How PropMason Scoring V3 Works
We score every plan a firm sells, not just its best one, so a single eye-catching offer can't carry a weak line-up. Reputation counts toward the score, and a bad enough rating over enough reviews now puts a ceiling on the whole score.
A term that applies to a plan and isn't published no longer drops out of the average: it scores low and keeps its full weight, so leaving details out costs a firm nearly every time. Nearly, not always - if a firm's real terms are worse than the cautious value we assume when we don't know, staying quiet still pays it. We publish that seam below rather than pretend we closed it.
What we also can't do is tell a firm's silence apart from our own crawler failing to read the page, so we take that low mark and stop there rather than adding a penalty on top of a gap that might be ours. Nobody can pay to rank higher.
What makes up a score
The method has ten dimensions, and nine of them carry weight in the score: how much room you have before you're stopped out (26%), how soon and how often the firm's published terms say you can withdraw (20%), how fair and clearly written the rules are (15%), what the challenge costs for the account you get (10%), how demanding the evaluation is (9%), how much of your profit you keep (8%), the firm's reputation from its review profile (7%), whether a public register still confirms the company you would be contracting with (3%), and how long the firm has been operating (2%). Those nine add up to 100. The tenth is data confidence, and it is weighted zero on purpose: how much of a firm we managed to verify is not a measure of how good that firm is, so we report it separately instead of letting it move the number. When one of those applies to a plan and the firm hasn't published it, we score that part at 40 out of 100 — 30 for the rules — and it still counts for its full weight, with one exception: a firm with no founding year on file is simply not scored on track record, so a missing year neither helps nor hurts it. A firm that publishes half its terms is divided by all of them, not just the half we could read. That mark is low on purpose but it is not a zero, because a missing term can be our failure to read a page as easily as the firm's decision not to publish it. Either way, holding something back lowers a score in almost every case - but not every case, and we would rather publish the gap than claim it away. Where a firm's real terms are worse than the cautious value we assume when we don't know, staying quiet can still pay it. We could also flatten the bands so no disclosed value ever scores below what we assume, but that would stop the harshest terms ranking against one another: every limit below the threshold would score alike, and you could no longer see which of two tight limits is the tighter. Things that genuinely don't apply to a plan stay out of the maths entirely.
| Dimension | Weight | Scope | What It Evaluates | Public Method Note |
|---|---|---|---|---|
| Profit Split | 8% | Offer level | The share of eligible profits a trader can retain, including whether advertised maximums are supported by the offer terms. | Clear, attainable terms score more reliably than headline-only or conditional claims. |
| Drawdown Risk | 26% | Offer level | Maximum and daily loss limits, in percentage or dollar terms, and how much practical trading room they leave you. Whether the drawdown is static or trailing makes up 20% of this dimension. A plan whose model we have not read from the firm's own page takes the low baseline for that part, like any other unstated term. | Risk structures are compared in context. An undisclosed daily loss limit no longer scores better than a disclosed one: an input the firm does not state keeps its full weight and is priced at the low baseline, instead of being set aside or filled in favourably on its behalf. Where the firm publishes none of this at all, the whole dimension takes that baseline. |
| Cost & Value | 10% | Offer level | What the offer costs relative to its account size. We score the list price anyone pays without a code: during a sale we score the struck-through price, not the sale price. | A monthly fee counts as two months, because the subscription renews each month until you pass or cancel; a one-time fee counts once. When a firm does not say what period its fee covers, we score the fee as published. A one-time activation fee charged when you pass is added once. A firm that says it charges none adds nothing, and if we have not found a firm's activation fee we score the fee alone. Codes, affiliate links and discounts that come with a condition never change this score. |
| Evaluation Structure | 9% | Offer level | The path to a funded stage, including evaluation steps, time constraints and minimum trading-day requirements. | Simpler is not always universally better; the score reflects the burden created by the full rule set. |
| Payout | 20% | Offer level | First-payout timing, recurring payout access and material withdrawal conditions stated in current firm documentation. | Published policy is scored separately from unverified testimonials or promotional language. A payout term we infer for a plan instead of reading it there, such as a firm-wide figure applied to each plan, can score at most 7.5 / 10 on this part, or 6.0 / 10 when the inference is less certain. |
| Rule Clarity | 15% | Offer level | How clearly the offer explains trading permissions, restrictions and rules that can affect eligibility or payout. | Explicit, internally consistent rules reduce uncertainty. A rule a firm never wrote down scores 30 out of 100 here, the lowest baseline we apply to anything, because an unwritten rule is the one most likely to cost you a payout. |
| Reputation | 7% | Firm level | Firm status and supported external reputation signals, with attention to source quality and review volume. | A rating alone is not sufficient; review volume and the integrity of the source matter too. A bad rating backed by enough reviews does more than cost a firm this 7% - it caps the whole score. See the ceilings below. |
| Counterparty Standing | 3% | Firm level | Whether you can name the legal entity you would be contracting with, and whether a public register still says that entity is there. Most prop firms are not regulated entities, and that is ordinary for this market rather than a scandal, so this is not a licence test: a regulator's register entry adds a small credit on top and its absence costs a firm nothing. | This is the only weighted input a firm cannot write itself - everything else here is its own pricing page or a review average it can influence. Being identifiable carries three quarters of the dimension and the supervision credit a quarter. A register we could not reach scores below one we could, but well above a firm that never named an entity, because a closed register is our failure and not the firm's. A company a register shows as struck off scores the same as one we could not identify at all: we have no evidence that either of those is the worse harm. |
| Track Record | 2% | Firm level | How long the company you would be paying has been operating, read from the founding year it declares. | A bounded ladder: ten or more years scores 100, five years 85, three years 70, one year 55, under one year 35. This is the same curve firm age already rode inside reputation - it now scores on its own two points, which is where reputation's old 9% went. A firm whose founding year we do not hold is not scored on this part at all: it leaves the average rather than taking the low baseline, because an absent year is not evidence of a young firm. The year is the firm's own declaration and we say so - we do not verify it against an incorporation register. |
What caps a score
The percentages above are how a firm earns its score. The limits below are the highest score it can hold, however well it does on everything else. Where more than one applies, the lowest one is the one you see. We only apply a limit over something we can verify about the firm itself. A page we couldn't load is our problem, not evidence against the firm.
| What happened | Highest possible score | Why it matters to you |
|---|---|---|
| The firm is flagged as a scam or has shut down | 0.0 / 10 | There is nothing to compare if you can't safely put money in. |
| We are actively looking into concerns about the firm | 4.0 / 10 | Good trading terms don't help if there's an open question about whether you'll be paid. |
| Trustpilot found that the firm's reviews were manipulated | No cap: a badge | We score the firm on its terms like any other and put a 'Fake reviews found' badge beside its score wherever the score appears, linked to Trustpilot's finding. The badge stays as long as the finding is on Trustpilot, even after a rating comes back. While Trustpilot has the rating withdrawn, reputation shows as not read and counts at the same low value as any other part we cannot read. Any other Trustpilot guideline breach that is not only about display gets a neutral 'Trustpilot flagged this firm' badge instead, also with no cap. A finding only about how Trustpilot content was displayed gets no badge. |
| A poor Trustpilot rating, over enough reviews to be sure of it | 2.5 to 5.5 / 10 | Terms on a web page are a promise. Thousands of former customers describing what actually happened are a record. Good paperwork should not outrank that. |
| One of the plans on sale has payout terms that score below 3.0 / 10 | 6.0 / 10 | You might buy that plan. An average across the others shouldn't hide it from you. The rule also covers drawdown terms below 3.0 / 10, but no plan can currently score that low on drawdown, so today only payout terms set it off. |
| The weakest plan we have full details on is far worse than the rest | 6.0 to 7.0 / 10 | A firm's best plan shouldn't paper over a much worse one sold right next to it. |
The reputation ceiling is graded, and both halves have to be true before it applies - a poor rating, and enough reviews to be confident the rating is real. At 2.0 stars or below across at least 200 reviews the score cannot exceed 2.5 / 10. At 2.5 or below across 200 reviews, 3.5 / 10. At 3.0 or below across 500 reviews, 4.5 / 10. At 3.5 or below across 500 reviews, 5.5 / 10. A firm with no rating, or with fewer reviews than the threshold, is never capped by this: we do not know enough about it, and not knowing is not the same as knowing something bad. It is a ceiling rather than a bigger share of the score because reputation carries 7%. Raising that weight would nudge every firm slightly and still let a badly regarded firm price and structure its way back up the table. A ceiling touches only the firms with a sustained bad record, and it cannot be traded away against a cheap challenge or a generous split. Where a confirmed-misconduct limit also applies (a scam or closure flag, or an open investigation), the two are worked out separately and the lower of them is the one that binds, so being two bad things at once can never score better than being one. What the misconduct finding does outrank is the wording: we describe such a firm by the more serious finding against it, even when the number you see came from this reputation rule.
A Trustpilot rating is never taken at face value. Before it is scored, it is pulled toward 4.1 stars, as if the firm had 250 extra reviews at that rating. A firm with thousands of reviews barely moves. A young firm with a handful of reviews is neither lifted by a few perfect scores nor sunk by one angry customer.
The reputation score is capped at 5.5 / 10 when the review record is too thin or too old to lean on: fewer than 100 reviews, or review data older than 180 days as of 2026-08-23. Such a firm can still earn a fair reputation score, only not a top one. A firm whose review date we do not have is not treated as old.
We tell you how much we know about a firm, and we charge it once. A term that applies to a plan and isn't published scores 40 out of 100 and keeps its full weight, so the score falls in proportion to how much is missing. We do not then add a separate ceiling for a firm being hard to read, because we cannot currently tell whether a term is missing because the firm withholds it or because our crawler was blocked, the page was down, or we failed to parse it. Charging twice for a gap that may well be ours would be unfair to the firm and misleading to you. Where a firm publishes nothing at all for one of the eight (every weighted part except track record, which drops out of the average when no founding year is on file), that whole dimension is scored rather than skipped, so silence across a dimension cannot pay.
Sources, Freshness and Corrections
V3 prioritizes official firm pages and dated canonical observations, then uses supported public platform data where relevant. Each scored value carries provenance and confidence metadata. Material conflicts are reviewed, and unsafe values are withheld rather than replaced with a misleading zero.
How sure we are isn't how good a firm is
A number we're less certain about doesn't make a firm worse. It means we checked and couldn't confirm it. When a firm publishes a term, it counts toward that part of the score. When it doesn't, that part scores 40 out of 100 and still carries its full weight - low, but not a zero, because our data cannot currently tell a firm that withholds a term from a page our crawler was blocked from or that was down when we looked. We can see that a value is missing. We cannot see why. Until we can, a gap costs a firm that one low mark and nothing more. Terms that don't apply to a plan are left out, and so is track record when no founding year is on file - an absent year is not evidence of a young firm. Some links here earn us a commission, and we disclose them, but no firm can pay for a rank, a score, or a tie-break. If a score or a link looks wrong to you, email support@propmason.com with the URL and we'll correct it.
Where firms without a score go
Not every firm has a score. Most often that is because it has closed, or because we cannot yet score a single one of its current plans to the standard this page describes. Those firms are not ranked, and they are not quietly dropped in among the ranked ones either.
An unscored firm sits below every scored firm, in both directions. Sorting the table from lowest score upward does not float them to the top - a firm we could not score has not beaten anyone. Its cell reads an em dash, never 0.0, because we have not judged it badly; we have not judged it at all. Among themselves, unscored firms are ordered by name, using a fixed character order rather than your language settings, so the same list comes back in the same order every time you load it and reads the same for every reader. In the default ranking you land on, scored firms that tie on the score itself are separated by reputation first, then rule clarity, then that same name order. Each firm still gets its own position, so firms with the same score take consecutive positions in that tie-break order and carry a Tied score label. If you re-sort a table yourself by clicking a column, that column decides the order and ties fall straight through to the name order, without the reputation and rule-clarity steps. Whether a firm pays us is not a tie-breaker at any level of any of those lists.
From offer evidence to a public firm score
V3 is the public scoring authority. It scores eligible active offers, balances results across programs and publishes only launch-ready firm results that match the active canonical dataset.
Offer-level scoring
Each active offer is evaluated on the terms that apply to that offer. One attractive tier cannot stand in for a firm's entire portfolio.
Program and firm aggregation
Related offers are grouped into programs, then Overall uses the balanced average across the full portfolio. A single showcase offer cannot boost the firm score.
Segment lenses
Overall, Beginner, Faster Payout Terms, Low Risk, High Capital, Instant, Futures and CFD lenses answer different comparison needs. A lens can be unavailable when no suitable offer exists.
Evidence-aware output
Confirmed misconduct, a punishing plan, a badly inconsistent line-up and a poor reputation across a large review base can each put a ceiling on Overall. None of them invent a score for a term we could not verify.
Structured data follows the same evidence policy: ranges and descriptive PropertyValue output are used when terms vary by offer, while unsafe or unsupported values are omitted.
Quality and Integrity Gates
These gates describe when a score may be limited, withheld or sent for review. They are evidence controls, not a public list of gameable penalty coefficients.
| Condition | Public Treatment | Review Level |
|---|---|---|
| A serious integrity concern is supported by reviewable evidence | Cap the public score at 0.0 / 10 and require manual review | manual review |
| Material evidence is unresolved or under review | Limit publication until the conflict is resolved | evidence gate |
| The firm or relevant offer is inactive | Exclude inactive offers and suppress stale ranking claims | evidence gate |
| An external reputation source reports an integrity issue | Exclude or reduce reliance on the affected signal | evidence gate |
| Key terms are missing, contradictory or marketing-only | Score the missing part low and keep its full weight; no separate cap, because the gap may be ours | evidence gate |
| Current terms are clear and supported by canonical evidence | Allow the verified evidence to participate normally | evidence gate |
Scoring Change Log
- ->A Trustpilot finding that a firm's reviews were manipulated no longer caps its score. Seven firms were held at 2.5 / 10 by that rule. They are now scored on their terms like any other firm, and a 'Fake reviews found' badge sits beside the score wherever it appears, linked to Trustpilot's finding.
- ->While Trustpilot has a rating withdrawn, reputation scores the same low value as any other part we cannot read, with no extra cap on top.
- ->Each firm page and the comparison tool now show the score of every weighted part. A part we could not read on any plan shows as not read, not as a number.
- ->This page now says what actually sets off the punishing-plan limit, that the drawdown model counts for 20% of drawdown risk, and how payout terms we infer instead of read are capped.
- ->A firm's score now reflects every plan it sells, evenly. Its single best offer no longer lifts the total.
- ->Reputation and track record count toward the score, and a poor rating over a large number of reviews now caps the whole score instead of only costing a firm the weight reputation carries. The graded thresholds are published above.
- ->A term that applies to a plan and isn't published now scores 40 out of 100 - 30 for rule clarity - and keeps its full weight in the average. It used to drop out of the maths entirely, which meant publishing less could raise a score. That is fixed wherever a firm's real terms score above the baseline, which is almost everywhere. Where they score below it, staying quiet can still pay, and we say so rather than flatten the bands and lose the ability to rank the harshest terms against one another. Within drawdown risk, an individual unstated limit now takes the baseline too; the other dimensions still assume a favourable value for some unstated fields, and that is pending work.
- ->An undisclosed daily loss limit no longer scores better than a disclosed one. Where a firm states nothing, we no longer supply a favourable number on its behalf.
- ->We still cap a score only over something we can verify about the firm. A page that wouldn't load for us is not evidence against it, and we cannot yet tell that case apart from a firm choosing not to publish - so we say so rather than pretending otherwise.
- ->Added limits for punishing plans and for line-ups where one plan is far worse than the rest. Firms with no score sort below every scored firm in both directions, then by name. When scores tie, we order by reputation, then rule clarity, then by that same name order. Who pays us never enters into it.
- ->Added a ninth dimension: whether a public register still confirms the company you would be contracting with. Its three points come out of reputation, which falls from 12% to 9%, and none out of the terms we read from firm pages. It is the only weighted input a firm cannot author about itself.
- ->Added a tenth dimension: track record, how long the firm has been operating, now scores on its own two points instead of hiding inside reputation at a fifth of its weight. Reputation falls from 9% to 7% and is now the review profile alone. The curve is unchanged - a decade scores 100, five years 85, three 70, one 55, under one 35. A firm with no founding year on file is not scored on this part at all: it leaves the average rather than taking the low baseline, because an absent year is not evidence of a young firm.
- ->Two brands belonging to one company now share one position in the ranking instead of holding two. Both are still named and both keep their own score, because the products differ and only the counterparty was ever counted twice. Firms that have announced a deal the ownership register does not yet reflect stay separate, and say what the register does not show.
- ->When a firm advertises "up to X%" profit split, we now score that number as a ceiling: it is the most you can reach, not a promised rate. Where the firm's own wording gives a guaranteed base plus a maximum, the base is what scores.
- ->Weekend holding now sits on the same ladder as the other permission rules: allowed scores 90, restricted - for example allowed only if you buy an add-on - scores 70, and banned scores below that. Until now anything short of a plain yes was read as a ban.
- ->We had the consistency rule backwards. The percentage is the most one day's profit may make of the total you withdraw against, so a higher allowed share is a looser rule, not a stricter one: a 50% cap treats you better than a 30% cap. A firm that plainly publishes no consistency rule now also scores better than one that says nothing.
- ->Adds conservative treatment for applicable terms without canonical evidence, so missing details cannot improve an offer or program score.
- ->Preserves public last-known-good V3 scores while all 23 replacement scores are staged, verified and activated as one cohort.
- ->Keeps not-applicable terms out of the denominator and does not expose crawler internals as a public confidence metric.
- ->Separates seven quality dimensions from scoring coverage and evidence reliability.
- ->Not-applicable fields leave the coverage denominator; missing applicable fields remain visible without automatically making all evidence low confidence.
- ->Commercial and affiliate relationships do not affect scores or sorting. Firms with the same score take consecutive positions in tie-break order.
- ->Moved from firm-level inputs to eligible offer, program and firm aggregation.
- ->Added evidence-based publication gates and market-aware comparison lenses.
- ->Became the sole public score authority with no V2.3 fallback.
- ->Improved score inputs with tier account sizes and stored rule relationships.
- ->Refined drawdown treatment and persisted refreshes from the crawler and webhook pipeline.
- ->Added integrity gates for serious reviewable risks and material data concerns.
- ->Introduced grouped sub-scores to make firm comparisons easier to read.
- ->Removed criteria that could not be sourced and verified reliably.
- ->Rebalanced scoring around verified reputation, profit split, drawdown and futures context.
- ->Introduced the crawl-backed nine-metric scoring model for published firms.
- ->Standardized comparison of offer terms, risk, payout and account-size data.
- ->Established the first public scoring baseline for published prop firms.
- ->Compared available offer terms, risk and payout information with the evidence available at the time.
Who collects these figures, how articles are checked before publication, and how corrections are handled is set out in our editorial and data policy.