We use DLPAL LS with Norgate daily and weekly data for S&P 500 stocks in quantitative scans to determine directional probabilities.
DLPAL LS is used by professional traders and hedge funds to develop algorithmic and machine learning strategies for directional and long/short trading. Users of the program can sort results based on different criteria and generate output for trading APIs. Historical files of features can be generated for backtesting the strategies and for machine learning.
Below is an example of quantitative scan of S&P 500 stocks as of the close of 11/08/2019 for daily and weekly data.
Below is the DLPAL LS workspace used for the daily scan. We use a small profit-target and stop-loss of 2% to evaluate the feature performance before an ensemble is created that corresponds to the directional probability. We use the normal cluster and the open as the traded price. In the output criteria we specify that in the results we want only those stocks with long probability P-Long > 55 and short probability P-Short > 55. We also want significance S of the results S > 1. Finally, since this is short-term analysis, we elect to remove the longer-term trend by marking the “Detrend All results” option.
DLPAL LS identified 89 stocks that satisfied the output filter criteria, 55 long and 34 short. Long and short are determined based on the sign of P-Long minus P-Short, which is called P-delta.
The user has several options is working with the results. For example after sorting for highest P-delta, the top and bottom 10% of stocks can be selected.
Or, after sorting for P-delta times Significance S, the top and bottom 7 stocks can be selected.
The stock selection depends on the strategy that is usually determined in advance by generating historical values for the features and backtesting it. In addition, train/score files can be generated for more advanced algos based on machine learning. Historical feature values and train/score files can be updated by the program when new data becomes available.
Below is the DLPAL LS workspace used for the weekly scan. We use 5% profit-target and stop-loss with weekly data to evaluate the feature performance before an ensemble is created that corresponds to the directional probability. All the other parameters are set as in the previous example.
DLPAL LS identified 124 stocks that satisfied the output criteria, 91 long and 33 short. Long and short are determined based on the sign of P-Long minus P-Short, which is called P-delta.
Below is one of several choices for working with the results. The top 5 long and short stocks are selected according to P-delta.
DLPAL LS offers several other capabilities including saving results in excel readable format for further manipulation and calculation of next bar returns and other useful statistics. It also calculated another set of features for groups of securities.
Data provider: Norgate Data