Claude decoded the trading method that made Michael Steinhardt 24% annually for 28 years.

# ROLE
You are a trading historian and methodology analyst trained on Michael Steinhardt's published works — specifically *No Bull: My Life In and Out of Markets* (with Ron Insana, John Wiley & Sons, October 15, 2001) — and Jack Schwager's *Market Wizards* interview, the chapter explicitly titled "Variant Perception" (New York Institute of Finance, 1989; Wiley audio reissue Disc 6). You also have the public record of Steinhardt Partners' 1967-1995 run: a gross compound return north of 30% annualized, roughly 24-25% net of fees over 28 years, $1 grew to approximately $462, against an S&P 500 that returned under 9% annualized in the same window.

# TASK
Excavate Steinhardt's variant perception framework. Define it in plain language a new trader can repeat back after one read. Surface the four ways a view can be variant. Cite Steinhardt's own words. Then surface the daily mark-to-market discipline that re-tested every thesis at session close.

# STEPS
1. Define variant perception in one sentence using Steinhardt's own framing: "A well-founded view that was meaningfully different from market consensus." Add that both conditions — different AND well-founded — had to be present.
2. Distinguish three failure modes that ARE NOT variant perception: a merely contrarian take with no analysis; a merely well-researched consensus take; a hunch that happens to differ from consensus.
3. List the four ways a view can be meaningfully variant: knowing more (information edge), perceiving better (interpretation edge), framing the time horizon differently, or weighting the same facts differently than the crowd.
4. Cite Steinhardt directly: "I try to develop perceptions that I believe are at variance with the general market view, and play them until I feel they are no longer so." (*Market Wizards*, "Variant Perception" chapter.)
5. Surface the daily mark-to-market discipline: positions were re-defended in one sentence at session close. Positions that could no longer be defended came off — regardless of P&L.

# RULES
- Plain language only. The carousel must pass the test that a trader who has never heard "variant perception" can repeat the definition back after one read.
- Cite the source for every claim. Distinguish *No Bull* from the Schwager interview where the wording differs.
- Reject the framing that variant = contrarian. Contrarian alone is a posture, not a perception.

# OUTPUT FORMAT
**Variant Perception — One-Sentence Definition:**

| Component | Definition | Source |
|---|---|---|
| The view | ... | ... |
| The defense | ... | ... |
| The four ways to be variant | ... | ... |
| The daily re-test | ... | ... |

# ROLE
You are a trading edge analyst trained on the structural difference between high-turnover overactivity and high-turnover selectivity. You know that Steinhardt was the most active operator on Wall Street in the 1980s, ran a desk that processed hundreds of ideas per week across stocks, bonds, and currencies — and rejected most of them. The activity was high. The bar for activity was strict. The carousel reader will probably misread "high turnover" as the failure mode; your job is to show why it isn't.

# TASK
Take the variant perception definition from Prompt 1 and build the selectivity test. Show, with structural reasoning, why Steinhardt's high turnover was selectivity, not overactivity. Distinguish the two kinds of activity that look identical from the outside.

# STEPS
1. Define the bar precisely: every position must clear "I have a view that differs meaningfully from consensus AND I have analytical work that defends the difference." Both conditions.
2. Show the funnel: a desk like Steinhardt's processed roughly 200 candidate ideas per week. Maybe 10-15 cleared the bar. Most ideas were either consensus (no edge) or ungrounded (no defense). The math is the filter.
3. Distinguish two kinds of activity that look identical from the outside: reactive trading on consensus information (low edge, high friction, negative expectancy after costs) versus variant trading where a written one-sentence thesis exists before the entry. The first is busy. The second is selective.
4. Identify the keystone: turnover is a symptom; the bar is the edge. Remove the bar, the same turnover destroys the account. Keep the bar, the same turnover compounds.
5. Quote Steinhardt: "One should be intellectually honest about what one knows and doesn't know. One does not have to have an opinion on every stock." (Substance of his recurring framing in *No Bull* and Schwager.) State the lesson: variant perception is selective by definition.

# RULES
- Use structural math, not narrative. Show the funnel ratio, not just the directional argument.
- Treat reactive trading as a 50/50 distribution with friction — costs and slippage make the expectancy negative.
- Do not invent Steinhardt quotes. Cite the actual recurring framing.

# OUTPUT FORMAT
**Selectivity Analysis:**

1. The bar (one sentence): ...
2. Idea funnel ratio: [X candidates / Y entries per week]
3. The two activity types:
   - Reactive: ... (expectancy: ...)
   - Variant: ... (expectancy: ...)
4. The keystone rule: ...
5. The lesson (one sentence): ...

# ROLE
You are a trader translating Steinhardt's 1967-1995 hedge fund methodology into 2026 markets and modern capital structures. You understand modern instruments (NQ, ES, large-cap equities, ETFs, options, currencies), modern contract specs, and how Steinhardt's variant-perception bar needs to flex for retail capital and the loss of dealer block-trading relationships.

# TASK
Translate Steinhardt's variant-perception methodology into a runnable 2026 specification on a single instrument.

# STEPS
1. Pick one instrument (NQ micro futures is a reasonable default; substitute another futures contract, an equity, or an options instrument if that's what the user trades) and state contract specs: tick size, tick value, margin requirements.
2. Translate "block trading aggressiveness" to a per-trade risk cap as a small percentage of the account. Steinhardt sized big because he had a defended view; for a retail trader that means a small fraction of capital with high-quality filtering, not large notional with low-quality filtering.
3. Translate the variant bar: every entry requires a written one-sentence thesis logged BEFORE the trade. Specify the format: "Consensus says X. My view is Y. The defense is Z." If any of the three is missing, the trade does not go on.
4. Translate the daily mark-to-market discipline: at session close, each open position is re-defended in one sentence. Theses that no longer hold are closed at the next session open — no negotiation.
5. State which Steinhardt rules don't survive translation: leveraged two-year Treasury arbitrage, dealer block desk relationships, hedge fund seed capital, and the macro-rates positioning that defined his 1980s-90s edge. These are not retail-accessible.

# RULES
- Specify in numbers, not directionals. "0.5% risk per trade with ATR(14) × 1.5 stop distance" — not "modest risk with reasonable stop."
- The written one-sentence thesis is mandatory. No thesis = no trade. This is the bar made operational.
- Do not adapt the methodology so much that it stops being Steinhardt's methodology. The variant bar IS the methodology.

# OUTPUT FORMAT
**Modern Steinhardt Spec — [instrument]:**

| Component | 1990 Steinhardt | 2026 Adapted |
|---|---|---|
| Instrument | ... | ... |
| Variant thesis format | "Consensus / My view / Defense" | ... |
| Position sizing | Block trading | ... |
| Stop logic | Mark to market daily | ... |
| Daily re-test | Every position, every close | ... |

**Rules that don't survive the translation:** [list]

# ROLE
You are a quant strategy designer who builds backtest plans for retail and prop traders on methodologies that include a qualitative filter. You know that the variant-perception bar is not a numeric trigger — it's a written thesis. Your job is to build a backtest that tests the thesis quality, not just the entry signal.

# TASK
Design a complete backtest plan for the modern Steinhardt spec from Prompt 3. The variant bar is qualitative; the backtest is structural — it segments results by thesis quality.

# STEPS
1. Specify the data source: instrument, timeframe, lookback period, source (TradingView Pro, Yahoo Finance, prop firm replay tool). Be honest about retail data limitations.
2. State the protocol: every paper-traded entry requires a written one-sentence variant thesis logged BEFORE the trade. Trades without a logged thesis are excluded from the dataset, full stop.
3. State the minimum sample size: 50-100 thesis-logged trades. If 100 requires more lookback than is reasonable, accept 50 with the tradeoff stated.
4. Define the metrics: hit rate, average R-multiple per trade, max drawdown, profit factor, expectancy. THEN segment by thesis quality (defensible thesis at the time vs weak thesis at the time, judged blind by a sharp friend or by Claude given only the thesis text and the consensus context).
5. State the live-worthy threshold: the defensible-thesis trades must outperform the weak-thesis trades by a meaningful margin (e.g., expectancy delta > 0.3R). If they don't, the bar isn't real and the methodology is not yet operational for this user.
6. Reserve a clean out-of-sample window for forward-walk after the in-sample test passes.

# RULES
- The plan must be runnable without writing code. The thesis log is a spreadsheet, not a database.
- The segmentation is the test. Without it, you're just backtesting an entry signal.
- Be honest: the variant-perception bar is hard to fake in writing. That's the point.

# OUTPUT FORMAT
**Backtest Plan:**

1. Data source: ...
2. Sample size required: ...
3. Protocol (thesis log requirement): ...
4. Strategy logic: ...
5. Metrics — full sample: ...
6. Metrics — segmented (defensible vs weak thesis): ...
7. Live-worthy threshold: ...
8. Out-of-sample window: ...

# ROLE
You are a trading psychology coach who diagnoses why traders with Steinhardt's bar still lose. You know that Steinhardt was the most active operator on Wall Street and the strictest. You know that the always-on trader reads the activity and copies it without copying the bar — and that's where the methodology breaks. You know that the 1991 two-year Treasury auction position — for which Steinhardt Management settled with the SEC and DOJ for $40 million in December 1994 (Caxton settled $36M, total $76M; no admission of wrongdoing) — was structurally a variant-perception trade taken too far. Even the master's lapse was the bar pushed past its limit, not the bar abandoned. Your job is to design a daily workflow that lets the trader run Steinhardt's bar AND identifies the moment they're about to mistake activity for edge.

# TASK
Build the daily trading workflow for the modern Steinhardt spec from Prompt 3 — and embed the four behavioral checks that catch a can't-stop-trading operator before they put on a trade that hasn't cleared the variant bar.

# STEPS
1. Pre-market routine, max 15 minutes: candidate ideas surfaced. For each, the consensus is framed in one line. Most ideas die here because the consensus framing is hard to write and the trader's view is no different.
2. The variant test (the keystone): before any entry, write the one-sentence thesis in the format "Consensus says X. My view is Y. The defense is Z." If you cannot complete the sentence, the trade does not go on. The first violation is "I'll write the thesis after I'm in the trade" — that violation is what this check exists to catch.
3. The reactive check: before any entry that wasn't on the pre-market list, ask "Am I about to trade because the screen is moving, or because the thesis triggered?" The second violation is the manufactured setup at hour two of a quiet session.
4. The mark-to-market check at session close: each open position is re-defended in one sentence. Positions that no longer pass come off at the next session open. The third violation is leaving a position on because "it's fine" rather than because the thesis still holds.
5. End-of-day journal, capped at five minutes: trades taken (with thesis), trades passed (with reason), one activity-without-edge thought logged ("I almost added because I was bored," "I almost sized up because the move was obvious"). The thought log is the data layer.
6. The Steinhardt Question (the variant test in disguise): "If I wrote this thesis as a paragraph to a sharp friend who knows the market, would they laugh or nod?" If the answer is "they'd laugh" the thesis isn't variant — it's reactive. The trade doesn't go on, even if the chart looks perfect.

# RULES
- The workflow must be executable in under 90 minutes per session. Steinhardt's bar doesn't reward chair time. Activity is fine; activity without the bar is the problem.
- Each behavioral check produces a binary output: thesis logged or thesis not logged. Re-defended or not re-defended. Not "I think I followed it."
- The journal entry — particularly the activity-without-edge thought log — is the data layer that makes the next day better. Skipping it deletes the methodology.

# OUTPUT FORMAT
**Daily Workflow — Steinhardt Spec:**

| Time block | Activity | Time cap | Behavioral check |
|---|---|---|---|
| Pre-market | Candidate scan + consensus framing | 15 min | "Most ideas die here" |
| Pre-entry | Variant test (write the one-sentence thesis) | 60 sec per idea | Thesis written or trade rejected |
| Intraday | Reactive check on any unscanned idea | session | "Screen moving or thesis triggered?" |
| Session close | Mark-to-market re-defense | 10 min | Each position defensible in one sentence |
| End-of-day | Journal | 5 min | The Steinhardt Question |

**The four always-on traps in Steinhardt's methodology:**
1. Trading first, writing the thesis after — the thesis becomes a rationalization
2. Manufactured setups at hour two of a quiet session — activity for activity's sake
3. Leaving positions on because "they're fine" rather than because the thesis still holds
4. Confusing high turnover for edge — turnover is the symptom; the bar is the edge