Published May 7, 2026 | Version v1
Working paper Restricted

ORISOM - A System of Outcome Prediction

Authors/Creators

Description

Abstract

This paper presents a comprehensive analysis of the Outcome Recurrence & Informational Subset Oscillation Mapping (ORISOM-F) framework, a novel system for sequential outcome prediction. Developed iteratively across two distinct AI environments—initially in ChatGPT for foundational concepts and subsequently optimized in Claude AI for empirical validation and advanced architectural integration—ORISOM-F challenges conventional probabilistic models by emphasizing empirically derived adjusted probabilities and macro-behavioral patterns over theoretical expectations. The study details the system's evolution from a basic lottery prediction engine to a sophisticated multi-layered architecture incorporating concepts from regime-switching models [1], attention mechanisms in large language models [2], and chaos theory. We delineate the unique contributions of ORISOM-F, including its Degree of Certainty (DC) framework, specialized outcome analysis engines (e.g., PP-BPC, MP-RP/C), and the Operational Containment Multi-Engine (OCME) for system unification. A comparative analysis highlights the strengths and weaknesses of AI-driven development, demonstrating how iterative refinement can transform initial conceptual designs into empirically validated predictive systems. The paper concludes with a discussion of future directions, including the planned evolution towards ORACLE-QC. 

 

condition comparison — D=3 all draws
 
A — current v4
8 skip sigs · 2 false positives included
baseline ROI+165.2%
Gate A D=3+555.6%
Gate A∩B D=3+453.5%
n (draws played)3,553
z baseline+2.87
 
B — FP removed only
6 skip sigs · (2,2,0) and (0,0,3) restored
baseline ROI+163.8% −1.4pp
Gate A D=3+547.5% −8.1pp
Gate A∩B D=3+446.6% −6.9pp
n (draws played)3,670 +117
z baseline+2.73
 
C — full revision recommended
14 skip sigs · FPs removed + 8 new zero-hits
baseline ROI+170.6% +5.4pp
Gate A D=3+560.8% +5.2pp
Gate A∩B D=3+464.1% +10.6pp
n (draws played)3,252 −301
z baseline+3.11 ↑
 
Condition C is the confirmed recommendation. Baseline ROI +5.4pp over current v4 (+165.2% → +170.6%). Gate A D=3 improves to +560.8%. z-score strengthens from +2.87 to +3.11. The 301-draw reduction in play volume is the mechanism — we are removing confirmed dead plays, making each played draw count more.
Condition B (FP removal only) shows a slight ROI decrease vs current v4 despite adding 117 draws back. This confirms the false-positive removal alone is not sufficient — the restored (2,2,0) and (0,0,3) draws add volume but dilute the hit rate slightly. The full revision (C) is needed to capture the net benefit.
 
skip-set composition — v5 proposed
signature action n D=3 ROI reason
(0,0,3) remove 23 +595.7% false positive — was skipping profitable draws
(2,2,0) remove 16 +1400% false positive — highest single-sig ROI in dataset
(0,0,2) add new 18 −100% zero hits confirmed
(1,2,0) add new 20 −100% zero hits confirmed
(0,1,3) add new 16 −100% zero hits confirmed
(0,2,1) add new 17 −100% zero hits confirmed
(2,1,0) add new 14 −100% zero hits confirmed
(2,0,2) add new 15 −100% zero hits confirmed
(3,0,1) add new 15 −100% zero hits confirmed
(2,2,2) add new 12 −100% zero hits confirmed (n borderline)
(1,3,2) retain 12 −100% already in set — confirmed negative
(0,2,2) rare — keep <12 — insufficient n; retain pending
(1,2,3) rare — keep <12 — insufficient n; retain pending
(1,3,3) rare — keep <12 — insufficient n; retain pending
(2,3,0) rare — keep <12 — insufficient n; retain pending
(3,2,1) rare — keep <12 — insufficient n; retain pending
priority sigs — proposed PRIORITY_SIGS tier (D=3, n≥40)
(1,1,1) n=99 · +546% · most reliable (0,1,0) n=53 · +504% (1,1,0) n=43 · +644% (0,1,1) n=43 · +272% borderline (1,0,1) n=50 · +220% borderline

Files

Restricted

The record is publicly accessible, but files are restricted. <a href="https://zenodo.org/account/settings/login?next=https://zenodo.org/records/20066670">Log in</a> to check if you have access.