From RevDog V2 to V3: Evolving Stock Analysis for Professional Capital
RevDog V3 evolves small-cap stock analysis from score-based conviction to institution-ready decision support. By validating factor persistence, stabilizing peer comparisons, and separating macro from stock signals, V3 prioritizes reliability, risk control, and capital protection.
RevDog V2 was built to impose structure and discipline on small-cap stock analysis—an area where outcomes are often driven by narratives, partial information, or hindsight.
Over extended live usage and repeated application across market conditions, one thing became clear:
the framework works, but certain design choices limit its use for professional capital.
RevDog V3 addresses those limits.
This is not about adding features or improving scores at the margin.
It is about making the analysis reliable, stable, and usable when real capital is involved.
What RevDog V2 Established
V2 introduced three foundations that remain central to the platform.
1. Eligibility Before Conviction
Stocks must pass hard filters on liquidity, accounting quality, governance, and balance-sheet integrity before any scoring is applied.
This avoids a common failure mode in small caps: strong conviction built on structurally weak businesses.
2. Sector-Relative Metrics
Growth, profitability, and efficiency are evaluated relative to sector peers rather than absolute thresholds.
This reflects how professional research is conducted and avoids penalizing industries with structurally different economics.
3. Deterministic, Reproducible Outputs
Given the same inputs, V2 produces the same result.
There is no adaptive tuning or narrative-driven override.
This consistency is essential when analysis is used repeatedly across portfolios and time.
Where V2 Shows Its Limits in Practice
These are not mistakes. They are natural constraints of a score-based framework.
Single Scores Compress Risk
A single conviction score hides uncertainty.
Two stocks can score similarly while having very different payoff profiles—one steady with limited downside, another highly asymmetric with deep drawdown risk. Treating them equivalently limits how analysis translates into risk control and sizing.
Macro Adjustments Blur Stock-Level Signal
In V2, macro conditions adjust stock scores.
In practice, this creates situations where business fundamentals are unchanged, but conviction rises or falls due to market regime. This blurs an important boundary between company quality and environmental risk.
Factor Persistence Is Untested
V2 reports strong historical performance, but factors are not explicitly tested for persistence across unseen periods.
Without this, it is difficult to separate signals that consistently add value from signals that benefited from a favorable regime.
Design Direction for RevDog V3
V3 focuses on making analysis more reliable for real capital deployment.
Four changes define the transition.
1. Walk-Forward Validation of Core Factors
What changes
Each major factor—growth, profitability, capital efficiency, balance sheet quality—is tested using walk-forward validation:
- trained on historical data
- evaluated on a subsequent unseen period
The output identifies:
- which factors remain predictive
- which weaken
- how much signal strength decays over time
Why
Backtests measure fit to history. Walk-forward validation measures generalization to the future.
Signals that do not persist tend to fail quietly in live portfolios. This step ensures factor weights reflect observed stability, not assumptions.
2. More Robust Peer Comparisons
What changes
V3 enforces:
- minimum peer group sizes of 12–20 stocks
- asymmetric market-cap expansion for thin sectors
- blended sub-industry comparisons where required
- winsorized medians to limit outlier influence
Why
In V2, comparing against a small peer set could create instability—one outlier quarter could swing relative metrics significantly, and peer groups themselves changed frequently.
Score stability is a prerequisite for actionable analysis.
If scores move materially due to peer noise rather than fundamental change, the signal is not usable.
This change reduces false signals, lowers turnover, and makes scores more comparable across time.
3. Macro Becomes a Portfolio Input, Not a Stock Adjustment
What changes
Stock analysis remains purely company-specific.
Macro indicators influence:
- exposure levels
- sector tilts
- position sizing
- cash allocation
Why
Market regime does not change business quality.
It changes how much risk should be taken.
Separating these layers prevents artificial inflation of conviction during market peaks and makes the analysis easier to integrate into portfolio construction.
4. Hard Governance Overrides
What changes
Certain high-information signals—particularly significant promoter stake reductions—function as hard constraints rather than scored factors.
If promoter holding decreases by more than 15 percentage points over a trailing 12-month period, the stock receives a governance warning that caps its effective score regardless of other metrics.
Why
In Indian markets, promoter behavior carries unique informational weight. When insiders reduce exposure at scale, this signal should not be diluted by averaging it into a composite score.
V3 treats thesis-killing signals as circuit breakers, not as inputs to a weighted sum.
Data Recency Awareness
V3 explicitly accounts for signal staleness.
Not all data is equally current:
- quarterly shareholding disclosures are 60–90 days old
- financial results are reported with delay
- price and volume signals are real-time
Signals are weighted with decay adjustments based on their freshness. This prevents stale information from exerting the same influence as current data.
Performance Expectations
V3 introduces structural changes that affect backtest characteristics.
Compared to V2:
- Raw CAGR may moderate as macro no longer inflates conviction during risk-on peaks
- Drawdowns are expected to improve due to regime-aware exposure control
- Risk-adjusted returns should improve even if absolute returns are marginally lower
These are intentional tradeoffs.
V3 prioritizes robustness and drawdown control over headline CAGR.
What V3 Is Not Trying to Do
V3 does not attempt to:
- predict short-term price movements
- replace portfolio construction
- auto-optimize position sizes
Those decisions belong at the portfolio layer.
V3’s role is deliberate and narrow:
provide stock-level analysis that can be trusted as an institutional input.
Looking Ahead
V3 establishes the foundation for institutional-grade stock analysis.
Future iterations will focus on:
- probabilistic return estimation
- factor interaction and covariance-aware weighting
- capacity-adjusted signals for larger deployments
These enhancements build on V3 once stability is demonstrated in live use.
Closing Perspective
RevDog V2 imposed structure on stock selection.
RevDog V3 imposes discipline on how that structure is used.
The evolution reflects a simple reality: in markets, avoiding avoidable losses matters as much as identifying opportunities.
That is the standard V3 is built to meet.