How rankings are calculated

Every PropPick ranking is produced by a fixed formula with no human discretion. No firm can buy a position or move one through a partnership. Below is that formula and its inputs, in full.

1. There is no single ranking

We don't publish a single “best prop firm” order. Each ranking answers a different question, and which question matters is yours to decide.

Ranking	Question it answers	Input data
By PropPick score	What do PropPick reviews add up to?	Formula in section 3
Most talked about	Which firms come up most in the communities?	Mention counts across each country's communities
Best rated	Do the communities speak well of them?	Share of positive mentions
Most payout talk	Which firms generate the most payout discussion?	Payout-topic posts in international communities + name mentions on Korean payout boards
Longest running	How long have they been operating?	Each firm's publicly stated founding year

2. There are only three inputs

(a) PropPick user reviews — reviews written directly on PropPick: an overall star rating, four sub-scores (rule clarity, support, usability, payout process), plus whether the payout was actually received and any dispute reports. The PropPick score is built from this data alone (section 3); payment outcomes and disputes feed the risk penalty subtracted from it (section 4). Firms with no reviews get no score, and hidden posts are left out of the tally.

(b) Public community posts, by country — public posts and comments gathered from the trader communities of each country. A thin sample turns a handful of voices into a ranking, so we collect as much as we can inside the window. We try to filter out promotion and off-topic noise: copy-pasted posts, bare lists of firm names and plain questions that carry no opinion are either dropped or marked neutral and left out of the ratio. Most talked about and Best rated are built on this data only , while the payout-talk ranking also uses (c) below. Current data is as of 2026-07-31 , covering the last 365 days. The sentiment of each post — positive, negative, neutral — and its topic (payouts, rules, support) are classified by manual review , and we show the source excerpt behind each classification together with a link to the original — so you can verify it yourself.

Community posts are the claims of their authors; PropPick does not verify or endorse them. In particular, “payout talk” is not proof of payment — it counts the posts where payouts were discussed, including both “got paid” reports and “never got paid” complaints. A high count means attention, not reliability. Automatic classification can be wrong, so if an excerpt seems to miss the point of the original, follow the link and read it. We do not scrape or repackage review data from other comparison sites.

(c) Mention counts on Korean payout boards — for the payout-talk ranking only, we also count how many posts on Korean payout-proof boards mention each firm's name. The current tally is payout reports on Naver cafés, covering everything up to 2026-07-29 — we scanned the titles of 826 posts and counted the 361 that named a firm. Because we only read titles, a post such as “third payout of July” that names no firm is never counted. The number is therefore a floor, and should be read as a rough order of magnitude rather than an exact comparison between firms. All we store is a count per firm. We do not collect or keep post titles, bodies, authors or links. Nothing is republished; only “how often is each firm mentioned” feeds the ranking. Frequent mentions mean many people use the firm, not that it pays out well.

3. The PropPick score formula

Only reviews left on PropPick count. Community figures can never be collected exhaustively, so they carry estimation, and they are other people's accounts that we cannot verify — we keep them out of a score that carries our name. Community tallies live in their own ranking tabs.

The five review inputs are rescaled to a 10-point range and combined with fixed weights.

Payout process  weight 0.30
Overall rating  weight 0.25
Rule clarity    weight 0.20
Support         weight 0.15
Usability       weight 0.10

base score = Σ(adjusted item ÷ 5 × 10 × weight) ÷ (sum of applied weights)
final score = base score − risk penalty

Payout process carries the largest weight. Choosing a prop firm ultimately comes down to one question: “does the money arrive?” Usability weighs least — a clunky platform still pays.

Sample-size correction (Bayesian shrinkage). To stop a perfect score from two reviews beating a 4.5 from fifty, small samples are pulled toward the overall mean. As reviews accumulate, a firm converges on its own average.

adjusted = (global mean × m + own mean × n) ÷ (m + n)

  n = number of responses for that firm
  m = shrinkage constant — overall rating 5, sub-scores 3, community 10
      (sub-scores are optional, so samples are thinner and shrinkage is kept light)
  global mean = not a benchmark we picked, but the mean of all data on this site

4. Risk penalties

Someone refused a payout leaves one star and moves on — and twenty five-star reviews bury it in the average. But “I never got paid” is the kind of fact that cancels out every other virtue. So it isn't averaged in; it is subtracted from the final score.

payout refusal   = 4.0 × √(unpaid cases ÷ (payment responses + 1))  max 4.0 pts
dispute reports  = Σ(reason weight × status multiplier)             max 4.0 pts

  reason weight     : unfair account suspension 1.6 · payout refused 1.6 · delayed payment 0.8
                      rule change without notice 0.8 · improper charge 0.6 · other 0.5
  status multiplier : filed / unresolved 1.0 · firm responded 0.7 · resolved 0.25

A resolved dispute is not reset to zero. That it happened at all is information. But a settled case and an abandoned one cannot weigh the same, hence the different multipliers. The +1 in the denominator keeps a single-case sample from taking the maximum penalty straight away. The square root on payout refusals exists because a straight proportion would turn “3 of 10 never got paid” into barely one point — the first unpaid case costs the most, and the curve flattens after that.

Missing data is not a zero. An unanswered item is dropped from the calculation rather than scored zero, and the remaining weights are renormalised — so a question nobody answered never turns into a bad review. A firm with no reviews at all gets no score.

Neutral posts are not counted. (Community sentiment rankings only.) The denominator of the positive share is positive + negative; plain questions and factual mentions carry no opinion and are left out.

Next to each firm's score you can see exactly which numbers were combined and how. The calculation lives in a single file, lib/score.ts.

5. Payout-trust index — until PropPick reviews accumulate

The ‘payout process’ ranking is meant to run on the sub-rating from PropPick reviews. With no reviews yet, that ranking sat completely empty. The first thing anyone asks about a prop firm is whether the money actually arrives, so leaving that slot blank means the page dodges the question. We built a temporary index from other people’s aggregates instead. Any firm that reaches 3 responses moves straight onto its own review value and ranks above this index.

Three kinds of data play three different roles. Public numbers from external sites produce the value; how much the community actually discusses payouts decides how much of that value we credit; and the source threads we have read and summarised adjust it up or down. Mention volume is deliberately not added to the score — being discussed a lot means a lot of people use the firm, not that payouts arrive. But an external score for a firm nobody has ever discussed is a score that has never been checked, so the less it is discussed the harder we pull it toward the overall average.

PropScorer payout-evidence / transparency score  weight 0.50
Trustpilot consumer review rating                weight 0.30
Prop Firm Match verified-purchase review rating  weight 0.20

exposure = log10(1+domestic mentions) + log10(1+overseas payout mentions)
shrunk = (overall average × 1.5 + observed × exposure) / (1.5 + exposure)
final = shrunk + source-tone adjustment (max ±1.5) − payout warning status 1.5

To appear in the ranking at all, payouts must have been discussed at least once in a Korean or an overseas community. With zero mentions the shrinkage simply returns the overall average, which made more than half the firms show the identical number — and a column of identical numbers is a list with a number attached, not a ranking. A firm nobody discusses is not a middling firm, it is a firm this ruler cannot measure, so it is left out and the count of excluded firms is shown above the ranking. Firms with payouts suspended get no index either — that is a job for the status badge, not for a ranking.

Source tone does not only subtract. Counting only negative threads would punish the firms we have read the most, since positive comments in the same thread would go uncounted. So we use the balance of positive minus negative, with slack in the denominator so a single thread cannot swing the full range.

6. The two PropPick picks on the home page

The home page carries two lists we compute ourselves. One answers the question you ask before buying, the other the question you ask after passing. The two disagreeing is normal, and the disagreement is the information — an account that is cheap to pass but awkward to withdraw from shows up in that gap.

Lowest expected cost to funding

Look only at the evaluation fee and the cheap account wins. But a high target with a tight drawdown means buying several times, so it ends up costing more. Only an expected cost divided by the pass probability measures rules and price on the same ruler. Since the assumed trader decides the result, we invent neither a win rate nor a risk-reward ratio and assume a trader with no edge at all (a random one-directional bet). Skill lifts every account's odds together, so the ordering barely moves.

Pass probability P
  End-of-day drawdown   P = D / (D + T)
  Intraday trailing     P = exp(-T / D)
    T profit target, D max drawdown. Both are independent of tick size and contract count.

Rule factors (rules that cannot enter the formula are docked by direction and size only)
  Tighter daily loss limit   up to 25% off
  Consistency rule present   15% off

Expected purchases = 1 / P
Expected cost = first payment + (expected purchases - 1) x retry unit + activation fee
  The retry unit is the cheaper of the reset fee and a fresh purchase.
  Monthly plans count one month — nobody knows the months before buying.
  Amounts use the promotional price currently in effect.

Best payout terms

You pass once, but you withdraw again and again. Six terms describe what happens to your money afterwards, and they fall into three kinds — how much you get (split), whether you get it at all (consistency rule, daily loss limit), and when you get it (first payout, cap, minimum). Do not underrate the middle pair. A consistency rule blocks money you have already earned, and a daily loss limit takes the account itself, at which point a 90% split is worth nothing. Split still carries the most weight for one reason: probability. It applies to every payout with certainty, while the other two only bite some traders. An unverified term is scored 0.5 rather than having its weight handed to the others, and an account needs four of the six verified to be ranked — not knowing something must never work in a firm's favour.

Profit split          0.28   50% → 0, 100% → full
Consistency rule      0.22   none → full, 50% → 0.7, 20% → 0
Daily loss limit      0.14   none → full, 4% of account → 0.7, 1% → 0
First payout          0.14   0 days → full, 10+ days → 0
Per-payout cap        0.12   uncapped → full, $3,000+ → full
Minimum payout        0.10   $0 → full,     qPayoutCode: ,000+ → 0

Unverified item       scored 0.5 (weight is never handed to the others)

Many firms step the split down (100:0 for six payouts, then 90:10). The summary uses the entry-stage value and the account page carries the original wording, because a summary must not hide that the split drops later. The daily loss limit is measured against account size, not as a raw amount: qNote: ,000 is 4% on a 25K account and 0.7% on a 150K one — the same number is breathing room in one case and a same-day exit in the other.

7. Minimum sample thresholds

Best rated: at least 5 positive + negative community mentions (sorted on the shrinkage-adjusted ratio)
Most talked about · Most payout talk: at least one relevant mention
Firms below the threshold are left out, and the number excluded is shown at the top of each ranking.

8. Partnerships and rankings are kept apart

Some outbound links are affiliate links and may earn us a commission if you make a purchase (Terms of Service, article 3). Whether a firm is a partner, at what rate, and whether it advertises enters none of the inputs above.

9. Disclaimer

Being listed or ranked on PropPick is not an endorsement, a certification, a guarantee of quality or a recommendation to use any particular firm. Regulation and firm policy in this industry change constantly, and past data guarantees nothing about future payouts or a firm's continued existence. Always confirm the terms on the firm's official site before paying; the final decision and its consequences are yours.

If you think the method is wrong, tell us. Whenever the formula changes, this page changes with it.