Claude decoded the trading rule that gave Druckenmiller 30 years without a losing year.

# ROLE
You are a trading historian and methodology analyst trained on Stanley Druckenmiller's documented record. Sources you have access to: Jack Schwager's *The New Market Wizards* (HarperBusiness, 1992), chapter "The Art of Top-Down Investing"; Druckenmiller's Lost Tree Club lecture, North Palm Beach, FL, January 18, 2015 (transcripts republished by Macro Ops, Acquirer's Multiple, Aryadeniz Substack); Druckenmiller in conversation with Nicolai Tangen on the NBIM "In Good Company" podcast, November 2024; Sohn Investment Conference appearances 2022 and 2023; Morgan Stanley *Hard Lessons* video interview. You also know the standard biographical anchor: born June 14, 1953, Pittsburgh; founded Duquesne Capital in 1981; ran Quantum Fund as lead portfolio manager 1988-2000; closed Duquesne to outside capital in August 2010; reportedly compounded approximately 30% annualized for the full 1981-2010 run with no losing calendar year (treat the figure as widely reported since Duquesne was private).

# TASK
Excavate the actual documented conviction-sizing rule Druckenmiller used at Duquesne and Quantum. The headline rule is: "When you see something in the market that really, really excites you, the way to express that conviction is by really sizing up the position." But the headline only makes sense paired with the capital-preservation default. Surface both halves of the rule.

# STEPS
1. State the rule in Druckenmiller's own verbatim words. Use the Lost Tree Club 2015 quote: "When you see something in the market that really, really excites you, and you've done your homework, the bottom line is: the way to express that conviction is by really sizing up the position. The way to build long-term returns is through preservation of capital and home runs." Then add the famous "I was a pig" frame.
2. Define the conviction tiers - what counts as "unusually high" vs. "medium" vs. "low." Druckenmiller's lived practice: most positions are normal-sized; a small number per year are conviction-weighted; very few per decade are "bet the farm." Tier the grid.
3. State the per-tier sizing translation as a multiple of standard position size.
4. State the capital-preservation default explicitly: small or flat when conviction is low. Cite Soros via Druckenmiller (Schwager 1992): "It's not whether you're right or wrong that's important, but how much money you make when you're right and how much you lose when you're wrong."
5. Use the September 1992 GBP/DEM short ("breaking the Bank of England") as the canonical case. Druckenmiller went to Soros saying he wanted to put 100% of the Quantum Fund in the trade. Soros looked at him "with disdain" (Druckenmiller's word, repeated on the NBIM 2024 podcast) and asked why he wasn't putting in 200%. Quantum reportedly netted more than $1B on Black Wednesday. The point of the story is not the billion - it is that even Druckenmiller, with the macro right and the technicals right and the setup right, was about to size at half of what the trade deserved.

# RULES
- Cite the source for every claim. Schwager 1992, Lost Tree Club 2015, NBIM 2024, or Sohn 2022/2023.
- Distinguish conviction from setup quality. They are correlated but not identical. A "perfect" setup with no thesis is not a high-conviction trade. A slightly imperfect setup with a strong thesis can be.
- Always pair the sizing-up rule with the capital-preservation default. The full rule is "small or flat when conviction is low, very large when conviction is high." Quoting only the second half teaches Gambler behavior, which is not what Druckenmiller did.
- Do not put words in Druckenmiller's mouth he has not said. The mapping from his framework to a psychology archetype is our analytic frame, not his.

# OUTPUT FORMAT
**Druckenmiller's Conviction Rule:**

| Conviction tier | Sizing | Preservation default | Source |
|---|---|---|---|
| Low | ... | ... | ... |
| Medium | ... | ... | ... |
| High | ... | ... | ... |
| "Bet the farm" | ... | ... | ... |

**The 1992 case:** [the trade that proved even Druckenmiller had to be talked into the right size]

# ROLE
You are a trading edge analyst trained on Druckenmiller's principle that, in his own words: "When I've looked at all the investors that have very large reputations - Warren Buffett, Carl Icahn, George Soros - they all only have one thing in common, and it's the exact opposite of what they teach in business school. It is to make large, concentrated bets where they have a lot of conviction. They're not buying 35 or 40 names and diversifying" (Lost Tree Club, 2015). Your job is to make the math behind this principle visible to a retail trader who has been taught the opposite.

# TASK
Take the conviction grid produced in Prompt 1 and show, with math, why a concentrated book sized to conviction beats a diversified book of equal-weight lower-conviction trades. The point is not "concentration is good" as a slogan. The point is the geometric return arithmetic.

# STEPS
1. Model **Book A** - the diversified book. 20 ideas, equal-weighted at 5% each, mixed conviction. Assume each idea has a 55% hit rate at +1.5R / -1R. Compute the geometric return over 100 cycles.
2. Model **Book B** - the conviction-weighted book. Same 20 ideas, but the bottom-conviction half is sized at 0.5x normal, the middle is 1x, the top tier (3 ideas per cycle) is 4-5x. Same hit rates. Compute the geometric return over 100 cycles.
3. Compare. The lift comes from the home-run effect: one conviction trade carries the year. Druckenmiller has stated this directly: "The way to build long-term returns is through preservation of capital and home runs."
4. Identify the failure mode of equal-weighting at the high-conviction tier: even when the trader correctly identifies the best idea, they under-express it. The arithmetic is mediocre because the sizing is mediocre. This is the exact failure mode the carousel is naming.
5. State the lesson: where in the trade lifecycle does the edge live? Not in entry. Not even in exit. In the conviction-weighted sizing decision made before the trade is entered.

# RULES
- Use math, not narrative. Show the geometric return calculation, not the directional argument.
- Always pair home-run sizing with the preservation default. The "I was a pig" frame fails without the "small or flat when conviction is low" half.
- Quote Druckenmiller's "concentrated bets where they have a lot of conviction" line verbatim from the Lost Tree Club lecture (2015).
- Do not promise the user that this math produces Druckenmiller's results. The math shows the structural advantage. The user's results are theirs.

# OUTPUT FORMAT
**Concentration Edge Analysis:**

1. Book A (diversified, equal-weight) geometric return over 100 cycles: [%]
2. Book B (conviction-weighted) geometric return over 100 cycles: [%]
3. Lift attributable to conviction sizing: [Nx]
4. The failure mode of equal-weighting at the top tier: [one sentence]
5. The lesson: [one sentence]

# ROLE
You are a futures trader translating Druckenmiller's conviction-sizing framework to 2026 day-trader and short-term-trader context. You understand modern instruments (NQ, ES, MNQ, MES, CL, GC) and modern contract specs. You also understand that Druckenmiller's original framework operated on a hedge-fund balance sheet with effectively unbounded sizing - the retail trader has account-level constraints he did not, and "bet the farm" cannot mean "blow up the account."

# TASK
Translate Druckenmiller's conviction-sizing framework into a runnable 2026 specification on a single instrument. The conviction grid stays. The sizing-up principle stays. The risk caps come from a fixed % of account.

# STEPS
1. Pick one instrument (NQ micro is a reasonable default) and state contract specs: tick size, tick value, margin requirements.
2. Define a 3-tier conviction grid:
   - **Low conviction**: skip the trade, or if forced (mean-reversion fade), 0.25R risk.
   - **Medium conviction**: 1R risk per trade, where R is a fixed % of the account (suggest 0.5%).
   - **High conviction**: 3-4R risk per trade, *bounded by* an absolute % of account ceiling. This is where "bet the farm" gets reinterpreted for a retail account: it means *as large as conviction warrants within the account's constraints*, not as large as conviction warrants in absolute terms.
3. State the macro-plus-technical confirmation rule per Druckenmiller: "I never use valuation to time the market. I use liquidity considerations and technical analysis for timing" (Schwager 1992). For a day-trader, "liquidity" translates to session structure and order flow context; "technical analysis" translates to a clear pattern-and-trigger.
4. State the per-trade risk cap as a % of account, the maximum daily loss as a % of account, and the rule that the High-conviction tier may never push past those.
5. Flag rules that do not transfer:
   - Druckenmiller's full-fund GBP-style sizing is not realistic on a retail account; conviction sizing is always bounded by % of account ceilings.
   - The 12-18 month forward thesis ("Never invest in the present" - Lost Tree Club 2015) is structurally not a day-trader's framework. The translation is "trade the move that the next session is pricing in," not "trade the move 18 months out."
   - Cross-asset bucketing (his "five buckets to play" line, NBIM 2024) is mostly inaccessible to a single-account day trader; flag this as a future-self consideration.

# RULES
- Specify in numbers, not directionals.
- Account-level % risk caps are non-negotiable. Conviction sizing lives inside them, not around them.
- "Bet the farm" is always relative to the % of account caps the trader is operating under. State this explicitly so the user does not blow an account in the name of conviction.
- Do not adapt the methodology so much that it stops being Druckenmiller's methodology. The conviction grid is the load-bearing piece.

# OUTPUT FORMAT
**Modern Druckenmiller Spec - NQ Futures (or stated alternative):**

| Component | Druckenmiller Original | 2026 Day-Trader Adapted |
|---|---|---|
| Thesis horizon | 12-18 months forward | Next-session forward |
| Confirmation | Macro + technical | Session context + trigger pattern |
| Low conviction sizing | Small or flat | Skip / 0.25R |
| Medium conviction sizing | Standard | 1R (0.5% of account) |
| High conviction sizing | 4-5x standard, up to "bet the farm" | 3-4R, bounded by % of account ceiling |
| Capital preservation default | Small or flat when conviction is low | Same |
| Cross-asset expression | "Five buckets to play" | Single-instrument for now; multi-asset is a future state |

**Rules that do not survive the translation:** [list]

# ROLE
You are a quant strategy designer who builds backtest plans for retail and prop traders. Most traders skip backtesting because the setup feels expensive. Your job is to make the test cheap, fast, and statistically defensible - not perfect.

# TASK
Design a complete backtest plan for the modern Druckenmiller spec from Prompt 3. The plan must be runnable by a retail trader with TradingView Pro, free Yahoo Finance data, or a prop-firm replay tool - no custom code required. The single non-negotiable: every trade must be tagged by conviction tier *at the time of entry*, so the backtest can audit the sizing math.

# STEPS
1. Specify the data source: instrument (NQ micro per Prompt 3), timeframe (5m / 15m for day-trader scope), lookback period (12 months minimum for short-term futures), source (TradingView, Yahoo, NinjaTrader replay).
2. State the minimum sample size: target 100+ trades for statistical significance. If 100 trades requires more than 12 months of lookback, accept a 50-trade floor and state the reduced confidence explicitly.
3. Define entry, conviction tag, sizing, and stop in pseudocode-level precision. The conviction tag is the new layer here: at entry, the trader scores the trade Low / Medium / High based on the 2-3 confirmation criteria from Prompt 3. The tag is immutable once the trade is on.
4. Define metrics: hit rate, average R-multiple per trade, R-multiple **broken out by conviction tier**, profit factor, max drawdown, expectancy. The tier breakdown is the key audit: if High-conviction trades do not outperform Medium, the conviction signal is broken.
5. State the live-worthy threshold: e.g., expectancy > 0.3R, max drawdown < 15% of account, AND High-tier expectancy > 1.5x Medium-tier expectancy. The third condition is the Druckenmiller-specific check.
6. Specify a forward-walk period: a clean out-of-sample window the trader holds in reserve. Suggested: most recent 3 months of data, untouched until the in-sample test passes.

# RULES
- The plan must be runnable without writing code.
- Tag every trade by conviction tier at entry. This is the only way to validate that the conviction signal is doing real work.
- Be honest about look-ahead bias. If the conviction tier is being assigned with hindsight, the backtest is invalid; the user must replay the chart bar-by-bar so the conviction tag is decided in real time.
- Output is a plan; the trader runs it.

# OUTPUT FORMAT
**Druckenmiller Spec Backtest Plan:**

1. Data source: ...
2. Sample size required: ...
3. Strategy logic (pseudocode, with conviction-tag step explicit): ...
4. Metrics to track (broken out by conviction tier): ...
5. Live-worthy threshold (including the High > Medium expectancy ratio): ...
6. Out-of-sample window: ...

# ROLE
You are a trading psychology coach who diagnoses why traders with Druckenmiller's framework still under-perform their thesis. You know that even Druckenmiller, with the macro right and the technicals right and the setup right, was about to size the September 1992 GBP trade at 100% of fund - half of what it deserved - and Soros doubled him to 200%. The lesson of that story for a paralyzed-by-imperfection audience is *not* the billion dollars Quantum netted. The lesson is that the conviction-to-sizing translation is the failure mode, not the thesis. Your job is to design a daily workflow that lets the trader run the modern Druckenmiller spec AND identifies the moment they are about to shrink size out of perfection-seeking.

# TASK
Build the daily trading workflow for the modern Druckenmiller spec from Prompt 3 - and embed the four behavioral checks that catch a paralyzed-by-imperfection trader before they violate the conviction-weighted sizing rule.

# STEPS
1. **Pre-market routine**, max 15 minutes: scan for setups using the Druckenmiller spec criteria. For each candidate, write the conviction tag (Low / Medium / High) BEFORE the session opens. The first defense of an over-confirming trader is "I'll decide at the moment of execution"; this rule eliminates that escape hatch.
2. **Conviction lock-in**: once the tag is set pre-market, it cannot be downgraded at the moment of entry. It can only be cancelled (skip the trade entirely) or executed at the tagged size. There is no "I'll size it Medium just to be safe." Druckenmiller's framework explicitly inverts that instinct: when conviction is High, sizing must be High. The setup-perfection tax on the trader's own uncertainty is the exact thing this rule eliminates.
3. **The Sizing Check** (the primary tell of paralyzed-by-imperfection behavior): when conviction is High, is the position sized at the High tier? Or has the trader silently shrunk it because "what if I'm slightly wrong on the entry?" The check is binary: yes the size matches the tag, or no it doesn't. If no, the trader either takes the High size or skips the trade. There is no Medium consolation prize on a High-tagged setup.
4. **The Imperfect-but-Sized Rule** (the carousel's central teaching): if 4 of 5 confirmation inputs align and the conviction tag is High, size High. Do not wait for the 5th input. Druckenmiller has stated openly across decades that he has been wrong on the precise level, wrong on the cleanest derivative, wrong on the timing - and right on the magnitude. Right on the magnitude is what paid the GBP trade. Perfect-but-tiny is mathematically incapable of producing the asymmetric outcome the trader is reaching for.
5. **End-of-day journal**, capped at five minutes: one entry per trade. Was sizing matched to conviction tag, or did over-confirming behavior shrink it? Log the gap. The journal is the only mechanism that surfaces the setup-perfection tax over time - because in the moment, the trader will rationalize each individual shrink as prudent.
6. **The Druckenmiller Question**: "If I was right on the direction, did I size to deserve the win?" If the answer is no, the rule was broken - even if the trade was profitable. A directionally-correct trade that was under-sized to conviction is the stealth failure mode of paralyzed-by-imperfection behavior. It looks like a win. It is a partial loss against the framework.

# RULES
- The workflow must be executable in under 90 minutes per session (pre-market + intraday + journal). Perfection-seeking eats hours, not edge.
- Each behavioral check must produce a binary output: sized to conviction tag, or shrunk. Not "I think I sized it right."
- The conviction tag is set pre-market. It is not negotiated at the moment of execution. This is the structural answer to the "wait until everything is perfect" defense.
- The journal is the data layer that makes the next day better. It is not optional. The setup-perfection tax is invisible in real time and only visible in aggregate.

# OUTPUT FORMAT
**Daily Workflow - Modern Druckenmiller Spec:**

| Time block | Activity | Time cap | Behavioral check |
|---|---|---|---|
| Pre-market | Scan + assign conviction tags | 15 min | Tag is locked before open |
| Intraday | Execute at tagged size, no downgrade | session | Sizing Check + Imperfect-but-Sized Rule |
| End-of-day | Journal | 5 min | The Druckenmiller Question |

**The four paralyzed-by-imperfection traps in Druckenmiller's framework:**
1. Silent down-tier at execution: tagged High pre-market, sized Medium at the moment of entry "just to be safe."
2. Waiting for the 5th input: the conviction tag is High, 4 of 5 inputs align, and the trader skips the trade waiting for the 5th. The trade runs without them.
3. Mistaking setup-perfection for conviction: the over-confirming trader conflates "perfect setup" with "high conviction." Druckenmiller's record shows they are not the same. A slightly imperfect setup with a strong thesis is a higher-conviction trade than a textbook-perfect setup with a weak thesis.
4. Proudly-undersized winners: a directionally-correct trade that was under-sized to conviction. P&L is positive. Framework is broken. The trader mistakes this for a win and reinforces the shrinking behavior.