Ports the fix already applied to the sibling Lotto 6aus49 project.
Since draws are i.i.d., only expected value (jackpot-sharing
avoidance) is improvable, not hit probability. Replaces the ad-hoc
"lucky numbers" list with empirically-documented picks, adds general
arithmetic-progression detection (not just +1 consecutive pairs), and
scores extreme odd/even splits higher (adjusted for 5 main numbers:
distance from the 2.5 midpoint, not 3 as in the 6-number Lotto case).
Deliberately omits a sum-near-mean component - testing against known
overplayed combinations (1-2-3-4-5, 5-10-15-20-25) during the Lotto
port showed it inverted the ranking for tightly clustered sequences.
Raises popularity's weight in the quality score from 20% to 45%
(main/euro/pattern/recency reduced accordingly), since those signals
carry no real predictive value for hit probability.
Also fixes _get_smart_euro_numbers(), which previously returned the
literal same Euro-number pair for tips #1-3 every single run (top-2
AI predictions, no randomization) and had zero EV consideration. Now
weighted-random across all 12 candidates with lucky numbers (3, 7)
downweighted - deliberately more conservative than the main-number
heuristic since 1-12 overlaps with calendar months and has no clear
birthday-range split to exploit.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
cbazzaandClaude Sonnet 5
0285a1c5fa