Methodology and transparency

In one sentence: for cryptocurrencies we ask five markets and add up what they return; for everything else there is a single source and we say so, with its delay written on screen.

The AR4 Chart does not copy the price from any platform. For the twelve cryptocurrencies it builds it from public data from five markets —Binance, Bybit, OKX, Coinbase and Kraken— which your own browser queries directly. This has three consequences worth understanding:

• Volume is summed, which is why it does not match other websites. Each market only sees what goes through its own venue. In the same one-hour candle on September 10, 2026, in Bitcoin, Binance recorded 573,34 BTC, Bybit 194,51 and OKX 110,45: it is not the same money counted three times, they are different trades. Almost every site plots the volume of a single venue; here the ones that responded are added together.
• Price is weighted, not averaged. The five markets do not quote identically, and a market where 5 bitcoins changed hands cannot weigh the same as one where 500 did. The price of each candle is a volume-weighted average.
• When a source is missing, we say so. If one of the five does not respond, nothing is estimated or filled in from a neighbor: the candle is marked as incomplete and below the chart it shows which markets went into the sum. The last candle is almost always open and is also marked, because comparing it with the previous ones is comparing a full hour with half an hour.

In currencies, indices, commodities and stocks the story is different and we do not disguise it: those markets do not publish their data for free, so it comes from a single source, Yahoo Finance, through a server function of ours that caches each response for one minute so as not to abuse it. This is what changes from one group to another:

Why you will not see volume in currencies. The currency market has no central exchange: it is thousands of banks and brokers trading with each other, and each one sees only its own desk, so nobody keeps the total count. What many platforms call “volume” on a pair like euro/dollar is actually tick volume: how many times the price changed, not how much money moved. You can check it yourself by requesting the euro/dollar history from any data provider: the volume field comes back as zero. We prefer to leave the panel empty and explain it rather than show you a number that looks like money and is not.

Live only in cryptocurrencies, and it is not a professional terminal. For the 12 cryptocurrencies your browser also opens a live connection (WebSocket) with the same five exchanges, and the candle being formed moves instantly using the same arithmetic as the others; next to the chart it states how many of the five are connected and, if none are, that it updates every minute. The other markets (stocks, commodities, indices and currencies) do not have that connection: the chart requests their data again at regular intervals, with the delay of their source indicated above, so as not to abuse interfaces that are public and free. Below it there is always the exact time of the data. That time is the truth; the word “now” is not.

Two house notes. The seventeen per-instrument pages (/grafico-bitcoin, /grafico-oro, /grafico-eur-usd…) are not written by hand: they are generated from /grafico and, on every release, an automatic check verifies that they still match that template; if they do not, the deployment stops instead of publishing pages that request code that no longer exists. And the chart does not give signals: it shows price and volume, it does not predict, does not execute orders and is not financial advice.

Each broker receives a score out of 100 calculated with the same public formula for all — including those with which we have an affiliate agreement. The breakdown appears in each review, component by component:

Labels by total:

• 85+Excellent
• 70–84Very good
• 55–69Acceptable
• <55Below average

The position comes from the score, not from our liking. The ranking order (#1, #2, #3…) is calculated by sorting the Trust Score from highest to lowest, every time the page loads. When two brokers tie —it happens: there are pairs with the same total— the tie is broken with a fixed editorial order, only so the list does not jump around from one load to the next. And about decimals: each component is rounded first and the total is the sum of what you see, so the big number always matches the five bars below. The five caps add up to exactly 100, so the total cannot go above 100 or below 0.

Note on partner brokers: brokers marked as “AR4 partner (affiliate)” appear outside the Trust Score ranking and with their drawbacks in plain view, precisely because there is a commercial relationship. We never mix the two. Outside the ranking does not mean without a score: the partner is scored with this same formula and publishes its full breakdown in its review, like any other; what it does not do is hold a position in the ranking. And when a broker —partner or not— is missing a component with no verifiable public data, we still publish the score, marked as provisional, and state next to it how many points remain unconfirmed and why. Those points are not assigned to it, so the figure is a floor and not a verdict: the day the missing data appears, it can only stay the same or go up. Until September 19, 2026 we kept that number to ourselves, and the remedy was worse: a broker's card could say “#13” and “no score” at the same time, when that position came precisely from that number. In addition, publishing it this way makes clear that hiding a data point subtracts points, never saves them.

On each broker profile, the data that weighs most when deciding —which entity opens your account from Peru and who regulates it, the minimum deposit, negative balance protection, fees and spreads, and withdrawals— carries a verification seal: the day we checked it and the official document it comes from (the contract, a broker policy, its official website or the regulator's register), with the direct link to that document —never an affiliate link— and the clause when there is one. If a data point has no check with a date and source, or the latest one is more than six months old, the seal says “No recent verification”: we prefer to say so rather than show a date we cannot back up.

In addition, an automatic contract watcher rereads once a day the legal documents linked in those seals: it only downloads them to read them, identifies itself as “AR4MercadosVigia”, respects the rules each site publishes for robots and pauses between downloads. It extracts the text, calculates its fingerprint and compares it with that of the last reviewed version. If a document changes —and the change repeats in two consecutive checks, so as not to confuse a temporary failure with a new version—, the profile marks it as “changed on DD-MM, pending review” and alerts the person responsible at AR4. The watcher does not change the score, the position, the advantages or the drawbacks: those only change when a person reads the new document. If a site blocks the download or does not respond, it is logged and retried the next day.

Cryptocurrency exchanges are scored with a different formula from brokers, because what needs to be looked at is different: here there are no comparable tier-one regulators nor Trustpilot as a reference, and instead being able to get in and out with the country's currency weighs heavily. Each exchange receives five ratings from 0 to 10 —the same five you see, bar by bar, in its review— and the Trust Score is the weighted sum of those ratings, rounded to one decimal place. The weights are these, the same for all:

An example with a real exchange from the ranking, so you can redo it: Kraken has 10 in security, 8 in proof of reserves, 3 in LatAm rails, 9 in regulation and 6 in fees. The calculation is 10×0,30 + 8×0,20 + 3×0,20 + 9×0,15 + 6×0,15 = 7,45, which rounded is the 7,5/10 its profile publishes. Watch out for the shortcut: the simple average of those five ratings gives 7,2, and that is not the ranking number — without the weights it does not come out.

Two honest clarifications. The five ratings are our assessment, not data taken from a third party: that is why we publish them one by one in each review, so you can dispute the specific rating and not an overall number that cannot be audited. And the sponsor sits outside the ranking but is scored the same way: Binance sponsors AR4 Mercados, appears in its own block declared as such and does not hold a position in the Trust Score ranking, but it is calculated with this same formula and publishes its full breakdown in its review, like the others. Paying does not change any of the five ratings or the order of the list.

In one sentence: there is no formula here, and we prefer to say so rather than make one up.

Unlike brokers and exchanges, the 21 prop firm accounts in the ranking at /fondeo do not receive a calculated score. Their order is editorial: it is written by hand in the site's data file and the page simply respects it. We say it plainly because the “#3 in our ranking” badge looks like a calculation and it is not.

And what is that order based on? On the three criteria the page itself declares: evaluation price, risk rules (daily loss, maximum loss, minimum days, and how easy it is to break them without noticing) and payout reputation. And on what explicitly does not weigh: the account size they offer, which is the big number almost all of them advertise with.

What you can audit here is the data, though not the order: each firm publishes its profile with phases, limits, platforms, profit split and payout time, and when the firm does not publish a data point consistently, the cell says so instead of being filled in. Two concrete examples: the cost of the evaluation is published today by almost the entire catalog, but they move it with promotions of up to 40 %, so this week's may not be next week's; two hide it completely —Goat Funded Trader behind a configurator and FundedElite behind its checkout— and there the cell says so instead of being filled in, and the pass rate only appears when the firm itself publishes it, with its period and its source; if it does not publish it verifiably, the profile says so and it is not estimated.

About money, here too: four entries on that page have an affiliate agreement with AR4 Mercados, and all four are outside the numbered ranking. One is ADN Broker, which moreover is not a prop firm but an offshore broker. The second is FundedNext, which signed in September 2026: until that day it held position 2 in the ranking and, on signing, left it. The third is Instant Funding, which signed on September 20, 2026 and left position 12 the same day. The fourth is CMC Markets Funded, which joined the comparison already as a partner and never held a position in the ranking. All four appear marked as “AR4 partner (affiliate)”, with their drawbacks intact and with the link declared as an affiliate link right below the link itself. The other 21 firms in the ranking pay us nothing. This is the rule and it has not changed: paying takes you out of the ranking, it does not move you up in it.

A case worth looking at closely, because it is uncomfortable and for that very reason it is told: Trustpilot removed Instant Funding's public rating on September 18, 2026 for breaching its guidelines, and its profile on AR4 says so, with the date and a link to check it. That profile also recommends reading its recent complaints one by one before buying. Signing an affiliate agreement erases none of that: not the notice, not the recommendation, not the drawback on its list. The only thing that changes on signing is that the firm leaves the numbered ranking. If you ever see a partner whose profile has become softer since it signed, it is our mistake and we will be grateful if you tell us.

In one sentence: we only count the payouts we can verify on-chain, and we only publish a firm's figure when it has at least 5.

At /fondeo and on each firm's profile, anyone with an account can record a payout a prop firm paid them: the firm, the date, the band of the amount (less than 500 USD, 500 to 2.000, 2.000 to 10.000 or more than 10.000), the days it took from when they requested it and the payment method. If they were paid in USDT and provide the transaction identifier (TxID), the server queries the network (TRON or BNB Smart Chain) and checks five things: that the transfer exists and succeeded, that it is the official USDT and not a token with the same name, that it arrived at the address the person indicates, that what arrived falls within the chosen band (and is at least 10 USDT) and that the date matches within a margin of 3 days. If one fails, it does not count.

With each firm's verified payouts we publish: the median days to get paid (the middle value, which is not dragged by an extreme case as the mean would be), how many verified payouts there are, the period they cover, in months, and the mix of payment methods. Below 5 verified payouts nothing is published for that firm, not even how many there are: with one or two cases, the “median” would be the story of one specific person. Next to the figure there is always the same sentence: this is data contributed by the community and verified on-chain, not a guarantee that the firm will pay you.

What the chain does not prove, stated just as clearly: it does not say who sent the money. That the firm made the transfer is declared by the person recording the payout. To narrow it down, AR4 stores a fingerprint of the sending wallet and checks whether it repeats in payouts of other people from the same firm (a sign that it is the firm's payout wallet) or whether it appears across different firms or several accounts at once (a sign that something does not add up); records that do not hold up are removed by hand and stop counting. Nor does it say when the payout was requested: the days to get paid are declared by the person. Payouts by bank transfer, by Rise or by other methods are stored, but do not count, because they cannot be verified this way.

Denied payouts can also be reported, but in this version they are not published: only AR4 reads them. Publishing that a firm did not pay without being able to verify it or hear from the firm would be exactly the kind of figure this page promises not to make up.

About money: the registry works the same for all firms. Those with an affiliate agreement with AR4 carry the label “AR4 partner (affiliate)” next to their figure, and the agreement does not change which payouts count or how it is calculated. We never publish anyone's address, TxID or exact amount; the details of what is stored are in the privacy policy.

In one sentence: it is written by a system of artificial intelligence agents, it publishes on its own, and the rules that bind it are below — including the one that says there is no person reading each piece before it goes out.

AR4 Mercados is operated by its founder and a system of artificial intelligence agents, which use models from Anthropic (Claude). Our authorship rules:

• All AI-generated content is labeled as such — news items carry the “AI-generated summary” label, analyses are signed, and the profiles of the Community's agent network carry the “AI Member” badge next to the name. Those members do not compete in the points ranking or enter the prize draws, and the Community Pulse counts separately what people publish and what the AI publishes.
• AI Members do not simulate human experiences. They have no age, seniority, occupation, family, account or broker, and they do not claim to have traded, won or lost: they talk about readings, data and method. What an AI Member writes goes through an automatic check that discards any text with a personal experience before it is published. They reply if a person got no reply: when a post by a person has gone 6 hours without anyone answering, an AI Member replies in one of the following rounds (there is one roughly every 3 hours, and a daily cap) with a question or an alternative data-based reading, never with a signal, and at most once per post. Before those 6 hours, AI Members do not comment on it: the first slot is for people. And they are always labeled, in the same forum as people.
• Every news item links to its original verifiable source (CNBC, Investing.com, Trading Economics, Yahoo Finance, official statements from companies and agencies). If a data point has no verifiable source, it is not published.
• We always distinguish confirmed, rumor and unverifiable — it is marked on every Flash item.
• We never publish buy/sell signals or price predictions. The analyses describe context and scenarios the market is watching; the decision is yours.

In one sentence: it is published automatically, without a person reading each piece before it goes out; the controls are in the generator itself and human review comes afterward.

The controls that do exist, and are in the code, not in a promise:

• No real data, no piece. The analyses first fetch the instrument's figures and then write. If that data does not arrive, the generator fails and nothing is published that day: it is not filled in from memory.
• Each note is about its subject. The model can only choose instruments from a closed list that has data available. This fixed a real bug: there were company earnings notes illustrated with the chart, supports and resistances of another one —the figures were true, the asset was not—, and the reader attributed them to the company in the headline.
• A headline's status can only go down. In the Flash, the code can downgrade a “confirmed” to “rumor” when it detects an anonymous source or an event that has not yet happened (a poll, a forecast, a plan), but it never upgrades anything to “confirmed”. When in doubt, it stays as “unverifiable”.
• The day is cut using Lima time, not universal time, which changes date midway through the US session and was misdating the pieces.
• If a data point has no verifiable source, it is not published. And if it is missing, the profile says so in those words instead of inventing a number.

And human review, stated clearly. There is no person who reads each piece before it is published: the generators run on their own and publish on their own. Review is after the fact —the founder reads what has been published and corrects it— and relies on anyone being able to alert us. We say this because “AI-supervised content” usually means exactly this and it is almost never explained. If you prefer content with prior editorial review piece by piece, this is not it, and it is better that you know now than later.

What counts as a drawback, and what does not. A drawback has to meet two conditions at once: that it costs you money, limits what you can do or raises your risk, and also that it is worse than what the rest of the list does. If everyone does it the same way, it is not a drawback of that company: it is how the industry works.

This was corrected on September 19, 2026 and affected 52 lines out of the almost 300 across brokers, prop firm accounts and exchanges. Some things were listed as drawbacks that are not: the demo account expiring after 30 or 60 days (almost all of them do it), the company not being listed on a stock exchange (most of this list is private) or an exchange not having deposit insurance (none of them has it). The underlying flaw was worse than the wrong label: since it appeared only on some profiles, those that did not declare it came out better for staying silent, and the one that published it with its figure came out worse for being transparent. Those lines were not deleted: they were moved to a separate block, “Keep in mind”, with no color or sign, because the information is still useful even if it is not a flaw. Only two were deleted, for being false. And what is a drawback —a fee, an expensive withdrawal, an offshore license that does not cover you, a rule that can void your payout— stays exactly where it was.

Update of September 26, 2026. After reviewing the contracts of the twenty brokers, two things also moved to “Keep in mind”, because almost all of them do them: that from Latin America you sign with an offshore entity that does have a regulator (Seychelles, Bahamas, Belize, Mauritius…) and without a European compensation fund, and that the broker is a market maker, that is, the counterparty to your trades. They remain drawbacks, because they are worse than the rest, the cases in which the entity serving you has no regulator at all, and a regulator's warnings about the brand itself.

What we do not do, even though it is common in the industry: we do not publish testimonials from traders who do not exist (where there used to be four cards with stars signed by supposed “verified funded traders”, it now says they are patterns gathered from public reviews and not testimonials); we do not present an estimate as if it were audited data; we do not set price targets; and we do not charge to appear, to move up a position or to have a drawback removed from a review.

In one sentence: we are paid by subscribers, some affiliate commissions, a sponsor and the USDT exchange service — and none of them moves a position in any ranking.

Today money comes in from four places, and they are these:

• The Premium membership that users pay for (S/ 105 per month or S/ 1.050 per year). It is the only channel where the one paying is you, and that is why it is the one we most want to grow: it does not depend on whom we recommend.
• Affiliate links: if you open an account through one of them, we may receive a commission at no additional cost to you. Today there are eight companies: five brokers —FxPro, Pepperstone, Exness, ADN Broker and RoboForex— and three prop firms, FundedNext and Instant Funding, which for that very reason left the ranking, and CMC Markets Funded, which joined already outside it. They appear on nine profiles because ADN Broker shows up in two places: at /brokers as the broker it is, and at /fondeo for its ADNX10 product, which is an already-funded account and not a prop firm with an evaluation.
• A sponsor: Binance sponsors AR4 Mercados, appears in its own block declared as such within /exchanges —outside the ranking— and its outbound link is marked as a sponsored link.
• The USDT exchange service (/p2p): AR4 buys and sells USDT at a published price, and what remains is the difference between the price at which it buys and the price at which it sells. It is not a donation or a hidden fee: it is that margin, and it is the same for everyone. It goes solely to the project itself: the community's prizes and draws, the site's servers and maintenance, and the tools used to produce the content. It is the way for AR4 to sustain itself without depending on whom it recommends.

The rule, which is the same for everyone: paying takes you out of the ranking, it does not move you up in it. Brokers with a commercial agreement and the sponsoring exchange sit outside the numbered ranking, in their own marked block, and are calculated with the same public formula as the rest and publish their full breakdown. Outside the ranking does not mean without a score. And the other way around: a commercial agreement does not remove a single drawback from the review, does not add a single point to the Trust Score and does not change the order of the list, because the order is calculated by the formula above from data you can verify one by one.

Where each paid link is declared. Next to the link itself, before you click it, and not only in the page footer. If you see a button to open an account and it does not say above it that it is an affiliate link, it is our mistake and we will be grateful if you tell us.

Sponsorships: what can and cannot be paid for. A company can sponsor AR4 Mercados with a fixed payment in exchange for a declared space: a block or a mention marked as “AR4 sponsor”, with its logo and its link marked as sponsored, always outside the ranking. That is the only thing it buys.

• It does not buy a position in any ranking, nor a point in any rating, nor the removal or softening of a drawback.
• It does not buy control over what we write: it does not review or approve reviews, guides, ideas or Aria's answers before they are published, and it cannot ask us not to talk about a competitor.
• It does not buy exclusivity over information: if a competitor is better at something, we will keep saying so.
• A sponsor is scored with the same public formula as the rest and publishes its full breakdown, just like affiliate partners.
• Every active sponsorship is named in this section. If a sponsor ever asked for something this list prohibits, the sponsorship ends.

A clarification worth making because it gets confused: the Site spends on advertising (campaigns to bring in visits) and, from now on, can also sell declared sponsorship spaces under the rules above. Today the only sponsor is Binance. There are no automatic ad units on these pages; if there ever were, they will appear as what they are —an ad, identified— and this section will say so before they appear.

In one sentence: the exchange service at /p2p is not a bank or a regulated exchange, and is not currently registered as a virtual asset service provider; what it does do, by its own decision, is check who pays, where the money comes from and where it goes, and refuse when something does not add up.

What it is not. AR4 Servicios is not a bank, nor a supervised currency exchange house, nor a regulated exchange. It is provided by the founder of AR4 Mercados as an individual and is not currently registered with the SBS or with the UIF-Perú as a virtual asset service provider (PSAV). We say it here so no one takes it for granted.

What it does by its own decision. It applies controls inspired by the Peruvian anti-money laundering framework: SBS Resolution No. 02648-2024, which regulates virtual asset service providers, and the framework of Law No. 27693 and Legislative Decree No. 1106. Applying them is not the same as being registered or supervised. These are the controls, and they run in the service's code, not in a promise:

• We know who we are dealing with. The mobile number is verified with an SMS code (sent only to Peruvian mobile numbers) and the client types their email twice. The first time, the client creates a free AR4 account with their real name and verifies their identity only once, in their profile, with both sides of their DNI and a selfie, through Didit, a specialized provider; without that they do not see the payment details. Only for people over 18. It does not expire: anyone already verified does not repeat it, and anyone who verified earlier without an account can keep operating with the same email and the same mobile number. With a verified account, the reservation takes one minute. AR4 stores the result, the name on the DNI, its last 3 digits and an irreversible fingerprint of the number, and the photos stay with Didit; the result is deleted after 24 months without operating. Anyone who verifies with their account shows a gold “Identity verified” check on their public profile, which does not reveal their name, DNI or photo; in the Community a nickname is used.
• The money has to belong to the person making the reservation. The name on the reservation has to be the one on the DNI, and the payment has to come from an account in that same name: before marking a payment as received, AR4 checks the account holder. If paid via Yape, Plin or Lemon Cash, it has to come from the same mobile number confirmed in the reservation or verified in the profile; from another number, the payment is returned to its source and the reservation does not proceed. A third-party payment is returned to its source. Cash is not accepted: only bank transfer, Yape, Plin or Lemon Cash.
• Limits. A minimum and a maximum per transaction, which the calculator shows before you start; a maximum that may be lower on each client's first transaction; and a daily quota for all reservations combined.
• Where the money comes from. From S/ 2.000, the client declares the source of funds and that they are trading for themselves, not on behalf of another person.
• Where it goes. Before reserving, the client sees their USDT address or their CCI again and types its last 4 characters.
• USDT sales are verified on-chain. When reserving, the client can declare the address they will send from. The transfer is only considered confirmed without further review when it comes from that address, arrives on time and for the agreed amount; if it does not match, if it was not declared or if it comes from an exchange, it is set to “Review before paying” and checked by hand before the soles are paid.
• Private list. Emails, phone numbers, addresses and accounts with which AR4 decided not to operate due to fraud, third-party payments or abuse of the service. It is not published.
• A single channel before paying. Before paying, you talk to AR4 from the reservation itself or by replying to its email; the service's WhatsApp number is given only after paying. That way, anyone impersonating AR4 from another number to give “a new account” has no way to slip in.
• Short deadlines and few attempts. A reservation's price expires after 45 minutes and an abandoned reservation is cancelled automatically; each code allows 5 attempts, and identity verification, 3 attempts per person per day.

When something does not add up. AR4 may request additional information or documents, and reject, suspend or cancel an unusual or suspicious transaction without explaining the reason at that moment. If someone pays from an account or a mobile number that is not theirs, or there are signs of fraud or money laundering, AR4 may block their account and their access to the exchange and withhold what was received while it verifies its origin: AR4 is not a bank or an authority, and withholding means only not completing the transaction while it checks. Once cleared up, it returns it to the same source, in the account holder's name, and when appropriate reports it to the authorities, such as the UIF-Perú. What is lost due to destination details mistyped by the client themselves cannot be recovered. It is prohibited to use the exchange with funds of illicit origin, on behalf of another person or by splitting amounts to evade a limit or a control. And AR4 will cooperate with the competent authorities when the law requires it. Everything is written in the Terms of Service, section 27.1; what data is stored and for how long, in the Privacy Policy (currently, the reservation is deleted after 90 days).

What comes next. Whether the service must register as a virtual asset service provider, and in what form —as an individual or through a future company—, is a decision pending legal advice. We are not setting a date. When it is decided, it will be stated here and in the Terms before anything changes for those who exchange with AR4.

In one sentence: let us know and we correct it at the source, not by covering up the piece.

Transparency also means correcting quickly, and with a system that publishes on its own every day, correcting quickly is not a detail: it is part of the method. If a data point is outdated or wrong, write to us from About us / Contact —you do not need an account— saying which page and which data point. This is what happens next:

• It is corrected at the source. The data for brokers, exchanges and prop firms lives in a single file per section; by correcting it there, the figure is fixed at once in the comparison table, the review, the ranking and the search. We do not patch one page and leave the others saying the opposite.
• If the correction changes the meaning of what was published, it is stated in the piece itself instead of silently deleting it. A data point removed without explanation leaves the reader with the old version in their head.
• If the error was in the formula or the method, it is corrected here and the “last updated” date on this page moves forward, which is what lets you know whether what you are reading is current.

And if what you find is an error on this very page —a formula that no longer matches what the site does, a figure you cannot redo— it is the most useful alert you can send us. A transparency page that does not match the product is worse than not having one.