LiveFelt — odds observatory
Realized casino odds each mode produces on each sport's REAL plays, vs the fair target. Goal: equal, fair odds across every sport = a real casino game. 2026-08-08 · accumulates daily.
Maps are unbalanced. Crash house odds swing 0 points of bust-rate between sports (should be identical). Worst craps dice distance from true craps: 0.189 (fair is under 0.04). Both modes are off-target on every sport measured.
Crash — target: same ~2% edge every sport (P≥2× 0.49 · P≥5× 0.196 · P≥10× 0.098)
| sport | events | rounds | bust rate | avg peak | P(≥2×) | P(≥5×) | P(≥10×) | verdict |
|---|
| cryptoBTC / USD | 1175 | 1 |
0% |
1.9439357979615941e+31× | 1 | 1 | 1 |
NEVER CRASHES |
|---|
| FAIR TARGET | — | — | ~fair | — | 0.49 | 0.196 | 0.098 | — |
|---|
bust rate should be the SAME across sports. MLB dies too fast (an out crashes it); crypto never crashes (no bust event). Both need re-mapped so P(reach X) matches the fair curve everywhere.
Craps — target: true dice distribution (P7 = 16.7%), same every sport
| sport | dice distribution (cyan = realized · gray = true craps) | seven | TVD | verdict |
|---|
| crypto1175 rolls | 23456789101112 |
P7 16.0%
true 16.7% |
0.1891 | BROKEN |
|---|
cyan bar = how often this sport's map rolled each total · gray bar = true-craps target. TVD = total distance from true craps (0 = perfect, <0.04 = fair).
Slots — target: 95% RTP, sensible symbol spread, same every sport
| sport | plays mapped | distinct symbols | top symbol | top share | verdict |
|---|
| crypto | 47 | 1 | PUMP | 100% | FREQ SKEWED |
|---|
RTP auto-calibrates to 95% under ASSUMED symbol frequencies; when the real frequency is lopsided the realized RTP drifts off target. Note: slots consumes the pitch-level feed, so per-sport slots must be measured against each sport's pitch feed (refinement in progress).
Blackjack & Poker — cards drawn by uniform hash = fair 52-deck (edge from rules only)
| sport | card draws | uniformity χ² (fair < 16.9) | verdict | house edge |
|---|
| crypto | 1175 | 6 | FAIR DECK | rules only (BJ ~0.5%) |
|---|
The deck draws each card by a uniform hash of the real play, without replacement. These two modes are fair BY CONSTRUCTION and identical across every sport — the observatory confirms it (χ² low = uniform). The map-distortion problem is crash, craps, and slots only.
All 5 modes now measured. Next: fix slots to read each sport's pitch feed, then re-calibrate the crash / craps / slots maps so every column reads FAIR (equal odds every sport = the moat). Refresh: cd /root/lf-loop && node lf-odds.js && node lf-odds-render.js (on-demand; accumulates into the day's file).