Entry-timing teardown

Goal: say whether and when to buy a name, grounded in live data and the underlying financials, delivered as an honest broker — not a cheerleader. Lead with the financials and the driver of price action before any buy call. Be willing to say "don't buy" or "not yet".

Before starting, read the investor's own rules if they exist (a journal with a firewall, position caps, whether they use stops, their edge). The edge this method is built for is being early on structural theses: size at entry, don't chase once it's consensus.

The 6 passes

1. Live snapshot. Current price; position in the 52-week range; % moves (1w / 1m / YTD / vs all-time high); valuation (forward P/E, EV/EBITDA, FCF yield) vs its own history and peers. Don't work from memory; pull it live.

2. Diagnose what's actually driving the recent move. Separate two things that look identical on a chart:

   • Fundamental thesis — demand, contracts, guidance, margins.
   • Technical / mechanical — what distorts price independent of the business: M&A stock issued to sellers (e.g. a utility's acquisition paid partly in ~50M shares on a staged 18-month lock-up, with sponsors distributing via block sales that cap the stock), lock-up expiries, secondary offerings, PE-sponsor distribution schedules, index add/drop flows, rate sensitivity, sector rotation.
   • The tell that it's technical: the stock falls on good news. A clean float with no overhang (e.g. a clean spin-off with no lock-up) means the price is "real" — no mechanical entry to wait for.
   • The absence of an overhang is a finding, not a non-finding — and it argues for SMALLER size. Always search for one explicitly and state which case you're in:
     • Overhang present -> the gap closes on a date, whether or not anyone changes their mind (e.g. a second lock-up tranche releasing on a known date, a secondary already cleared above the current price). A beat nobody rewarded is then holdable, because "why won't it move" has a mechanical answer.
     • No overhang -> the de-rate is narrative, so the re-rate needs the market to genuinely revise a belief. Slower, less certain, no forcing date. A company that is a net buyer of its own stock with a shrinking share count can still be a good business at a real discount — but size it as a starter and say so.
     • Check the buyback/share-count direction while you're here: a shrinking share count means the mechanical flow is a bid, not an offer.

3. Map the catalyst calendar. Earnings date (binary for rich momentum names — never initiate the day before at a high multiple); regulatory decisions; policy/legislative events; seasonality tested with actual data, not eyeballs (price_history.py) — flag small samples and data-mining, and check whether a "seasonal" is really a recurring catalyst (e.g. a sector's "August" that is policy-news clustering, not a calendar law). Identify the dated trigger the entry hangs on.

4. Locate on the recognition spectrum — FOUR ORDERED TESTS, all must clear. Usually the crux of the buy/no-buy call. Run recognition.py first; it measures R1-R3.

   #	test	fails when	example kill
   R1 — is there a crowd?	>= 4 covering SELL-SIDE brokers with live targets. Quant/model services (Zacks, Weiss, CO FORECST, CFRA) are excluded from every count.	< 4 -> UNCOVERED: no crowd whose mind can change	a name with 2 brokers whose last opinions were 20 and 37 months old
   R2 — is the crowd falsifiable?	>= 1 sell-side SELL.	0 sells -> CONSENSUS: no bear to convert	names rated 20/4/0 and 21/3/0
   R3 — has the crowd ALREADY turned?	consensus target <= ~10% off its own trough**, OR **>= 20% upside left	already re-rated -> MID-FLIGHT	a target +42% off its trough with only +6.7% upside left
   R4 — is the crowd WRONG?	forward EPS must EXCEED trailing, AND forward P/E must be BELOW trailing P/E	the multiple rises going forward -> "cheap" was a trailing artifact	fwd P/E 17.6 vs trailing 11.6 with EPS guided -36%

   R4 is the load-bearing test and it is deliberately NOT automated. R1-R3 describe the shape of opinion; only R4 asks whether the opinion is RIGHT. Without it the framework walks into a falling knife: the one name in 86 where the crowd had not turned (0 Buy / 7 Hold / 2 Sell, target still falling) was a kill. The crowd had not turned because the crowd was right — a value screen with no wrong belief to arbitrage.

   R4 needs a second feed and a guidance check. Take forward EPS from a second source (e.g. your broker's quote API — E*TRADE /v1/market/quote/<SYM>?detailFlag=ALL -> estEarnings), then verify it against CURRENT company guidance: a consensus figure that predates a guidance cut inverts the test.

5. Underwrite the financials before advising. Revenue growth and quality (organic vs acquisition vs cannibalization), gross/operating margins, FCF and GAAP net income (is "adjusted EBITDA" masking a net loss?), balance sheet (net debt, cash), share-count trend (serial dilution?). Real business vs story stock. Pull figures with sec-financials.

5b. Underwrite the PEOPLE — run leadership-teardown. Who runs this, what have they said in public, how did they leave their last job, and — if the company is splitting, merging or reorganising — where did the operating talent go? This pass can downgrade or explain, but it can NEVER promote a name that failed pass 1 or pass 4: a great new CEO does not fix 80x forward with zero sell ratings.

6. The buy plan — concrete and honest.
   • A specific answer: buy now / wait for <dated trigger> / don't add. Pick real dates (overhang-clear estimate, earnings date, catalyst window).
   • Scale in in 2-3 tranches. If no stops are used, size is the only risk control. A starter answers FOMO; the reserve respects buying ahead of a catalyst.
   • Write every tranche as a PERCENT OF BOOK plus a dollar figure — never as a share count. A share count silently inflates the position if the price rises between plan and fill (a plan written as "37 shares = 5% of book" at $216 is 5.23% at a $226.74 fill, overrunning the next tranche's earmark). State the target percent and recompute shares at the price on the day the order is placed.
   • Give every tranche a gate that is a condition, not a date alone — a price ceiling with its valuation logic ("<= $227 = 19.5x fwd; above that you're paying a premium multiple for a de-rating story, which is self-contradictory"), or a named print. A gate whose why is written down survives being tested; a bare number gets rationalized around.
   • Sizing inside a firewall — speculative single-name bets live in the active sleeve only, sized like money you can watch go to zero; the retirement core stays diversified.
   • Honest-broker delivery — distinguish thesis from price; push back plainly if it's FOMO/chasing; recommend against low-value adds. Make a framework point once, then advise forward; don't turn past losses into a recurring scold.

Tooling (bundled)

• Live quotes / news / analyst targets / deal terms: web search. Verify suspicious prints (a "crash" with no news is usually bad data).
• Price history & seasonality: python price_history.py <TICKER> — ~10y of daily bars from api.nasdaq.com, no key, browser User-Agent; prints trailing windows and month-by-month seasonality. Dead sources, don't retry: the Yahoo chart API hard-429s, Stooq sits behind a JavaScript proof-of-work wall. Fallback for live quotes only: quote.cnbc.com/quote-html-webservice/restQuote/....
• Recognition distribution AND its derivative: recognition.py — R1-R3 automated.

  pip install requests
  python recognition.py DKS --series --brokers      # one name, full history
  python recognition.py @tickers.txt --csv out.csv  # a list
  

  Source: api.nasdaq.com/api/analyst/<T>/targetprice, which returns historicalConsensus, a ~13-month MONTHLY series of {buy, hold, sell, consensus target}. The 0-sells test is a SNAPSHOT; the series gives the derivative — whether the crowd has ALREADY changed its mind. y in that series is the consensus TARGET, not the stock price; the tool re-proves this at runtime and refuses to score if it fails.
• Market-implied macro: python kalshi_macro.py fed cpi — Kalshi's public contracts give a distribution for the Fed path, CPI, payrolls, unemployment and recession timing, so "the market says a cut" becomes "73% hold, 25% hike." Read it, never trade it: the fee is 7%·p·(1-p), and benchmarks of LLMs trading real order books mostly lose money. Where it disagrees with CME fed funds futures, the gap itself is worth explaining.
• Seasonality is ~10 observations per month. Report n, mean, median AND win-rate, and say plainly that small-n is data-mining. Check whether a lopsided month is just the name's beta to a market-wide seasonal. Never let a seasonal override a thesis; at most it is a pre-commitment device ("red in this month is expected, not information").
• Sector context first: sector-acceleration. Positions: etrade-pull.

Output shape

Snapshot -> what's driving the move (thesis vs technical) -> catalyst/earnings/seasonality calendar -> recognized-or-not -> financials -> people -> dated, tranched buy plan with a clear recommendation. End with the single most important date or trigger to watch.

Not investment advice; a research method.