Version: 3.0 | Status: Live validation — Phase 4 Author: Rahul Sai | Started: May 2026
TPAAM is a regime-aware probabilistic capital allocation framework. Central hypothesis: which theory of markets is valid right now is itself regime-dependent. Six regimes (Expansion / Contraction / Mania / Panic / Transition / Innovation) each assign different epistemic authority to six scoring factors (Q/V/N/A/P/T).
Current status: Behavioral discipline layer empirically validated in live trading. Mathematical parameter layer in active validation. Whether specific weights add edge beyond the behavioral layer alone is an open and formally unresolved question.
Layer 1 — Market Regime Detection (probabilistic, non-exclusive)
↓ MMAC: posterior blending, not hard-switch
Layer 2 — Theme Discovery
Layer 3 — Probabilistic Asymmetry Engine
Layer 4 — Multi-Factor Scoring (Q / V / N / A / P / T)
Layer 5 — Dynamic Weighting (regime-conditional)
Layer 6 — Position Sizing
Layer 7 — Portfolio Construction
Layer 8 — Feedback Loop
Normalises regime scores into a posterior and blends all six weight vectors proportionally. A barely-dominant regime produces cautious blended weighting. Missing input data pulls toward uniform (maximum uncertainty) and raises a degraded-confidence flag — never silently substituted.
Formal analogue: Bayesian Model Averaging over regime hypotheses.
AsymmetryRatio = (P_up × M_up) / (P_down × M_down)
2.0 investment minimum
3.0 tactical trade minimum
4.0 aggressive allocation eligible
| Regime | Q | V | N | A | P | T |
|---|---|---|---|---|---|---|
| EXPANSION | 0.35 | 0.25 | 0.15 | 0.10 | 0.10 | 0.05 |
| CONTRACTION | 0.35 | 0.20 | 0.05 | 0.10 | 0.15 | 0.15 |
| MANIA | 0.10 | 0.05 | 0.40 | 0.25 | 0.15 | 0.05 |
| PANIC | 0.40 | 0.10 | 0.05 | 0.20 | 0.20 | 0.05 |
| TRANSITION | 0.20 | 0.15 | 0.15 | 0.20 | 0.15 | 0.15 |
| INNOVATION | 0.10 | 0.05 | 0.35 | 0.30 | 0.10 | 0.10 |
Weights are hand-set priors, not estimated from data. Whether they outperform equal-weight is what the Part 6 three-arm backtest tests.
| Track | Count | Win Rate | Status |
|---|---|---|---|
| FTMO demo | 22/30 | ~64% | 8 trades to milestone |
| Crypto (clean Cat A) | 22 | 63.6% | Gate-conditional |
Validated findings:
- HTF alignment = Y: ~80-83% win rate
- HTF = N: 0% win rate across clean sample
- Override rate: 0% wins in pre-override clean sample
Open question: Whether the specific quantitative weights add edge beyond the behavioral discipline layer alone. Pre-registered three-arm test (TPAAM vs equal-weight vs MMAC blend) will answer this.
- PANIC over-detection bias: 38.3% on GBM null test
- INNOVATION detection gap: 0.0% on null test
- US-centricity in regime proxies
- Weight matrices are unvalidated priors
Entropy Haircut: Shannon entropy of MMAC posterior scales position size down as regime uncertainty rises.
IMM Transition Layer: Interacting Multiple Model filter adds regime transition prior. Stepping stone toward a learned HMM. Directly attacks the PANIC detection bias.
Python 3 | yfinance | Alpha Vantage | AngelOne SmartAPI
"The system should never be more certain than the evidence warrants."
The framework is built to be falsifiable. Parameters are logged. Open questions are documented. Things that don't work are recorded alongside things that do.
Nothing here is financial advice. This is a research and validation project.