Automated cross-session video game testing

A system comprising:

• an electronic data store configured to store data objects corresponding to key performance indicator data;
• and
• an interactive computing system comprising one or more hardware processors, the interactive computing system configured to execute specific computer-executable instructions to at least:
• receive a byte stream for a test session of a video game, the test session of the video game associated with an execution of an instance of the video game;
• decode a portion of the byte stream into a data object, the data object comprising data encapsulating an event associated with the execution of the instance of the video game;
• determine that the data object corresponds to a key performance indicator;
• generate a plurality of prediction models by using as training data historical data associated with the key performance indicator, the historical data including a data portion associated with execution of one or more of a plurality of additional test sessions of the video game on an older version of a video game engine than a version of the video game engine on which the test session was executed;
• refine generation of the plurality of prediction models by reducing a weight of at least one weighted parameter associated with the data portion of the historical data corresponding to the older version of the video game engine in response to receiving user feedback to reduce the weight, thereby decreasing impact of the data portion of the historical data on the generation of the plurality of prediction models;
• store the data object at the electronic data store;
• aggregate the data object with additional data objects stored at the electronic data store that correspond to the key performance indicator to obtain a first set of aggregated data objects;
• generate a user interface element based at least in part on the set of aggregated data objects;
• output the user interface element for display to a user;
• evaluate the first set of aggregated data objects against a trend generated from an occurrence of the plurality of additional test sessions of the video game;
• in response to a determination that the first set of aggregated data objects does not satisfy the trend, select a first test procedure based at least in part on a result of said evaluating by at least:
• applying the plurality of prediction models to the first set of aggregated data objects, each prediction model having weighted parameters and each prediction model configured to output a particular test procedure;
• selecting, using an automated test selector and without user input, a first of the prediction models from the plurality of prediction models, wherein the automated test selector is configured to select, without user input, different prediction models from the plurality of prediction models, and wherein the automated test selector is generated based at least in part on a machine learning algorithm configured to select the first prediction model as a prediction model associated with a highest accuracy of generated output of the plurality of prediction models;
• and
• using the particular test procedure output by the first prediction model as the test procedure;
• perform the test procedure on the video game by:
• modifying code corresponding to the video game;
• and
• initiating another test session of the video game using the modified code of the video game;
• and
• in response to a determination that a second set of aggregated data objects determined for the first test procedure does not satisfy the trend:
• repeat the selecting to select a second test procedure different from the first test procedure, wherein selecting the second test procedure comprises selecting from the plurality of prediction models a second prediction model associated with a next highest accuracy of generated output;
• and
• perform the second test procedure on the video game.

A non-transitory computer-readable storage medium storing computer executable instructions that, when executed by one or more computing devices, configure the one or more computing devices to perform operations comprising:

• receiving a byte stream for a test session of a video game, the test session of the video game associated with an execution of an instance of the video game;
• decoding a portion of the byte stream into a data object, the data object comprising data encapsulating an event associated with the execution of the instance of the video game, wherein decoding the portion of the byte stream into the data object transforms data from a machine-readable format to a human-readable format;
• determining that the data object corresponds to a key performance indicator;
• generating a plurality of prediction models by using as training data historical data associated with the key performance indicator, the historical data including a data portion associated with execution of one or more of a plurality of additional test sessions of the video game on an older version of a video game engine than a version of the video game engine on which the test session was executed;
• refining generation of the plurality of prediction models by reducing a weight of at least one weighted parameter associated with the data portion of the historical data corresponding to the older version of the video game engine in response to receiving user feedback to reduce the weight, thereby decreasing impact of the data portion of the historical data on the generation of the plurality of prediction models;
• aggregating the data object with additional data objects that correspond to the key performance indicator to obtain a first set of aggregated data objects;
• outputting the first set of aggregated data objects as a performance metric corresponding to performance of the test session of the video game;
• evaluating the first set of aggregated data objects against a trend generated from an occurrence of the plurality of additional test sessions of the video game;
• in response to determining that the first set of aggregated data objects does not satisfy the trend, selecting a first modified test procedure based at least in part on a result of said evaluating, wherein selecting the first modified test procedure comprises:
• applying the plurality of prediction models to the set of aggregated data objects, each prediction model having weighted parameters and each prediction model configured to output a particular test procedure;
• selecting, using an automated test selector and without user input, a particular prediction model from the plurality of prediction models, wherein the automated test selector is configured to select, without user input, different prediction models from the plurality of prediction models, and wherein the automated test selector is generated based at least in part on a machine learning algorithm configured to select as the particular prediction model a prediction model associated with a highest accuracy of generated output of the plurality of prediction models;
• using the particular test procedure output by the particular prediction model as the first modified test procedure;
• and
• performing the first modified test procedure on the video game by:
• modifying code corresponding to the video game;
• and
• initiating another test session of the video game using the modified code of the video game;
• and
• in response to determining that a second set of aggregated data objects determined for the first modified test procedure does not satisfy the trend:
• repeating the selecting to select a second modified test procedure different from the first modified test procedure, wherein selecting the second modified test procedure comprises selecting from the plurality of prediction models a second prediction model associated with a next highest accuracy of generated output;
• and
• performing the second modified test procedure on the video game.