Dispositions — Tier-2

Tier 1 is the personality reading of a stock — four panels, each describing a distinct character dimension. Tier 2 is what happens when you take two of those dimensions and ask the dashboard to show you only the stocks that score well on both. That's what a disposition is in the LENS view: a deliberate 2D sub-space of the character grid, with a target corner that defines what "well" means.

This page does two things. It walks through what a disposition is — the X axis, the Y axis, the geometric-mean combination, the target-quadrant filter, the directional-confirmation rule. And it answers an honest question with a real 19-year backtest: if you held the top 20 picks of each disposition for a month and rebalanced, would you beat SPY?

What a disposition does

Each disposition combines two percentile inputs into an X axis (sometimes one input, sometimes a geometric mean of two or three), and another two into a Y axis. Every stock's position in the X/Y plane is the (X-percentile, Y-percentile) of its composite scores. A target corner — top-right for most — defines the sub-region of the plane the disposition cares about.

A few rules sharpen the framework:

• Geometric mean, not arithmetic. A 50-and-50 stock scores 50 on the X composite. A 99-and-1 stock scores about 10. The geometric mean penalizes single-leg weakness, so a name has to score well on both inputs to land in the target zone — not just on one.
• Inputs that are null mean the composite is null. A name missing one of its underlying percentile inputs gets no disposition score. The framework refuses to fire when any leg is missing, which is the dashboard's directional-confirmation principle made explicit.
• Target-quadrant restriction. Each axis is shown as a 50-wide half — only the target quadrant of the percentile plane is plotted, and only target-quadrant rows enter the Top Picks panel. This is a deliberate narrowing, not a sort.
• Geometric-mean fit ranks within the target. A stock at (95, 95) sits closer to the target corner than one at (60, 95). The fit score is the geometric mean of the two coordinates (after flipping for non-TR targets), and the top 20 by fit are the disposition's Top Picks.

The framework gives the dashboard a way to ask compound questions of the universe: "Which stocks are growing fast AND outperforming?" (Beats) "Which are persistent leaders AND quietly volatile?" (Surge) "Which are cheap AND generating cash?" (Bargains). The 2D structure forces simultaneity. The percentile floor forces magnitude. The directional-confirmation rule keeps null-prone names from sneaking in.

The eight dispositions in the dashboard today:

• Surge — Momentum × Amplitude (high-momentum, high-volatility breakouts)
• Alpha — Performance × Persistence (durable outperformance)
• Beats — Market Beat × Business Beat (price meets fundamentals)
• ETFs — Beats lens restricted to the ETF watchlist
• Compound — Operational Quality × Compounding Growth (own-it-for-a-year names)
• Stable — Operational Quality × Storm Stability (drawdown discipline)
• Bargains — Cheapness × Cash Generation (mean-reversion candidates)
• Bear — Price Weakness × Business Weakness, both axes flipped (short candidates)

The Tier-2 test

The natural test at this tier — and the one we ran — is the mechanical buy-signal version: at each month-end, pick the top 20 by fit, hold them equal-weighted for 4 weeks, rebalance, repeat for 19 years on a point-in-time S&P 500 universe. Did the basket beat SPY?

What the evidence is telling you

Each testable disposition produces modest positive alpha — and that's exactly what the framework should produce. Over 19 years of PIT-correct backtests:

• Performance × Persistence (Alpha) is the strongest: +6.9% CAGR excess, Sharpe 0.60, MaxDD −22%. Persistence is the most orthogonal Tier-1 dimension, and pairing it with Performance produces the cleanest 2D filter.
• Market Beat × Business Beat (Beats) comes next: +4.4% CAGR, Sharpe 0.45, MaxDD −27%. Pairing price strength with fundamentals strength filters out the empty-momentum names.
• Surge (Momentum × Amplitude) returns +3.8% but with the highest disposition-level volatility — Sharpe 0.36, MaxDD −30%.
• Bear is the weakest: +2.7% CAGR, Sharpe 0.22, MaxDD −44%. Inverting both axes is a real signal, but it's a long-only short-candidate proxy that catches false positives in bull regimes.

These are real but modest numbers. Sharpe ratios are all sub-1, meaningful drawdowns are real (Surge gets to −30%, Bear to −44%), and none of the baskets is a turnkey strategy. The dispositions are doing what their construction implies: deliberately narrowing the universe to a sub-region of character where a sub-Sharpe-1 signal exists, then leaving the rest of the work to the investor.

That's why Tier 2 is still a watchlist tool, not a strategy. The alpha that this architecture can produce — the kind that turns into a deployable portfolio — compounds when you stack additional disciplines on top: multiplicative composites across three to four orthogonal dimensions (not just two), sector caps, per-position trend exits, regime overlays. That layer is still in progress and not yet ready to publish; what ships here is the descriptive surface it builds on.

The four deferred dispositions need fmp-derived ratios. Compound, Stable, and Bargains depend on point-in-time valuation and quality percentiles (P/E, EV/FCF, ROE, FCF yield, gross margin) ranked against sector peers. The fmp pipeline ingests the underlying fundamentals but the historical sector-relative ranking layer isn't wired yet. ETFs needs a survivorship-aware historical ETF panel that doesn't exist in the current stack. These will land once the fmp/ pipeline gets the next pass; the rows will fill in.

What dispositions are actually for

Tier 2 is a focused watchlist surface, not a turnkey strategy. The point of a disposition isn't to hand you a deployable portfolio — it's to narrow the 500-stock universe to a sub-region of character that matches the question you're asking right now. The modest alpha visible in the backtests above is real, but it's modest enough that the right primary use is editorial: a starting set for investor judgment, not a basket to hold mechanically.

When you open LENS and pick a disposition, what you get is:

1. A scatter plot of stocks in a 2D character region — visually showing which names cluster near the target corner and which are barely in the quadrant.
2. A Top Picks list of the 20 names with the highest target-fit — your starting set for judgment.
3. The two-axis structure itself — names being plotted on (Momentum, Amplitude) tells you something about what you're optimizing for and what you're trading off. Looking at Compound vs Bargains is looking at the same universe through two different questions.

The right use is editorial: "Show me the top 20 stocks where Persistence and Performance both score high. Now I'll go look at each one, read their full SYMBOL card, apply my own judgment, and decide what (if anything) to do." That's a fundamentally different activity from "buy the top 20 mechanically every month," and the activity that is the dashboard's intended use.

The Top Picks panel on the LENS view also runs a 1-year equal-weighted backtest of the displayed basket vs SPY — but as forward-looking decoration showing what the current set of names has done recently, not as a backtest claim for the disposition's predictive power. The 19-year results above are the honest baseline.

Going further

• The four panel deep-dives (Perform, Persist, Profit, Peril) — explain what each character dimension is and why, with Tier-1 face-validity / orthogonality / marginal-IC evidence.