TL;DR
Calculate floors, store picks, generate graphics
# player_stats_v2.py
def calculate_floor(games: list) -> int:
"""Floor = minimum value"""
return min(games)
# Anthony Edwards - Last 10 games
pts = [32, 28, 41, 24, 35, 29, 27, 31, 26, 33]
floor = calculate_floor(pts) # 24
line = 22.5
if floor > line:
print("✓ FLOOORGANG") # 100% hit rateGame history visualization:
Floor = min(recent games)
INSERT INTO picks
(player, stat, line, floor, odds) VALUES (...) -- 224 rows inserted
Picks stored in Supabase
NBA @FlooorGang - Nov 26
Odds from DraftKings • Up to 10 recent games shown
Minnesota Timberwolves
PTS 109.5 OVER
Houston Rockets
PTS 108.5 OVER
LaMelo Ball
AST 6.5 OVER
Dylan Harper
PTS 9.5 OVER
Floor > Line = 100% hit rate
Auto-generated Twitter pick card
Project Killed
67.7% win rate sounds good until you factor in the juice. The books always win.
What We Built — a lesson in why the house always wins
The Floor Concept
A player's "floor" is their minimum performance over the last 10-20 games. If Anthony Edwards has scored at least 24 points in every single game, his floor is 24. Simple.
The theory: sportsbooks set lines based on averages. But if a player's floor exceeds the line, they've hit it 100% of the time recently. That's not a prediction — that's a pattern.
The catch: The books know this too. Lines below floors come with heavy juice (-200 to -400), eating your edge.
The Pipeline
A Raspberry Pi woke up at 8 AM, checked the NBA schedule, and calculated when to run — 3 hours before the first game. It fetched alternate lines from The Odds API (DraftKings), pulled game logs from the NBA API, and calculated floors for every player with lines.
When a floor exceeded an available line, it was a pick. The system generated a mobile-optimized graphic and posted it to Twitter automatically. A second cron job scored yesterday's picks at 9 AM.
The verdict: Cool automation, bad business. Shut down after 23 days.
Final Results
224
total picks generated
67.7%
win rate (126-60)
23
days running
$0
profit (juice ate it all)
Tech Stack
The pipeline worked flawlessly. Free NBA stats, $30/month for odds data, automated on a Raspberry Pi. The engineering was solid — the edge wasn't. Check out the picks archive at @FlooorGang.