LucidDaily?
Don't buy it.

Ten pricing-card screenshots plus eight support articles. Everything below is confirmed unless marked ⚠ — the 100K tier and all four of its coupon prices are assumed, and 100K has no LucidPro comparator at all, so no 100K number can ever carry a Daily-vs-PRO claim.

Payouts and the trail — the two rules that decide everything

Note the pricing inversion: the harsher rules are the cheaper card. Intraday eval drawdown is $15–$45 cheaper than EOD; DLL ON is $15–$50 cheaper than DLL OFF. That discount is exactly what the grid says not to take.

Both products are walked over the same calendar windows, on the same instrument, with the same method family, so the delta is dominated by rules rather than market luck. To be precise about it: this is paired at the cohort / market-window level, not a byte-identical trade list — each arm carries its own re-optimized sizing, and sizing shifts the 2R day-stop, so the two books differ by a handful of trades (at 25K, 2 trades in the primary window). Pre-registered decision rule, set before the run: a 95% CI that straddles zero means "don't replace", full stop. Realistic framing = our actual position (3 funded PRO 25Ks, 2 free slots). Greenfield = a clean start, shown for context only.

Every framing, every horizon

The one positive cell is mean-driven, not median-driven (median only +$2,678), rests on 18 heavily overlapping cohorts, has a negative 10th percentile, and is the same cell whose funded arm carries 43.75% death by 90 days and 69.23% by 180 days. Primary-window 18m cells are n=1 and are excluded from the verdict entirely. Treasury arms are a no-op — median annual fee burn is $342 / $693 / $781 ⚠ / $1,621 against a $6,000/yr budget, so the budget never binds and reinvesting payouts has nothing to fund.

And the "daily payouts" headline doesn't even buy faster cash

At 6 months the gap is starker still: median first cash day 160 / 137 / 141 / 115 for Daily vs 85 / 78 / 81 for PRO. The intraday trail plus the buffer-at-start+MLL+$100 floor means a Daily account has to climb further before a single $500 request is even legal — and it dies more often on the way.

The corrected metric is the ratchet premium: over the pre-lock phase only, every rise in the floor caused by a new unrealized high-water mark, minus the part later re-earned by realized P&L — measured against a counterfactual floor that ratchets on realized balance alone.

At every tier ratchet_gross == ratchet_capture, so the floor's climb to its permanent lock is fully re-earned by realized P&L and the median funded premium is exactly zero; the worst case is a third to a half of one MLL, not nine of them. EOD-trail SKUs have no unrealized channel into the high-water mark at all, so their premium is zero by construction — the correct answer, not a missing measurement. No chosen cell and no paired delta moved when the metric was corrected.

So what does explain a $7.9k–$9.0k gap?

Three structural facts, none of them a per-trade tax.

The transient pre-lock premium and the tighter basis are the enforcers of point 3, not a separate ongoing cost. Compounded across a slot-capped conveyor, that is the $7.9k–$9.0k.

The unrealized ratchet is still real and still causal: re-run with the high-water mark ratcheting on realized equity only and the intraday eval SKU gets 20–30% cheaper at every tier, while all 26 EOD cells are unchanged on every reported metric. It is the size of its cost that was wrong, not its existence.

Pass rate is 100.0% in all 104 primary-window eval cells, so cost per slot is purely attempts × price. The answer is the same at every tier: EOD drawdown + DLL ON, at the lowest taper base.

Three things worth internalising. (1) The cheaper card is the expensive one — intraday drawdown costs 33–51% more per funded slot despite a lower sticker. (2) DLL ON is free money at these bases — it saves 17–19% per slot and cost exactly 2 lockout days across 24–25 attempts (n=2, a measurement, not a law); the lockout starts biting at roughly 2× the chosen base. (3) Never transplant PRO's taper base — 50K Daily wants $800, not PRO's $1,600, and PRO's base is both dearer and slower. The eval is only 4.0–7.3% of a slot's yearly value, so this is a rounding error — but it's free to get right.

Why EOD wins, restated after the correction. The draft justified this pick with a heuristic — "intraday wins when the trail tax is ≤ 10–15% of MLL" — fed the retracted 189–269% figure. On the corrected metric the intraday cards sit at 7.8–11.3% of MLL, i.e. inside the band where the heuristic would pick intraday. The pick doesn't change, because it isn't made on the heuristic: it's made on directly measured cost per funded slot ($120.40 vs $160.59 at 25K; $162.85 vs $223.73 at 50K), driven by a median 2 attempts on intraday versus 1 on EOD. Measurement beats heuristic — and that makes this a pricing verdict, not a trail verdict.

If the launch coupon lapses, this flips at 25K and 50K. At list prices the EOD+DLL card loses to EOD + DLL-OFF ($595.80 vs $193 at 25K; $656.50 vs $261 at 50K) — their EOD+DLL list prices are a flagged pre-sale baseline. 100K and 150K do not flip. The full-price campaign arm makes Daily worse overall: 25K −7,891 → −10,140, 50K −8,974 → −12,486.

Size at 13–15% of MLL — $150 / $300 / $400 — not at PRO's 30–40%. Transplanting PRO's funded risk onto Daily's intraday trail kills ~15 of 16 cohorts inside 90 days at both small tiers and earns a quarter to a third less. Note also that the ladder is not monotonic: 25K r200 is strictly dominated by r150 — lower EV and four times the death rate. Do not interpolate it.

The structural finding: withdrawal policy cannot buy survival

Nineteen withdrawal policies were swept. At 25K r150 all nineteen produce exactly 12.50% death-by-90-days — the only variation anywhere in the policy dimension is the 180-day figure, and that's one cohort moving one 6.25pp step. Same at 100K r400. Same at 50K r300 except weekly. Meanwhile policy moves EV by up to 2.7×.

This is a rule fact, not a fluke. The trail locks permanently at start+$100 once the balance exceeds the Initial Trail Balance — and that threshold equals the payout buffer. So payout eligibility requires the account to already be past the lock. Before first eligibility no policy can differ, because there's nothing to withdraw. After the lock, a max sweep always leaves exactly one MLL of headroom and the floor never moves again — withdrawals cannot lower the floor. At sizings where deaths happen before first payout, policy is irrelevant to survival: at 25K r150 all three deaths never took a payout at all.

Two corollaries worth keeping. sweep0 == promimic to the cent — imposing a PRO-style $500 cycle cadence on a Daily account costs literally nothing, because the $500 minimum is already the binding cadence constraint; Daily's rules are what let cash out early, not the policy. And deep-threshold policies push median first cash from day 81 to day 267 at 25K while dropping "gets paid at all" from 81.25% to 43.75% — a way to convert cash into nothing.

Where policy does matter — hot sizing at 150K. There deaths are mostly post-payout, so retained cushion is real protection: skim25 buys −12.5pp of 90-day death for −5.8% of PV3, but median extracted cash falls 30%, $9,477 → $6,619. This is the one genuinely close call in the study — if the objective is median cash rather than the death guardrail, sweep0 is the 150K pick, not skim25.

The gate was pre-registered: a "Daily beats PRO" claim must survive the rule matrix, fill realism, degraded edge and winners-half. Since the headline is already negative, the question becomes whether any stress flips it into a defensible Daily win. It does not.

CIs, in order: slippage [−8,514, −4,336] / [−9,414, −3,738] / [−2,433, +10,084] · degraded [−1,588, −1,106] / [−2,823, −1,080] / [−6,906, −6,225] · winners-half [+402, +572] / [+169, +908] / [−1,441, +308] · full price [−11,564, −8,766] / [−15,464, −10,130] / [−13,258, +9,615] · PRO DLL 1200 150K [−12,502, +9,958].

* Precision fix. The draft called the two HWM arms "bit-identical"; Codex refuted that as over-claimed. Corrected: the arms are identical on every headline metric — PV3, net-cash medians and quartiles, all death rates — for every chosen cell, but 5–9 of the 16 cohort detail rows differ (individual death dates and lock dates move). No selection flips; the row still reads HOLD. Only the strength of the wording changes.

The one sign flip is degenerate — read it before quoting it

Both arms make zero payouts and both are net-negative. Daily "wins" by exactly the difference in fee burn — it is a cheaper way to lose. Under the same stress, Daily's 150K eval pass rate is 0.0% in every window and the funded cells show near-total death (15–16 of 16 cohorts). This must not be quoted as a flip in Daily's favour.

One honest note against our own side: the slippage stress moved opposite to the prior. We expected fill optimism to flatter Daily, so slippage should have widened PRO's win. It narrowed the gap at all three tiers — PRO's cycle-goal structure loses more to slippage than Daily's continuous sweep. It doesn't rescue Daily (no cell reaches significance), but the known-bias argument can't be used to inflate PRO's margin either.

Before publishing, both reviewers demanded checks constructed to cut toward Daily — the point being to see whether a negative verdict survives evidence selected against it. Two things moved: one claim was retracted, and one cell's basis changed. No selection flipped, and 25K/50K never moved off "PRO wins" in any of the 18 verification cells.

The finding that isn't about Lucid: our edge is about one tick thick

The ladder exists to prove PRO wasn't shielded from the haircut. It wasn't — and what it exposed is a shared execution knife-edge that neither product causes and neither escapes. At one tick of adverse slippage per side, standalone funded net-cash medians collapse on both sides:

PRO's medians go to zero at both small tiers; Daily's fall 74–81% while its accounts die 3–4× as often. PRO's eval time also doubles-to-triples by ×0.70 (median days to pass 14 → 30 at 25K, 19 → 41 at 50K, 22 → 54 at 150K). Whichever firm we trade, the fleet's edge is about one tick per trade thick. That is the largest risk on this page — and it is ours, not Lucid's. It is also why Daily's campaign delta narrows under slippage (−7,891 → −6,419 at 25K): PRO's cycle-goal structure loses more cash to slippage, while Daily's accounts are simultaneously dying three to four times as often. Cash gap narrows, survival collapses.

The 150K flip — and why it means "not measurable", not "Daily wins"

Exactly one cell flipped anywhere in the round: 150K at edge ×0.90, mean Δ +$11,633 (CI [+5,736, +17,941], P(Δ≤0) = 0.00, positive 10th percentile) — the only non-straddling positive produced anywhere. It is stated here rather than buried, and it is not evidence for Daily.

The response reverses direction twice — −827 → +11,633 → +3,958 → −3,816. Real economic effects don't do that. The per-cohort book shows why: shrinking every trade by 10% takes the worst Daily cohort, 2025-04-01, from $1,342 to $27,607 — not a Daily edge appearing, but a floor breach not happening because a 10%-smaller drawdown missed the trail by a tick. Corroborating: daily150k deaths fall 93.75% → 50.0% under that haircut while its cash median falls ($6,619 → $5,138). The price axis misbehaves the same way — making Daily 30% cheaper moves the delta +$8,012 and making it 30% dearer moves it +$1,890, both the same direction, which no fee effect can do.

Add the identification problem — seven monthly cohorts inside a 19-month window overlap almost completely, so the bootstrap treats ~1–2 independent draws as 7 — and the conclusion is not "Daily wins at 150K" or "Daily loses at 150K". It is: the 150K comparison is not identified with this data. The recommendation is unchanged — don't buy — but its basis is "not measurable", not "measured negative". 25K and 50K are untouched: at ×0.90 they move to −$8,631 and −$9,999, further against Daily.

Everything above is a selection. This is the whole file: every cell the study ran, unfiltered, with the picks marked but nothing hidden. Click a column header to sort (click again to flip). Every number here is read straight out of the study artifacts by lucid_daily_site_grid.py — nothing on this page is hand-typed, and the source SHA-256s are listed at the bottom of the section.