Double Color Ball ML Prediction

Problem context

Double Color Ball selection is often treated as a one-off random event without a structured, reviewable data-analysis workflow. This skill converts historical draw records into structured features, helping inspect frequency, misses, consecutive numbers, matching tails, zones, and parity/big-small patterns before producing red and blue ball recommendations.

How it works

• Data and training: the source notes 3,431 historical periods, using the most recent 2,000 periods as the training window to form 62,700 red-ball samples.
• Feature engineering: it builds 15 features, including recent_10, recent_30, recent_50, current_miss, avg_miss, max_miss, is_repeat, has_consecutive, same_tail_count, freq, in_zone1/2/3, is_odd, is_big, and is_prime.
• Model integration: it uses a random forest with 100 trees and maximum depth 12, allocating 50% of the ensemble weight to the ML branch while combining it with a rule engine.
• Usage entry points: it exposes basic prediction, backtesting, and model inspection, which makes it easier to compare the documented V3.6/V3.7 rule-based versions against the V3.8 ML-enhanced version.

Boundaries

• The skill explicitly states that it is for entertainment only and does not guarantee wins.
• Lottery draws are random events; historical patterns and model scores do not change the underlying probability.
• It is better treated as an entry point for feature engineering, model backtesting, and data analysis experiments, not as a betting decision tool.