The CoinPicks Method.

Every term below is defined before it is used. If a word on this page isn’t here, it isn’t jargon.

This method was not derived from chart patterns or indicators. It was derived from the trading mechanism itself. Crypto introduced a market structure that had never existed: the automated market maker, where price is set by a formula against a public pool of money. That structure publishes, in real time, the one variable that governs how hard an asset can move — how much money is in its pool.

A thin pool has little money inside; a deep pool has a lot. Thin or deep decides almost everything that follows in this paper: how hard a coin moves, how large we are allowed to size, and how fast we take profit. Traditional markets hide this variable inside dealer inventories and dark pools. AMMs print it on the wall.

Drop a large buy into a thin pool and the price doesn’t drift — it gets displaced, shoved upward, because there is so little money in the tank. That splash is the strategy, and it names the whole method. The core discipline is simple to state: find low-liquidity opportunities that graduate into high-liquidity opportunities — and be in the pool before the crowd’s money arrives.

The identical $50,000 that barely nudges a $1M pool can multiply a $50K pool. The smaller the pool, the more work every dollar does. Because pools are public, this math can be run on any asset before a dollar is committed — the edge is verifiable, not claimed.

Timing follows from the same physics. Crowds arrive at tops and flee at bottoms; the operator must do the opposite. In practice the method accumulates into extreme fear and de-risks into extreme excitement — mechanically unremarkable, emotionally brutal, which is precisely why the edge persists. The one signal treated as a true risk-off event is the start of capitulation; short of that, positions are held through drawdowns by design.

The money gets rules before it gets coins. Every $100 in the portfolio is split by time horizon into three sleeves, each with one job, one hold period, and one exit style. The split is enforced in software (§09) — every trade is tagged to its sleeve when it is journaled.

The 70% sleeve is the deliberate center of gravity — position trading, which this method holds to be the most favorable risk/effort/return balance available to an individual: below investing, far above day trading, and never day trading. The 10% sleeve lives where the 100× outcomes are born; every miss is capped at its small stake, and its real second function is reconnaissance — the best coins it surfaces graduate up into the 70% engine. The 20% sleeve exists so that no sequence of events can remove the operator from the game.

The deepest property of the split is that it is one mechanic at three pool scales. The 20% runs displacement in the deepest pools — small moves, nearly impossible to sink. The 70% runs it in medium pools — large moves under explicit rules. The 10% runs it in the thinnest pools, where displacement is violent in both directions, which is exactly why it stays capped. Same physics, three risk envelopes.

Two structural rules complete the framework. Concentration: never more than seven active positions — beyond that, tracking quality collapses. Contribution: monthly additions may be held as dry cash and deployed into weakness rather than dripped in blindly.

The best traders in the world are right roughly 51% of the time. What makes them profitable is not accuracy — it is what happens to position size when they are wrong. This method treats risk management as the profit engine, not the seatbelt, and begins from the asymmetry every passive holder ignores:

A 68% drawdown needs a 233% recovery; an 85% drawdown needs 726%. Losses compound against you exponentially, which is why “do nothing and hold” is not a strategy — it is an unpriced bet that the asymmetry never lands on you. The method’s standing defenses:

• Laddered entries. Never one buy — a pre-planned ladder of roughly five orders whose sizes increase as price falls. Recovery mathematics improve with every rung, and the plan is written before the first dollar moves.
• The swing de-risk rule. The moment a short-term position shows profit, the initial capital plus 10% is pulled out as fast as possible. What remains is house money, managed on trend and Bitcoin’s price.
• Dry capital doctrine. Held cash is the substitute for forecasting accuracy. With reserves ready, being early or imprecise is survivable; without them, even correct theses die in drawdowns.
• Profit is extraction. Winning is defined as cash pulled out of the market — denominated in dollars (USDC on-chain), not in tokens held. Paper gains are not results.
• Sell strength, not weakness. When cash must be raised, elevated positions are trimmed — never the beaten-down bag at the bottom of its range.
• One rule above all. Never bet money you cannot afford to lose, and size every trade so one bad day cannot wipe you out.

Selection runs as a funnel with the same three gates in the same order, applied to every candidate without exception. Notably absent from the gates: market capitalization. After a thousand-plus recorded iterations of the checklist, market cap was dropped entirely as a screening variable — pool liquidity replaced it, because the pool, not the ranking table, is what a position must eventually exit through.

Every graded coin receives a stat card, every stat carries its reason, and every report is required to state the bear case — what would make this fail — before any bull case is considered. A refusal is a result: “no” protects capital exactly as effectively as “yes” grows it.

Above the per-coin funnel sits a research program built on narrative economics — the study, formalized by Nobel laureate Robert Shiller, of how stories spread through markets. The program’s working hypothesis: crypto assets appreciate through virality, and virality clusters into a small set of persistent story families that repeat every cycle. The utility claim on a project’s website is itself an unverified story; the measurable object is the narrative and its historical price behavior.

• The winner census. Thousands of prior-cycle high-multiple coins (top ~700 per cycle) analyzed and reduced to 326 full-cycle winners, each classified into a persistent narrative family and scored.
• The window study. A separate model isolates the coins that produced the highest multiples specifically in the window between a cycle bottom and the return of momentum — a different question than "what won eventually."
• The consensus filter. Only assets confirmed by BOTH independent models survive to the working shortlist. Agreement between two differently-built models is the point; either alone is a weaker instrument.
• The frontier: crowd tracking. The next instrument follows the buyers themselves — tagging the early-buyer wallets of past 20–30× coins on-chain and measuring where that cohort rotates next. Seven years of wisdom in one sentence: study the people buying.

Direction is read at three nested scales — the multi-year credit cycle (the tide), Bitcoin’s four-year halving cycle (the engine), and the short altcoin bursts that fire after halving tops (the spray). When the cycles agree, the method presses; when they argue, it waits. Technical analysis is deliberately demoted to one automated input among many — price behavior is evidence, not oracle.

The program’s proprietary instrument is Timeline Price Analysis — a methodology in continuous development since 2023 — industrialized as the Timestamp system. Its premise: price is the only ground truth in markets, so an event’s importance is measured by what price actually did when it hit, not by how loud the coverage was.

The daily self-verification step is the part with no precedent we know of: the research maintains its own truth instead of decaying, and a report’s trustworthiness becomes an explicit function of the computational power spent verifying it. Structurally, it behaves like a proof system — agents are to these reports what hash power is to a chain.

The discipline layer matters as much as the instrument: the thesis changes on structural evidence, never on price alone. A rally is not a regime change; a fundamental break in the model is. And “unclear” is a real answer — when the picture is mixed, the system says so rather than manufacturing confidence. This caution is grounded in one more research finding: much of what markets read as organic narrative is manufactured perception, seeded deliberately and traded up the media chain. Stories are therefore scored and verified — never trusted.

The split (§04) sets each sleeve’s budget; then each coin’s liquidity sets its size. Thin pool: smaller buys, faster profit-taking. Deep pool: larger size, more patience. Every trade enters with a written plan — entry ladder, exit stages, profit target — and a journaled conviction level and risk classification, before a dollar moves.

The long sleeve has its own machinery. Selling is forbidden there by mandate, so liquidity comes from borrowing against the position instead: a planner computes safe loan size from real income, expenses, and cash — not from appetite — and live-tracks loan health (LTV) across venues. Conservative defaults sit far from liquidation, deviations require written reasons, and the mandate’s side effect is structural: unrealized gains finance life without ever becoming taxable sales.

Tax strategy is otherwise kept strictly separate from trading strategy: a 10–20% tax on realized profit is never allowed to argue against exiting an asset that can drop 60%. Profit first; optimization after.

Each research program above ran as a manual standard operating procedure for seven years — hypothesis, test, analyze, adjust, thousands of recorded iterations — before any of it was automated. The automation principle is deliberately unromantic: the AI is not smarter; it does brute-force labor no human team could afford.

The throughput deltas are the argument. Coin research: 3 fully-researched coins per day with a full-time human researcher, versus 100+ per day — from thousands scanned — with agent fleets. Direction research: the equivalent of hundreds of parallel research operations dispatched in minutes. Execution routing: a ~45-minute manual scan of venues for best price, reduced to about one minute of automated pool-scanning across centralized and decentralized exchanges at once.

The member-facing machine is the CoinPicks Terminal: one interface connecting exchange accounts, on-chain wallets, and bank accounts; orders placed in natural language; entry ladders and staged take-profits placed cross-venue; and the discipline layer — sleeve tagging, conviction journaling, risk parameters — built into the order flow itself. Bypassing the discipline is possible, but the system makes rebellion harder than compliance.

Two architectural boundaries are absolute. First, analysis is separated from execution: the research layer vets and emits; a deterministic execution layer, outside the model, places orders under the operator’s standing mandate. Second, the tools ship as readable files members can inspect, modify, and extend in their own AI environment — connect a new venue, lower the risk profile, embed a strategy — because the software is not the moat. The refined judgment inside it is.

Binary signals destroy information. Every buy and sell call in this system is graded on a seven-level intensity scale — from very light to all-in — so conviction travels with the call, position sizing has a native input, and the record can be audited honestly afterward.

Calls are published to the community as they are made, timestamped, wins and losses alike — the operator trading his own decisions in front of the room. The scale is also the bridge between research and sizing: a level-2 buy in a thin pool and a level-6 de-risk into strength are complete instructions, not vibes.

The method’s development loop is the scientific method run against live markets: form the hypothesis, commit real capital under rules, record the outcome, adjust the checklist, repeat — for seven years, across thousands of iterations per program. Abandoning market cap as a variable (§06), demoting technical analysis (§07), and the never-more-than-seven rule (§04) are all scars from that loop, not aesthetic choices.

Larger claims are decomposed before they are believed: a macro theory is broken into individual points of truth, each verified separately, before the theory earns any weight in the model. The same standard is applied inward — this paper states where its own numbers are clean-room simplifications, and the Timestamp system re-verifies its own published research daily rather than letting it decay.

And the receipts are structural, not curated: timestamped calls with graded intensity, losses published alongside wins, public pool data anyone can recompute, and a free curriculum that teaches the entire method to anyone who wants to check it against reality.

• Crypto is volatile. You can lose money — including money you were sure about. Nothing in this paper is financial advice, and nothing here promises a return.
• Displacement is symmetric. The thin pools that multiply gains multiply losses identically; the sizing rules are the method, not an accessory to it.
• Direction is probabilistic. The instruments produce a weighted read, never a certainty, and the system reports "unclear" rather than inventing confidence. Predictions carry error and always will.
• The models vet narratives, not coins. No shortlist from any instrument in §06 is a buy list; per-asset diligence is never skipped.
• Hit rate honesty: the best operators are right barely more than half the time. Anyone promising accuracy is selling something else.
• The automation frontier is young. Autonomous execution runs small and supervised; capability claims in §09 describe our systems as tested to date, not guarantees of behavior.
• Narrative scoring quantifies qualitative reality. It is scored interpretation with stated confidence — treated as evidence, never as fact.
• Borrowing against collateral carries liquidation risk. LTV discipline is mandatory, and no loan is safe by default.
• This is not a get-rich-quick system. It asks for real hours (§04’s sleeves carry real weekly time budgets), real discipline, and tolerance for being early.

The method was developed by Alexander Lorenzo across seven years of live trading, refined through the community he teaches, and encoded into software by the CoinPicks engineering effort. This paper is maintained as a living document: it is corrected when the method is corrected, and each revision is logged below.