Large, established companies sometimes fall hard on a single headline. We research whether the drop is actually justified, and when it is not, we still do not buy — not until price itself confirms the recovery. Positions are held for a 5 to 15 percent move.
This is the slower, smaller sibling of the SPY options service. Gains per position are more modest and they take weeks rather than hours. Most candidates never become trades at all — they are discarded, invalidated or expired, and every one of them stays on the ledger with its reason.
3 candidates on the ledger · 0 reached entry
Once a day, after the close, we screen a universe of large, liquid companies for drops that are unusual for that specific stock — not just large in absolute terms.
Live parameters, read from the engine. Every candidate also stores the exact values it was judged under.
A stock that fell because the market fell is not a dislocation. It is beta.
Both tests have to pass, and the sector-relative one is the test that matters. A six percent drop means something very different in a utility than in a high-beta name, so the first test asks whether the move is extreme relative to that company's own trailing volatility.
The second asks whether the company fell further than its sector did on the same day. If the whole sector was down and this name simply came along for the ride, there is no company-specific news to be wrong about, and no overreaction to recover from. Buying it is a bet on the market, which is not what this strategy is.
Symbols already holding an open position or a live candidate are skipped, so the same dislocation is never counted twice.
Every candidate is researched against the actual news that caused the fall, then classified into exactly one of three verdicts. The full reasoning, the biggest risk and what would invalidate the thesis are all published on the ledger.
The news permanently impairs earnings power. A guidance cut with a real demand story behind it, accounting irregularities, a dividend cut, the loss of a major customer, regulatory action that changes the business, emergency dilution, a patent cliff.
Discarded permanently. No override, at any conviction.
Real, but recoverable. A weak quarter, a currency or input-cost headwind, a one-off charge, a supply disruption, litigation with bounded exposure. The damage is genuine and the business is not broken.
Eligible, if conviction clears the bar.
The selloff is not supported by new information. A sympathy move off a peer's bad news, a downgrade containing nothing new, index or rebalancing flows, a thin short report, a misread headline.
Eligible, if conviction clears the bar.
A favourable verdict never buys anything. It only arms the candidate — and only at a conviction of 3 or better out of five. Price confirmation is still mandatory, and an armed candidate that never gets that confirmation simply expires after 15 sessions, unbought.
Structural verdicts are discarded and can never reach entry through any path, including a manual override. If the read was wrong, the candidate has to be researched again from scratch.
This is what separates this from buying the dip. Being right about the news is not enough — price has to prove the recovery in three explicit stages before a single share is bought.
Price closes at least 2% above the post-drop low, and that low is at least one session old. The low and the high of the bounce are both recorded. Nothing is bought here.
The knife has stopped falling.
Price pulls back toward the low, coming within 3% of it, but holds above it — a higher low. If it closes more than 1% below the recorded low instead, the candidate is invalidated on the record and is never re-armed.
The low is holding, not breaking.
Price closes back above the bounce high. Only now does a position open, in one normalized lot, with the stop already placed under the retest low. This is the only event that ever opens a position.
The recovery is moving, not hoped for.
Every exit is decided before entry. There is no discretionary holding on and no adding to a loser.
At most 10 positions open at once, and at most 2 in any single sector. One position per symbol, ever — a second dislocation in a name we already hold is not an invitation to double up.
New entries halt entirely while SPY is more than 10% off its recent peak. A market-wide dislocation breaks the idiosyncratic premise the whole strategy rests on: if everything is falling together, a single name falling is not telling us anything about that name.
Blocked entries are recorded as blocked. They do not quietly disappear from the ledger.
Second Helping has no track record yet. Join the list and you will see it accumulate in public from its very first candidate — including the ones we pass on.
You would be the first on this list
Drop us a line at hello@poutinecapital.com.