Getting the direction right and still giving the money back on the way down is one of the most common ways a good call turns into a break-even trade. This is why exits are first-class in Quantustik's trade plan: a TP1/TP2/TP3 ladder for banking profit in stages, and a stop that scales with volatility instead of sitting at an arbitrary fixed percentage.
A single fixed take-profit target forces an all-or-nothing bet on exactly how far a move goes: sell too early and the position leaves obvious upside on the table; wait for a distant target and a round trip back to break-even (or worse) is the most common way a correctly-timed trade ends up going nowhere. Banking most of a move in stages beats holding out for all of it and giving the gain back on the way down — a discipline preference this feature exists to serve, not a measured performance claim.
A single fixed-percentage stop has the mirror problem on the downside: a stop set at, say, a flat 5% is far too tight for a stock that normally swings 4% a day and far too loose for one that rarely moves 1% — the same number means a completely different amount of risk depending on the ticker's own volatility.
Every forecast run produces thousands of simulated price paths, not a single point estimate. Both legs of the live trade plan (visible on any ticker page) are read off what those paths TOUCH, because that is the event an order fills on: a limit sell fills the moment price trades up through the rung, and a stop closes the moment price trades down through it. So TP1 / TP2 / TP3 are the 60th, 75th and 90th percentiles of how HIGH each path gets — levels that 40%, 25% and 10% of paths reach at some point — and the stop is the 25th percentile of how LOW each path gets, or a volatility-scaled floor (1.5σ over the forecast horizon), whichever sits closer to spot. Reading the rungs off where paths finish instead would understate how often each one is reached, and would leave reward and risk measured against two different events.
Reward-to-risk is reported against the conservative TP1 rung (the reward the ≥2:1 house risk gate checks), with the more optimistic TP2-based ratio shown separately and clearly labelled — never used to justify a trade on its own.
Beyond the static ladder, Quantustik runs a live paper-trading automation that advances a tracked signal's state as price moves: once TP1 prints, the position moves into a "trailing toward TP2" state, and if price then falls back to the original entry, the trade closes at break-even instead of riding a full round trip back into a loss. That is the mechanical version of "lock in the gain, don't give it all back" — a stop that moves up once the trade has already proven itself, rather than sitting fixed at the original entry-time level for the life of the trade.
A separate, explicitly volatility-proportional trailing-stop formula — the higher of a small fixed buffer below entry or price minus a multiple of the 20-day ATR — is used on the market-conditions overlay's own exit plan, illustrating the same "ATR-proportional, not fixed-%" principle applied to a portfolio-level position rather than a single ticker.
The core idea generalises beyond any one formula: a stop or a trailing distance expressed as a multiple of the instrument's own recent volatility (ATR, σ) automatically tightens for a calm stock and loosens for a choppy one, instead of forcing every ticker through the same fixed-percentage sieve. A fixed 5% stop is effectively tight for a high-beta name and effectively loose for a low-beta one — volatility-scaling removes that inconsistency without requiring a different manual rule per ticker.
None of this is presented as a tested source of edge on its own. It is an exit-discipline framework layered on top of the forecast and the position-sizing rules described elsewhere in the AI Lab — see the position-sizing article for how the same forecast distribution feeds the entry side of the same trade plan.
This exit methodology is explicitly untested for P&L publicly: there is no backtest showing the TP1/TP2/TP3 ladder or the break-even stop advance improves realised returns versus a simpler exit rule. The ladder levels come from the forecast's own simulated-outcome percentiles, so a mis-calibrated forecast distribution mis-places every rung in the ladder along with it — the exit plan is only as good as the distribution it is built from. Laddered exits also mechanically truncate the right tail: selling rungs early caps the occasional winner that would have kept running, and whether the smoother outcome is worth that cost is precisely the untested part.
Educational research only — not investment advice.