Where 14, 45 and 7 come from
Firms that publish pass rates cluster around 14% of evaluations passed. Of those funded, roughly 45% ever receive a payout — the rest breach, quit, or never clear the gates. Multiply them and about 7 in 100 buyers see money come back.
We use the industry figures only to frame the question. A firm's own score never uses them: it is computed from that firm's reviewed rulebook, its published payout evidence and its conduct record.
The trader every figure is run on
One deliberately marginal trader: a 38% win rate at 2:1, 4 trades a day, risking 5% of the drawdown per trade, paying commission and slippage on every trade.
Marginal on purpose. A great trader survives almost any rulebook and a bad one survives none, so neither tells you anything about the rules. Everything discriminating happens near the edge — which is exactly where most funded traders sit.
Reaching a target before hitting a floor is a gambler's-ruin problem; day-level rules are evaluated on the exact binomial distribution of a day's outcomes rather than an approximation. Change the trader in the live model at the foot of the page and every number follows.
The reference account
A $50,000 account with a $2,500 maximum drawdown, a $3,000 profit target, a $99 evaluation with a $49 reset, a 90% split, five minimum trading days and weekly payouts.
The advertised account size is simulated buying power, not deposited capital. The drawdown is the real budget, and it is the number every dollar rule is normalised against — which is why a "$50K account" with a $1,000 drawdown is a much smaller product than one with $2,500.
Why phases multiply against you
Each phase is another independent walk to a target without touching the floor. Two phases is not twice as hard, it is the square of one — and the trailing floor compounds it.
Expected net turns negative at two phases for this trader: the ticket, the expected resets and the time are worth more than the payout they buy. That is the arithmetic behind "ROI is the wrong question".
Why a moving floor is so much worse
A static floor sits still, so every dollar banked becomes cushion. An end-of-day trail follows your closing balance. An intraday trail follows the highest point you touched — profit you never actually held.
For this trader a never-locking intraday trail leaves 38% of the advertised $2,500 usable. The odds fall faster than the cushion, because ruin probability is convex in the cushion: halving it more than doubles the chance of dying.
This is why intraday accounts are the cheap ones. The discount is priced.
A cliff, not a dial
A limit that ends the day is a time-out and costs you almost nothing. A limit that fails the account is a second kill floor, re-armed every morning.
This trader's worst possible day is four losses — almost exactly $500. A fatal limit at $500 is therefore near-certain death across a 60-day evaluation; the identical rule at $1,000 can never be reached and never binds.
The number alone tells you nothing. What matters is where it sits relative to your worst realistic day, which is why we score the consequence and the size together.
Rules that charge you in time
Consistency cap. Your best day may not exceed a set share of total profit. A higher percentage is easier — readers invert this constantly. Under a 20% cap a single $760 day silently moves your finish line to $3,800 of profit.
Winning-day punch card. Most futures firms want five winning days, each above a dollar floor, and the card resets to zero after every payout. A $40 green day counts for nothing. A $150 floor changes nothing for this trader; $250 triples the wait.
Neither rule reduces your profit directly. Both keep you exposed to the floor for longer, which is the same thing.
What the fine print actually does
"100% split" is nearly always the first slice only. A per-cycle cap meters a good week out in instalments while the remainder stays exposed to the floor. A payout buffer is profit you must leave behind to keep trading.
And at many firms a withdrawal does not lower the floor — so taking all of your profit can execute your own account.
The retries are the real price: a $29 evaluation bought six times costs more than a $99 evaluation bought once, which is why cheap tickets do not automatically rank higher.
Simulated forever versus a live path
Simulated forever. Evaluations and "funded" accounts are both simulations. Revenue is your fees and resets; your payout is a cost. Every trader who improves is a trader the firm earns less from, so the rulebook is tuned to keep you close to getting paid without quite getting there.
A live path. The firm puts up real capital and splits real wins, so its revenue is a share of your profit. It earns when you do, and the rulebook tends to show it.
That is why a published live-capital path is a layer-1 signal that multiplies every score, and why the Best-for-traders lens weights it hardest. Prop firms are still the cheapest place in finance to learn — a $50 evaluation beats the thousands most people lose figuring it out live — but at a sim-only firm you are meant to stay a customer.
A lens is a weight set, not a different truth
Every lens scores the same signals from the same evidence and only changes the weights. Layer 1's weights are identical on every lens but one — Best for traders raises the live path and retroactive changes, because that lens asks whose side the firm is on.
Layer weights and risk-curve strength for each board lens| Lens | Firm | Rules | Money | Curve |
|---|
| PropRank | 35 | 35 | 30 | 0.6 |
|---|
| Best for traders | 45 | 40 | 15 | 0.8 |
|---|
| Easy payout | 25 | 35 | 40 | 0.6 |
|---|
| Max allocation | 30 | 50 | 20 | 0.6 |
|---|
| Lowest account cost | 25 | 20 | 55 | 0.6 |
|---|
| Lowest cost to payout | 25 | 25 | 50 | 0.6 |
|---|
| ROI | 25 | 20 | 55 | 0.6 |
|---|
Deterministic, capped, human-published
Nothing here is scored by a language model, a clock or a coin. The same reviewed evidence always produces the same board, and a human publishes every board.
An unknown signal sits at a low-neutral 40 and is listed by name, so an unresearched firm can never outrank a researched one by being vague.
Two findings skip the arithmetic entirely: a documented retroactive rule change caps a firm at 20 on every board, and an unresolved critical flag caps it at 25 (major flags at 45). Those were terminal signals in every collapse we have on record.
A methodology change requires a new algorithm version. This page describes proprank.brain.v1 and extraction schema proprank.rules.v4; historical boards keep their original version and as-of date, so a later formula cannot rewrite an earlier ranking.
What we refuse to do
Compensation is never an input. Partner status, commission rates, promo depth and sponsorship are excluded from every score and every ranking input.
Featured is a label, not a rank. Featured firms sit in their own marked strip, disclosed whenever a partner relationship exists, and never move up a board. No firm can buy its way to the top.
We do not rank by popularity. Reviews are read for direction, not level; a loud community is not evidence of a payout.
Creators keep their own codes. Creator mode shows a community their creator's codes inside PropRank and the iPhone app, and PropRank takes no cut of them. Rankings are never sold to a community through a creator either.
Retroactive rule changes
Has the firm rewritten rules and applied them backwards?
- What it does
- Reads approved tos_retroactive_change risk flags. An unresolved flag scores 0 and caps the whole composite at 20 on every framing; a resolved one scores 40; a researched firm with no such flag scores 100.
- Why it matters to you
- Retroactive application was a terminal signal in every documented collapse (FundingTicks, The Funded Trader, Fast Track Trading). Booked profit vanishing by email is the single most reliable death rattle in the industry, so it is treated as close to disqualifying.
- Where the curve bends
- Binary with memory. There is no partial credit while the flag is open; resolution restores less than half because the behaviour, once shown, predicts repeat.
- Where the data comes from
firm_risk_flags (approved) · flag_type = tos_retroactive_change- When we don't know
- A firm with no approved research at all has no basis for 100 and is imputed at the low-neutral 40.
Rule-change velocity
How often, and how hard, the rulebook moved in the last year.
- What it does
- Counts approved rule_change_events detected in the trailing 365 days, weighted minor 1 / notable 2 / major 4, with harder-direction changes weighted 1.5×. Score = 100 − 10 per weighted point.
- Why it matters to you
- A rulebook that moves monthly cannot be priced by the trader who bought it. Lucid rewrote rules six-plus times in year one; churn precedes most distress sequences and erodes the one thing a comparison site sells: the rules you read are the rules you get.
- Where the curve bends
- Linear and unforgiving: two major harder changes (12 points) already cut the signal to 0. A single minor easing costs 10.
- Where the data comes from
rule_change_events (approved / auto_approved) · severity, direction, detected_at- When we don't know
- No events on a researched firm scores 100; monitoring coverage is short for new firms, so pair this with operating history.
Regulatory standing
Named regulators vs. regulatory action against the firm.
- What it does
- Unresolved regulatory_action flag → 0; resolved → 50. Otherwise a firm that names a real regulator for its parent scores 100, and a researched firm naming none scores 70.
- Why it matters to you
- Almost nobody in this industry is regulated, so the scorer rewards relative posture (regulated parent, clean warning-list record) and punishes enforcement, rather than pretending a licence exists.
- Where the curve bends
- Step function: action is a cliff, a named regulator is a modest bonus.
- Where the data comes from
firm_risk_flags · regulatory_action; extracted facts · regulators[]- When we don't know
- Unresearched firms are imputed at 40.
Payout distress
Delay/denial patterns, lost processors or platforms, insolvency chatter, collapse.
- What it does
- Starts at 100 and deducts per approved distress flag by severity (critical 100, major 60, notable 30, minor 10; resolved flags cost half). Separately, any unresolved critical flag of any type caps the composite at 25 and any major flag at 45, on every framing.
- Why it matters to you
- Every 2024–2026 collapse looked generous on paper while it died. Slowing payouts come first, always; this signal is the solvency gate — generosity multiplied by distress, not summed with it.
- Where the curve bends
- Steep: one unresolved major flag removes 60 points here and caps the whole score regardless of how the other layers look.
- Where the data comes from
firm_risk_flags (approved) · payout_delay_pattern, payout_denial_pattern, processor_loss, platform_loss, insolvency_chatter, exit_scam_or_collapse, rebrand_of_failed_firm- When we don't know
- No flags on a researched firm scores 100; unresearched firms are imputed at 40.
Jurisdiction
Where the company actually sits.
- What it does
- Reads the headquarters country code from reviewed facts. United States 100; established legal jurisdictions (UK, Canada, Australia, EU members, Switzerland, Singapore, Japan) 75; the UAE and other identifiable jurisdictions 55; a mailbox jurisdiction or none stated 30.
- Why it matters to you
- A trader's only remedy against a firm is the firm's home legal system. US-based firms run on US-regulated market infrastructure and can be served; an anonymous offshore entity cannot.
- Where the curve bends
- Tiers, not a gradient. Moving a firm between tiers is an editorial decision with a quote behind it.
- Where the data comes from
extracted facts · headquarters_country_code- When we don't know
- Not stated on a researched firm scores 30 (opacity about domicile is itself a signal).
Named leadership
Is there a findable human accountable for the firm?
- What it does
- A named chief executive in reviewed facts scores 100; a researched firm with none scores 25.
- Why it matters to you
- Anonymous teams are the first line of the extraction profile. Public CEOs carry reputational cost when payouts stop; mailboxes do not.
- Where the curve bends
- Binary: there is no partial credit for a first name, a pseudonym, or a LinkedIn ghost.
- Where the data comes from
extracted facts · chief_executive- When we don't know
- Unresearched firms are imputed at 40.
Operating history
Years in operation.
- What it does
- 10 + 9 × years, capped at 95: one year 19, three years 37, five years 55, nine years 91, ten-plus years 95.
- Why it matters to you
- Most prop-firm deaths are firms under two years old. Longevity is the cheapest solvency proxy there is, and it gates rather than merely contributes on the Best-for-traders framing.
- Where the curve bends
- Linear to nine years, then flat; age cannot buy a firm past 95.
- Where the data comes from
prop_firms.founded_year (falls back to extracted facts)- When we don't know
- Unknown age sits at 40 — an unverifiable history is not worth the benefit of the doubt.
Live-capital path
Does the journey ever reach real money?
- What it does
- Any published plan whose funded stage trades live capital scores 100. A firm whose plans are all stated as simulated scores 20. Mixed or unstated sits at 40.
- Why it matters to you
- A firm that only ever makes money on simulated accounts profits when its traders fail; its rules will drift toward passing as few as possible and keeping the rest almost-winning. A firm with a live path profits when traders succeed, which aligns every rule after it.
- Where the curve bends
- Binary at the firm level; the Best-for-traders framing weights it as the heaviest universal signal.
- Where the data comes from
firm_account_plans.live_capital (published)- When we don't know
- Unstated is imputed at 40; the operator should push research to state it explicitly.
Payout transparency
Strongest evidence tier the firm's payouts can be verified at.
- What it does
- Processor-attested (on-chain, audited) 100; itemised firm disclosures 80; third-party trackers 60; marketing counters 35; social proof 15; anecdote 10; nothing approved 10.
- Why it matters to you
- Payout numbers are the most inflated data in the industry. The three largest claimed totals are the ones with zero on-chain visibility; opacity must never outrank verification.
- Where the curve bends
- Stepped by provenance tier; a bigger claimed number on a weaker tier never beats a smaller attested one.
- Where the data comes from
firm_payout_stats (approved) · provenance- When we don't know
- No approved observations scores 10, not neutral.
Community trajectory
Direction of member reviews, not their level.
- What it does
- Needs at least three published reviews. Centred on the recent-90-day average (3.0 stars = 50, each star ±20), shifted 15 points per star of drift versus earlier reviews and capped at ±15; a falling payout rating costs up to 15 more.
- Why it matters to you
- Star levels are gamed in both directions; a 4.2 falling fast is worse than a steady 3.8. Complaint velocity is the second-earliest collapse indicator after payout velocity.
- Where the curve bends
- Bounded: the level term moves 20 per star, the drift term is capped at ±15, and a falling payout rating can cost 15 more — a month of one-star reviews after years of five can take the signal to 0.
- Where the data comes from
firm_reviews (published) · overall_rating, payout_rating, created_at- When we don't know
- Fewer than three reviews is unknown and imputed at 40.
Denial discretion
How much room the terms reserve to refuse a payout.
- What it does
- Narrow, quotable denial language scores 55; none 100; broad 'at our sole discretion' / 'simulated — no obligation to pay' language scores 0 and is named as a deal-breaker.
- Why it matters to you
- Discretionary denial is the most weaponised clause family — TFT admitted ~10% of payouts denied pre-collapse; Fast Track cancelled already-approved payouts. Vagueness is measurable and predictive of disputes.
- Where the curve bends
- Three steps; there is deliberately no credit between narrow and broad.
- Where the data comes from
extracted facts · payout_denial_discretion (quoted from the terms)- When we don't know
- Unknown is imputed at 40.
Drawdown survival
Probability the reference trader survives the funded phase to a first payout.
- What it does
- Runs the survival model: the plan's drawdown becomes an effective cushion (static 100% of it, EOD trailing ~83%, balance-based ~74%, intraday trailing ~50%, a never-locking intraday trail ~38%), and gambler's-ruin math gives the chance of reaching the payout threshold before touching the floor. Scored as that probability × 100 on the firm's best plan.
- Why it matters to you
- Same trader, same trades: only the floor type decides whether the account dies. The trailing-lock axis is first-class because a trail that never locks makes eventual breach a near-certainty for any strategy with variance.
- Where the curve bends
- Convex in the cushion: halving the usable drawdown more than doubles the ruin probability for a marginal trader, while a strongly profitable trader barely notices. This is exactly the asymmetry that makes intraday accounts cheap.
- Where the data comes from
firm_account_plans · max_drawdown_usd, drawdown_type, trailing_lock, payout_buffer_usd, payout_consistency_rule_pct; trader model- When we don't know
- A plan without a stated drawdown cannot be simulated; the firm is imputed at 40. A plan whose drawdown TYPE is unknown is simulated at a 67% cushion — between end-of-day and intraday — and the unknown is listed.
Drawdown in honest trades
How many full-stop losses fit inside the drawdown at the smallest tradable size.
- What it does
- Risk per trade is the larger of the model's share of the drawdown and one micro contract at a normal stop (ticks cannot be split). Headroom = drawdown ÷ that risk; 30 consecutive losses of room scores 100, 10 scores 33.
- Why it matters to you
- A $1,000 daily limit is 500 MNQ points or 50 NQ points — the rule did not change, the contract did. Tiny drawdowns on big headline accounts force oversized trades, which is the quiet way cheap accounts become unwinnable.
- Where the curve bends
- Flat at the model's chosen risk until the micro-contract floor binds, then it falls linearly as the drawdown shrinks.
- Where the data comes from
firm_account_plans · max_drawdown_usd; reference instrument per market- When we don't know
- No drawdown stated → imputed 40.
Daily loss regime
The curfew: how many honest trades fit in a day before it ends.
- What it does
- No daily loss limit scores 100. With a stated limit, room = limit ÷ risk per trade; eight trades of room scores 100, two scores 25. A limit whose size is not published scores 45. A limit whose breach FAILS the account (hard_breach) is discounted to 70% and also enters the survival model as a per-day ruin hazard.
- Why it matters to you
- A tiny limit beside a big advertised drawdown makes the number you paid for unreachable in practice — one bad hour ends the day, and months of re-bought tests follow.
- Where the curve bends
- Linear in trades of room up to eight, then flat.
- Where the data comes from
firm_account_plans · daily_loss_limit_usd, daily_loss_consequence; extracted facts · daily_loss_limit- When we don't know
- Unknown → imputed 40. An unstated consequence is read as a lockout (the industry default), never as fatal.
Consistency gate
No-one-lucky-day: the best-day share of profit the firm allows.
- What it does
- Uses the payout-time percentage when stated, else the evaluation percentage, else the qualitative fact. Score = (pct − 15) × 2.4: 50% → 84, 40% → 60, 30% → 36, 20% → 12. No rule → 100.
- Why it matters to you
- A HIGHER percentage is EASIER — people invert this constantly. A $1,500 best day under 30% locks the first payout behind $5,000 of total profit; under 50% behind $3,000. Harsh rules applied only at payout punish exactly the big wins good trading produces.
- Where the curve bends
- Linear in the percentage; the trap end (20–30%) sits near zero on purpose.
- Where the data comes from
firm_account_plans · payout_consistency_rule_pct, consistency_rule_pct; extracted facts · consistency_rule- When we don't know
- Unknown → imputed 40.
Winning-day punch card
Qualifying days required before each payout, and the wait before the first.
- What it does
- Score = 100 − 9 × required days (0 days 100, 5 days 55, 10 days 10), blended 70/30 with the stated first-payout wait (100 − 1.5 × days).
- Why it matters to you
- Most futures firms will not pay until five winning days are on the card, each above a dollar floor — and the card resets to zero after every payout. It is a permanent treadmill between a trader and their own money.
- Where the curve bends
- Linear in days; every extra qualifying day costs nine points because the card resets each cycle.
- Where the data comes from
firm_account_plans · min_trading_days, winning_day_min_profit_usd, first_payout_after_days; extracted facts · minimum_trading_days- When we don't know
- Unknown → imputed 40. With no stated dollar floor, any green day punches the card; a stated floor lowers the chance a day qualifies and lengthens the expected path inside the survival model.
Payout fine print
Split, caps, cadence, the buffer trap.
- What it does
- Weighted blend: split (40% → 0, 90% → 100) 28%; cadence (daily 100, monthly ~0) 22%; buffer as a share of drawdown (a full drawdown of unwithdrawable profit → 15) 15%; per-cycle cap ($5k stops binding) 10%; first-payout wait 13%; payout-time consistency 12%. A live-capital funded stage lifts the blend.
- Why it matters to you
- '100%' is nearly always the first slice only; early caps meter your $4,000 week out in $2,000 instalments while the rest stays exposed to the kill floor; and at many firms withdrawing does not lower the floor, so pulling all your profit can execute your own account.
- Where the curve bends
- Each part is linear; the buffer term is the steepest because it is the least-known and most expensive rule.
- Where the data comes from
firm_account_plans · payout_split_pct, payout_buffer_usd, payout_cap_per_cycle_usd, first_payout_after_days, payout_consistency_rule_pct, live_capital; extracted facts · payout_cadence_days- When we don't know
- Parts that are unstated drop out and the rest renormalise; nothing stated → imputed 40.
Max allocation
How many machines one trader may play.
- What it does
- Log ladder on the published allocation ($100k ≈ 43, $400k ≈ 72, $1M ≈ 92, $2M+ 100) 50%; account count (30 + 7 per account) 25%; per-plan exposure (contracts or leverage) 25%.
- Why it matters to you
- Every rule runs per account, so the same edge copied across five accounts is five paydays. Experienced traders filter on this early — and a firm's generosity here interacts with its solvency (account stacking is what stressed Apex's treasury).
- Where the curve bends
- Logarithmic in dollars: the first $100k matters far more than the fifth.
- Where the data comes from
extracted facts · max_allocation_usd, max_accounts; firm_account_plans · max_contracts, max_position_notional_usd- When we don't know
- Unknown → imputed 40.
Conduct freedom
News, overnight, and automation permissions.
- What it does
- Average of three permissions: allowed 100, restricted 45, prohibited 0.
- Why it matters to you
- Money rules decide IF you get paid; conduct rules decide whether you still exist. Clear permissions stated before purchase are a rule-quality signal; bans applied after you win are the predatory version.
- Where the curve bends
- Linear average; one prohibited permission costs a third.
- Where the data comes from
extracted facts · news_trading, overnight_positions, automation_trading- When we don't know
- Unknown permissions score 50 each; all unknown → imputed 40.
Entry cost
Cheapest evaluation dollar per $1k of drawdown.
- What it does
- The drawdown is the trader's real account, so price is normalised by it: $0–$100 per $1k of drawdown maps 100 → 0. A $39 eval on a $2.5k drawdown scores ~84; $540 on $10k scores ~46. Price alone is the fallback when no plan states its drawdown.
- Why it matters to you
- The nominal account size is marketing. Comparing fees per dollar of usable drawdown is the only like-for-like price.
- Where the curve bends
- Linear in $/k of drawdown.
- Where the data comes from
firm_account_plans · eval_cost_usd, max_drawdown_usd- When we don't know
- No priced plan → imputed 40.
Cost to first payout
Expected spend before the first dollar comes back, retries included.
- What it does
- Survival model: evaluation fee × expected attempts (1 ÷ pass probability, resets priced at the reset fee when stated) + activation + monthly fees over the expected months. $0 → 100, $500 → 50, $1,000+ → 0.
- Why it matters to you
- FPFX data: the average customer spends ~$800 across ~3 attempts. A cheap ticket with a brutal floor costs more than an honest ticket, because retries are where the money goes.
- Where the curve bends
- Linear in expected dollars; the pass probability makes it convex in drawdown hostility.
- Where the data comes from
firm_account_plans · eval_cost_usd, reset_fee_usd, activation_fee_usd, monthly_cost_usd + survival model- When we don't know
- No simulable plan → imputed 40.
Payout probability
End-to-end chance the reference trader is ever paid.
- What it does
- Pass probability × first-payout probability on the firm's best plan, × 100.
- Why it matters to you
- Only ~7% of challenge buyers ever receive a payout. The funnel end-to-end is the trader-relevant number, and the rules decide it long before skill does.
- Where the curve bends
- Product of two ruin probabilities; hostile floors and heavy gates compound.
- Where the data comes from
survival model over published plans- When we don't know
- No simulable plan → imputed 40.
Time to first payout
Expected trading days from purchase to first cash.
- What it does
- Evaluation days (target ÷ daily expectancy) plus the longest of: profit required before withdrawal ÷ expectancy, qualifying days ÷ chance of a qualifying day, and the stated first-payout wait. 0 days → 100, 100+ days → 0.
- Why it matters to you
- Instant-funding with an intraday trail can still pay faster than a static account that needs five winning days — the Easy-payout framing exists to rank that trade-off honestly.
- Where the curve bends
- Linear in days.
- Where the data comes from
survival model; firm_account_plans · profit_target_usd, min_trading_days, first_payout_after_days, payout_buffer_usd- When we don't know
- No simulable plan → imputed 40.
Expected net return
Expected first-payout dollars minus expected spend, per dollar spent.
- What it does
- ROI = (P(pass) × P(payout) × first payout after split and caps − expected spend) ÷ expected spend. −100% → 0, break-even → 50, +100% → 82, +300% → 97.
- Why it matters to you
- This is the number the ROI framing ranks on: not usable drawdown per $100, but money back per money in once the rules have had their say.
- Where the curve bends
- Saturating above break-even so a single outlier plan cannot dominate the board.
- Where the data comes from
survival model- When we don't know
- No simulable plan → imputed 40.