Best Forex Brokers for Scalping

Scalping is defined by holding time and target size rather than by any particular indicator. A scalper is in the market for seconds to a few minutes and is trying to capture a small number of pips, which changes the arithmetic of everything else. On a five pip target, an all-in cost of 0.8 pips consumes roughly a sixth of the gross move before the trade has done anything wrong. That is why scalpers gravitate to raw-spread accounts with an explicit commission, and the full cost comparison - converting commissions into pip equivalents, comparing structures like for like and modelling what a month of trading actually bills - is the subject of our low-spread broker comparison rather than this page.

What that arithmetic does not capture is the thing that decides whether a scalping strategy survives contact with a live account: whether the price you clicked is the price you get. Every element between the click and the fill can move it - your own reaction time, the terminal, the network hop to the broker's trade server, the broker's order handling and risk checks, and the liquidity provider's response. The scale is easy to underrate. At roughly ten units of the quote currency per pip per standard lot, a $7 round-turn commission is worth 0.7 pips; one pip of adverse slippage on the entry and another on the exit costs nearly three times that. A broker quoting a marginally wider spread with cleaner fills can be genuinely cheaper in practice than the tightest number on a comparison table.

It is also worth being precise about what retail scalping is not. It is not high-frequency trading: the latency budget is milliseconds rather than microseconds, and the returns to shaving further flatten quickly once your platform is hosted near the broker's server. The failures that actually kill retail scalping strategies are more mundane - an order rejected because the stop was closer to price than the server allows, a requote that costs the move you were trying to catch, a position that cannot be closed during the fast few seconds when you most want out, and a clause in the client agreement that nobody read. The sections below work through each of those in turn.

Retail forex execution falls into two broad shapes. In a principal or market-making model the broker is your counterparty: it quotes its own prices, takes the other side of your trade and manages the resulting risk on its own book. This is legal, disclosed and not inherently improper, but the firm's own position can sit opposite yours, which is why the model attracts scrutiny. In an agency, STP or ECN model the broker routes orders to external liquidity providers and earns a commission rather than a marked-up spread. IC Markets is described in our broker dataset as offering institutional-grade ECN/STP execution; for every other firm on this page our dataset records no execution-model claim at all, which is itself worth knowing. In practice most retail firms operate some hybrid, internalising part of the flow and hedging the rest, so the label on a marketing page tells you far less than the order-execution policy document does.

The mechanism most relevant to scalpers, and the least discussed, is last look. In wholesale FX a liquidity provider may be given a brief window after receiving an order in which to accept or reject it, during which it checks whether the price has moved. That window is where rejections and requotes on fast-moving quotes come from, and it falls hardest on strategies that click precisely at the moment price moves - which is the definition of scalping. The FX Global Code, the industry's voluntary good-practice standard, addresses last look and asks participants that use it to be transparent about how. Our dataset records nothing about whether any broker on this page uses last look, so the broker's own execution policy is the document that has to answer the question for your account. If it does not, ask.

MetaTrader exposes the consequence directly through execution mode, which the broker sets per symbol. Under Instant Execution you either get the price you requested or you get a requote. Under Market Execution there is no requote: you get the next available price, with slippage possible in either direction. For most scalping approaches Market Execution is the better of the two, because a requote costs you exactly the move you were trying to capture, whereas slippage at least puts you in the trade. The trade-off is that you have accepted an unknown entry price, which is why position sizing for a scalper should assume the fill lands slightly worse than the click. Where to read the mode differs by platform: MetaTrader 5 lists an Execution field in the symbol Specification window, whereas on MetaTrader 4 you infer it from the order dialog, which offers a maximum-deviation setting on Instant Execution symbols and not on Market Execution ones.

Slippage is the difference between the price you asked for and the price you got, and it is not inherently a bad thing - it should occur in both directions. The single most informative test you can run on a broker is symmetry. If your history shows fills worse than requested when price moves against you but never better when it moves in your favour, that asymmetry is a finding, and it is a much stronger signal than any advertised execution-speed figure. Positive slippage should appear in a large enough sample. Its absence is worth asking about.

The data to run that test is already in your account. MetaTrader 5 keeps orders and deals as separate records - the order carries the price you requested, the deal carries the price at which it actually executed - so the difference is directly computable per trade. On MetaTrader 4 the same comparison has to be reconstructed from the terminal Journal, which records the request and the price the order was opened at. Export a few hundred round turns rather than a few dozen, because slippage distributions are noisy and a small sample tells you about one week rather than one broker. Then split the results by session and separate out any entries placed within a couple of minutes of a scheduled release, because those will otherwise dominate the average and make an ordinary broker look erratic.

Size introduces a second effect that is easy to confuse with poor execution. As your order grows it consumes more of the available depth, so the average fill price drifts away from the top of the book - a mechanical consequence of liquidity rather than misconduct. Below a certain size this is invisible; above it, it becomes the largest single component of your cost. A depth-of-market ladder with a volume-weighted average price mode is the most direct way to see where that threshold sits for you, and cTrader - which our dataset records as available at IC Markets, Pepperstone and FxPro - is the platform on this page most commonly used for it. Our dataset records which brokers offer cTrader, not what any broker enables within it, so confirm the ladder is switched on for your account. Be clear what a retail ladder represents in any case: spot forex has no central order book, so the depth shown is your broker's aggregated feed, not the whole market.

Log rejections and requotes separately from slippage, because they have different causes and different fixes. A rejection may be a liquidity provider declining the order, a server-side constraint such as minimum stop distance, an insufficient-margin condition, or a volume the broker does not accept on that symbol. A requote is specific to Instant Execution. Both are far more informative when timestamped, because a cluster at the same minute each day points at something structural such as the rollover window rather than at anything about your strategy.

The constraint that ends more tight-stop strategies than any spread ever has is the stops level - a band of prices around the current market, set in points by the broker per symbol, inside which the trade server will not accept a stop loss, a take profit or a pending order. The scalping-specific problem is one of scale: where a broker sets that band above zero, it can be wider than your entire pip target, which makes the strategy unplaceable on that instrument no matter how attractive the pricing looks. It is not negotiable, it is not a platform setting you can override, and the broker can change it. Check it on the exact instruments you intend to scalp, on the exact account type, before you fund anything - our guide to MT5 brokers explains where the field sits in the platform.

Its companion, the freeze level, is the one that bites during the move rather than before it. Inside that distance from the current price an existing order or position cannot be modified or closed at all, which removes your ability to act at precisely the moment price is running through your level. A strategy that depends on manually cutting a trade the instant it goes wrong is relying on an ability the server may withdraw exactly when it is needed. Read the two together rather than separately: a workable stops level sitting alongside a wide freeze level is a worse combination for a scalper than either number suggests on its own, and neither appears anywhere in a spread comparison.

Several other server-side limits reject orders that look perfectly reasonable on the chart, and a scalper hits them more often simply by placing more orders. Minimum volume, volume step and maximum volume per order are set per symbol by the broker, as is any cap on aggregate exposure in one direction or on the number of pending orders an account may hold. An automated strategy that computes a size the broker will not accept has its orders refused, usually with an error that reads like a platform fault. The practical response is to keep a rejection log alongside your slippage log: if a fifth of your refusals turn out to be one repeated constraint, that is a symbol or an account to change, not a strategy to abandon. One more distinction is worth internalising - stop loss and take profit are held on the server, but MetaTrader's trailing stop runs inside your terminal, so it stops trailing the moment the platform closes and only the last level it managed to write to the server remains in force.

Latency is not one number. The round trip from decision to confirmation contains your own reaction time, the terminal's processing, the network hop to the broker's trade server, the broker's order handling and risk checks, the liquidity provider's response including any last look window, and the confirmation travelling back. Only the network leg is under your direct control, and at retail scale it is frequently not the largest component. This is why the honest measurement is order-to-fill time recorded from your own trade history rather than a ping figure. A ping tells you how long a packet takes to reach a server; it says nothing about what happens after it arrives.

When you do measure it, look at the shape of the distribution rather than the average, because the average is not what hurts you. A median fill time is close to irrelevant to a scalper: what decides whether a strategy is viable is the slowest few percent of fills, and those cluster in exactly the conditions that generate your signals. Sort your round turns by order-to-fill time, look at the worst five percent, and check what they have in common - a session, an instrument, a size, a minute of the day. If the tail is fat and concentrated, that is actionable. If it is fat and evenly spread, the problem is more likely to be upstream of you than in your connection.

Two published speed figures appear in our data and they should be read carefully. Pepperstone's profile cites sub-35ms execution as reported, alongside more than 900,000 clients and monthly volume above AU$947 billion as an indicator of the depth of flow behind its pricing. FxPro's profile cites an average execution under 12 milliseconds. These are broker-reported figures measured on bases the firms define themselves - what is being timed, from where, on which instruments and over what sample are rarely stated - so they are not comparable with each other, and FXMARE has not independently verified either. Treat them as a reason to run your own test, not as a substitute for one. Our dataset holds no rejection rates, requote rates or fill statistics for any broker on this page, and no server or data-centre locations, so there is no ranking here to give you.

Hosting is the one part of the network leg you can actually change, and our guide to MT5 brokers covers the options and their conditions in detail rather than repeating them here. What belongs on this page is the ceiling. Shortening the round trip improves fills at the margin and does nothing whatever to a strategy that has no edge, and our dataset does not record whether any broker on this page provides hosting or on what terms, so that has to be confirmed directly. If your strategy is mechanical, the largest single latency saving available to most retail traders is not the network at all - it is removing human reaction time by automating the entry, which the order-entry section below returns to.

What client agreements restrict is rarely scalping by name. It is a cluster of related behaviours gathered under a heading such as abusive trading or market abuse: latency and price arbitrage, exploiting a stale or manifestly erroneous quote, trading patterns the firm characterises as taking unfair advantage of its systems, and occasionally a minimum holding time. The remedies these clauses reserve are severe - voiding the trades in question, adjusting the account, restricting it to closing orders only, or terminating the relationship. Read that clause and the order-execution policy before you deposit rather than after your first profitable week, because that is when it tends to be invoked. Our dataset records nothing about whether any broker on this page permits scalping, EAs, news trading or hedging, so this page can only tell you which clauses to look for.

There is a specific list worth confirming in writing for the account you intend to open. Are Expert Advisors permitted, and on this account type rather than somewhere in the product range? Is the account hedging or netting - a strategy that stacks or opposes positions in the same symbol behaves entirely differently on a netting account, and the mode is fixed by the broker when the account is created. Is there any minimum holding time between opening and closing? Is external hosting or a VPS permitted? Are there restrictions on trading around scheduled releases, or on the number of orders per minute? Policies differ between entities of the same brand, which is why the availability note on this page tells you to confirm on the specific account and jurisdiction rather than on the brand.

Where you live narrows the shortlist before any of this matters. Our dataset records that Pepperstone does not accept residents of the United States, Canada, New Zealand or Japan. For a US-based trader that leaves the CFTC and NFA regulated route, which on this page is FOREX.com, operating through Gain Capital Group LLC under NFA ID 0339826, with US retail forex leverage recorded at 1:50 on majors and, per our profile, no negative balance protection for US clients. US retail forex rules also require offsetting positions in the same pair to be closed on a first-in, first-out basis, which removes hedging-style entries from the toolkit entirely - a structural constraint on strategy design rather than a broker preference, and one that no amount of shopping around will change.

In the seconds around a scheduled release the mechanics of execution change, and understanding what changes matters more than any view on the number itself. Liquidity providers widen their quotes or step back from them entirely, the depth behind each price thins, and the book can be momentarily one-sided. Price can print through your stop without ever trading at it, because a stop loss is an instruction to exit at the next available price, not a guarantee of a price. Orders that would fill instantly at midday in London may be rejected outright, and fills routinely land well beyond the requested level. The cost side of the widening - what it does to your realised spread and therefore to your average trading cost - is covered on our low-spread broker page; the point here is that the risk on the trade is not the distance to your stop, it is the distance to wherever the fill actually lands.

That has direct consequences for sizing. If you intend to hold through a release, size the position on the assumption that the exit is worse than the level you set, not on the assumption that it is exact. Some brokers offer a guaranteed stop as a chargeable order type that removes this specific risk in exchange for a premium and, usually, a wider minimum distance; our broker dataset does not record which of the firms on this page offer one or at what charge, so confirm it directly rather than assuming. Be aware too that firms commonly increase margin requirements ahead of scheduled high-impact events and publish notices when they do - a position that was comfortably margined the day before can require more collateral by the time the event arrives.

There is also a contractual dimension. Some agreements restrict trading inside a defined window around a release, or reserve the right to void trades placed in it, and if news trading is part of your approach that clause is not a detail - it determines whether the strategy is permitted at all on the account you are opening. Our dataset does not record any such restriction for the brokers on this page, so the agreement itself is the only place to check. The alternative that many experienced scalpers choose is simply to be flat before the release and to re-enter once the spread has normalised and depth has returned. That is a strategy decision rather than a recommendation, and either choice needs to be made deliberately rather than discovered mid-position.

Platform choice constrains how quickly you can express a decision. Our dataset records IC Markets on MT4, MT5, cTrader and TradingView; Pepperstone on the same four plus its own mobile app; FxPro on MT4, MT5, cTrader, its proprietary FxPro Edge platform, TradingView and a mobile app; and FOREX.com on its Advanced Trading platform, MT4, MT5, TradingView, NinjaTrader and mobile apps. For a scalper the relevant question is not how many platforms a broker offers but whether the one you will actually use supports the entry method your strategy needs - a clickable price ladder, one-click dealing, keyboard shortcuts, a pre-set default volume, and protective orders attached as early as the account allows.

That last point needs stating carefully, because the common advice is wrong on a large share of accounts. On an Instant Execution symbol a MetaTrader order can carry its stop loss and take profit at the moment it is submitted, and it should. On a Market Execution symbol - which is what most raw-spread forex accounts use - many brokers disable those fields on the order dialog entirely, so the position exists unprotected until you modify it after the fill. If that is your account, the gap is real and it sits in exactly the seconds your strategy operates in, so the correct response is to have the modification ready as a single action, or to let an EA place it, rather than to assume protection you have not got. Check which mode applies on your symbols before you rely on either behaviour. One-click trading has a related character: it removes the confirmation dialog and with it the accidental brake, so pair it with a fixed default size you have deliberately set rather than whatever the last trade used.

If the strategy is mechanical, automation is the strongest single improvement available, because it removes human reaction time entirely - usually a larger saving than hosting provides. It also moves the failure modes somewhere new. Symbol naming is not standardised between brokers, so a robot that hard-codes a symbol string breaks silently when moved; netting and hedging accounts execute the same instructions differently; and per-symbol volume steps and minimums cause rejections that look like platform faults. Our MetaTrader 5 broker guide works through the mechanics of running Expert Advisors and testing them properly. Whichever route you take, automation changes how consistently a strategy is applied, not whether the strategy has an edge.

Start by discarding the most common source of false confidence: a demo account cannot answer the question this page is about. A demo does apply the symbol's stops level, freeze level and volume constraints, so it will tell you whether your orders are placeable. What it cannot reproduce is the part that matters most - a liquidity provider's response, last look rejection, the real distribution of slippage, or how the book behaves at your size in a fast market. Use a demo to learn the interface, confirm the instruments and hours you need are available, and rehearse order entry. Then stop drawing conclusions from it about fills.

The test that does work is small, live and parallel. Fund two brokers with an amount you are entirely prepared to lose, run identical logic at identical size in the same sessions, and log four things per trade: requested price, executed price, timestamp, and any rejection or requote. After a few hundred round turns you can compute mean slippage per side, the ratio of positive to negative slippage, the rejection rate, and your realised spread at the moment of entry rather than the broker's advertised average. Low minimums make this affordable: our dataset records no hard minimum deposit for most Pepperstone regions with a nominal $10 figure, $0 at IC Markets' global entity with around $200 suggested for institutional-grade pricing, $100 at FxPro and $100 at FOREX.com. Complete one withdrawal from each account while the sums are still small, so you learn the real timeline before it matters.

Be careful about what the result licenses you to conclude. A month of data at one size in one set of sessions describes that broker, at that size, in those sessions. Scale the size and the depth you consume changes; move the hours and the liquidity changes. It is also worth reading a firm's operational record alongside its execution claims - our profile for FOREX.com records a $700,000 NFA fine in 2022 for improper account adjustments following a trading system malfunction, which is precisely the category of event no amount of latency tuning protects against, and a good reason to keep your own independent record of every fill.

None of this changes the underlying odds. Most retail accounts lose money trading leveraged forex and CFDs, and firms authorised in the UK and EU are required to publish the percentage of their own retail client accounts that do - read that figure for the entity you would actually open with. Scalping's high trade count means execution quality and costs compound faster in both directions than in any other style, which is an argument for measuring them rather than a promise about the outcome. A better fill improves an edge; it does not manufacture one.