Testing Probabilistic Business Interfaces with Variable Winning Probability

Following the previous discussion on testing probabilistic business interfaces, a new requirement introduces a non‑fixed probability algorithm where the user’s winning probability is calculated as P = p(1+0.1*N). Here, p is the original winning probability, N is the number of consecutive losses (capped at 5), users cannot win consecutively, and each gift’s probability is rounded to the nearest 1%.

The system provides three APIs: a lottery draw interface, an activity configuration interface (including gift settings), and a personal activity detail interface (user info, draw count, win history).

The testing tool is implemented in Java, wrapping the three APIs into methods that are repeatedly invoked to gather data and compute statistics; the test runs for one day.

The core testing difficulty lies in verifying the correct implementation of the formula P = p(1+0.1*N) under the constraints of no consecutive wins and variable probability.

Solution 1: Compute the theoretical overall winning probability for various p values through mathematical analysis, then perform massive lottery simulations and compare the observed results with the theoretical expectations, using the same standards as the previous probabilistic interface test.

Solution 2: Conduct a large‑scale simulation (e.g., 10,000 draws), record each draw as 1 (win) or 0 (loss), build a table of P versus N for a specific p, and count occurrences of patterns such as "1000" and "1001" to derive the empirical probability after two consecutive losses; compare this empirical probability with the theoretical Pn, again using the prior test standards.

Both approaches involve substantial test volume and time consumption; because the probability is evaluated per user, multithreaded testing must bind each thread to a distinct virtual user.

The article concludes that such requirements are best addressed with white‑box testing techniques.