Statistical Targeting
Jump to navigation
Jump to search
A family of Targeting strategies that learn about the enemy's movement through the accumulation of relevant statistics. In particular, these methods tend to process input on the fly, updating an internal data structure from which they can quickly generate a firing angle in future situations; this is in contrast to log-based targeting methods that store raw data and analyze it at fire time.
The statistics gathered could be anything that can be used to reconstruct a firing angle, such as the success rates of different simple targeters in a Virtual Guns array, the last bearing offset that would have hit the enemy, or a segmented array of Visit Count Stats.
Forms of statistical targeting
- GuessFactor Targeting (traditional) - The most common and successful form of statistical targeting involves the combination of GuessFactors, segmentation, and Visit Count Stats. Few modern bots use any other form of statistical targeting.
- Neural Targeting - This method of targeting trains a neural network to output correct firing angles based on input given at fire time. While not a commonly used method, it has been proven to have strong potential by Engineer, by Wcsv.
- Averaged Bearing Offset Targeting - A simple approach to statistical targeting. This method works by tracking the bearing offset that would have hit the enemy each time a bullet is fired (using Waves) and then firing at that average of those offsets.
- Angular Targeting - The original form of Angular Targeting was the same as Averaged Bearing Offset Targeting. A more effective version was used in Gouldingi, called Angular Targeting/Factored, which multiplied the bearing change by a factor that was continually averaged and dependent upon the enemy's current direction and velocity.
- Laser Targeting - The next step up in complexity from Averaged Bearing Offset Targeting, this method kept a log of correct firing angles, from which it would randomly select one at fire time.
- BestPSpace - While not really a firing technique in itself, this method describes segmenting the probability graphs based on different variables to select the best probability.
