a backtesting strategy which equally weights all stocks and rebalances every 20 days
nAsset = size(priceTT, 2);
initial_weights = ones(1,nAsset) / nAsset;
equalWeightFn = @(~, ~) initial_weights;
strategy = backtestStrategy('EqualWeighting', equalWeightFn, ...
'InitialWeights', initial_weights, ...
'TransactionCosts', 0 , ...
'RebalanceFrequency', 20);
%% create a backtest engine and run backtest.
backtester = backtestEngine(strategy);
bt = runBacktest(backtester, priceTT);
%% summry the backtest result.
summary(bt)
However, when the prices of 'Stock_d' is replaced with zeros (please uncomment the third line in the example, "prices(:,4) = 0"), all the backtesting results becomes as below.
EqualWeighting
______________
TotalReturn NaN
SharpeRatio NaN
Volatility NaN
AverageTurnover NaN
MaxTurnover 0
AverageReturn NaN
MaxDrawdown NaN
AverageBuyCost 0
AverageSellCost 0
Please advise me whey an asset data including a stocks with zeros does not work.
Thank you for any comment in advance.
  댓글 수: 3
Brendan
Brendan 2021년 8월 16일
Peter is correct. In the latest version of the financial toolbox you will get a more helpful error message if you attempt to invest in an asset that has an invalid return (for example and Inf for NaN return).
backtest strategy rebalance functions should take care to avoid investing in assets that encounter invalid return data.
Doheon Lee
Doheon Lee 2021년 8월 27일
Thank you, Peter and Brenda,
My MATLAB is 2021a, and I still have thie problem. Instead of zeoring for NaN values in a table containing historica stock prices, I replaced NaN with 1e-10, which is close to zero and it works fine. That is, replacing the third line with the following,
price(:,4) = ie-10;
Thank you for the comments. :)

댓글을 달려면 로그인하십시오.

답변 (0개)

카테고리

Help CenterFile Exchange에서 Portfolio Optimization and Asset Allocation에 대해 자세히 알아보기

제품


릴리스

R2021a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by