7 Anti-Arbitrage Plugins Brokers Use — Named, With Detection Mechanics 2026年04月27日 – Posted in: Arbitrage Software

Investigation · April 2026

Most retail forex brokers run at least one anti-arbitrage plugin. Few publicly admit it. This article names the seven most-deployed plugins in 2026, explains exactly how each one detects arbitrage flow, and shows the signs each leaves in your trading statement before you lose your account.

🕵️ 7 plugins named
🛠️ Detection mechanics
📊 Statement signs
🛡️ Mitigation playbook

What is a broker anti-arbitrage plugin

A broker-side anti-arbitrage plugin is a software module that runs inside the broker’s trading server and detects trading patterns associated with latency, lock or news arbitrage. When detection fires, the plugin degrades execution for the targeted account: by widening spreads, adding asymmetric slippage, requoting orders, holding fills for review, or silently rejecting specific order types. The goal is to make arbitrage strategies unprofitable on the broker’s own server without formally banning the account — preserving the deposit while neutralising the edge.

Table of Contents
  1. Why anti-arbitrage plugins exist
  2. The 7 plugins, named with mechanics
  3. Statement signs each plugin leaves
  4. How to test a broker for plugins
  5. Mitigation playbook
  6. FAQ

Why anti-arbitrage plugins exist

Three commercial pressures push brokers to deploy these plugins, and understanding which pressure applies to your broker tells you which plugin you are most likely facing.

B-book brokers internalise client trades against their own book. When an arbitrageur trades on a stale price for 50–200 ms, the broker is the direct counterparty losing money on every fill. For a B-book operator, an arbitrage strategy is not a market participant — it is a leak in the P&L. Plugins are the patch.

A-book / STP brokers route client orders to liquidity providers. LPs categorise flow as “good” (long-hold, wide-spread, retail) or “toxic” (short-hold, tight-spread, profitable). Toxic flow gets penalised — the broker pays wider markup or loses the LP relationship entirely. The broker’s choice is to filter the toxic flow before it reaches the LP.

Hybrid brokers run both books and use plugins as a sorting mechanism: profitable, fast accounts get pushed to the B-book where the broker can degrade execution silently; unprofitable accounts get sent to the A-book where the broker earns commission without risk.

The result is that almost every retail broker — regardless of what it advertises — runs at least one of the seven plugins below. The differences are which one, how aggressively it is tuned, and whether you can detect it before the strategy is dead.

~85%Retail brokers running ≥1 plugin (BJF 2026 audit)
2–7Days from first detection to account restriction
3.4×Avg slippage on flagged accounts vs baseline
25+Years BJF tracking broker plugin evolution

The 7 plugins, named with mechanics

Each plugin below describes its mechanical operation — not its marketing name. Different vendors sell variants under different brand names, but the underlying detection logic falls into one of these seven categories.

1
Virtual Dealer Plugin (VDP)
The original · still widely deployed
What it does
Simulates a human dealer on the broker server. Intercepts every market order and decides — based on configurable rules — whether to fill at the requested price, requote, slip, or reject.
Detection trigger
Two conditions in combination: order arrives within X milliseconds of a price move (configurable, typically 50–500 ms), and the order direction matches the move. The plugin assumes the trader saw the move on a faster feed.
Punishment mechanic
Asymmetric slippage — negative slippage (against the trader) on profitable opens; symmetric or zero slippage on losing opens. Over weeks, this drains the strategy without obvious account-level intervention.
Common deployment
Mid-tier B-book brokers, white-label retail platforms, prop-firm platforms (especially for arbitrage prevention during the challenge phase).

2
Anti-Latency Plugin (ALP)
Surgical · purpose-built for latency arb
What it does
Compares server timestamp on incoming orders against the price-feed timestamp at order receipt. If the difference is less than the typical “human reaction” floor (configurable, often 200–400 ms), the order is flagged.
Detection trigger
Order placement RTT below threshold + order direction correlated with last tick direction. The plugin essentially asks: “Did this trader place this order faster than a human plausibly could?”
Punishment mechanic
Hold the order in a delay buffer (typical 100–500 ms hold), revalidate price on release, fill at the new price. The faster the original arbitrage, the more it is hurt by the delay.
Common deployment
Brokers that publicly tolerate “trading bots” but quietly deploy ALP to neutralise latency-sensitive strategies. The broker can claim it allows automation while still defending the B-book.

3
Slippage Engine (asymmetric slippage module)
Statistical · invisible at first
What it does
Tracks per-account slippage statistics. Once an account exhibits “abnormally low” execution variance (a sign of automated, fast strategies), the plugin actively biases slippage on subsequent fills — negative on profitable directions, neutral on losing ones.
Detection trigger
Statistical fingerprint: average fill consistency, ratio of profitable fast-closes to total trades, and short hold-time clustering. The plugin builds a per-account “automation score” over time.
Punishment mechanic
Slippage is added at order match time, drawn from a non-zero-mean distribution biased against the account. The trader sees occasional large negative slips that always seem to occur on the best opportunities.
Common deployment
Mainstream retail brokers that brand themselves as “no-dealing-desk” or “STP”. The slippage engine is technically a market-maker tool, but it is layered onto an STP route as a “best-execution adjustment”.

4
Last Look / Hold Window
LP-side · institutional origin
What it does
Holds the order for a short window (typically 30–250 ms) before either filling or rejecting. During the hold, the LP watches the market and decides whether the price is still favourable to fill.
Detection trigger
Last look does not “detect arbitrage” — it is structurally biased against fast flow. Any trader whose orders are more profitable to fill than to reject statistically gets a higher rejection rate.
Punishment mechanic
Outright rejection on the most valuable fills. The trader sees high rejection rates (sometimes 30–50%+ in volatile periods) on exactly the orders that would have been most profitable.
Common deployment
All ECN, prime-of-prime and LP-routing brokers. Last look is built into the LP layer itself — the broker often cannot turn it off, only renegotiate the hold time.

5
AI Pattern Detection (modern, 2023+)
Machine learning · hardest to fool
What it does
A trained machine-learning model scores each account in real time on a multi-feature vector: order timing distribution, hold-time distribution, win/loss asymmetry, lot-size patterns, instrument selection, session-time clustering, and correlation with broker’s own price moves.
Detection trigger
No single feature triggers detection; the model classifies the entire account’s trading pattern as “arbitrage-like”, “news-trading-like”, or “retail-like” with a confidence score. Above threshold, the account is flagged.
Punishment mechanic
Routing change. The broker silently re-routes flagged accounts to a higher-friction execution path (extra hop, wider spread, longer hold). The trader sees gradually deteriorating execution that “tracks” their profitability.
Common deployment
Top-tier retail brokers and prop firms with the engineering budget to train custom models. Increasingly common since 2024. This plugin renders most pre-2023 masking strategies ineffective on its own.

6
Dynamic Spread Widener
Per-account · spread-engine integration
What it does
Maintains a per-account spread offset on top of the broker’s base spread. Once an account is flagged, the plugin adds 0.5–3.0 pips of additional spread visible only to that account, on the instruments the strategy uses.
Detection trigger
Often paired with the AI Pattern Detection or Slippage Engine. Once those modules have flagged an account, the spread widener is the response — implemented at the spread-aggregation layer rather than at execution.
Punishment mechanic
The arbitrage opportunity disappears mathematically. If the price gap captured by the strategy is 0.6 pips and the broker adds 1.0 pip to the spread, every trade is now negative-expectancy at entry.
Common deployment
Brokers operating their own spread aggregator. Hardest to detect because the trader sees a spread chart that “looks normal” — the widener only applies when the trader’s terminal queries quotes for execution.

7
Account-Level Throttle (Speed Bump)
Order-rate limiter · last-resort tool
What it does
Limits the rate at which an account can place orders — typically expressed as a minimum interval between orders (e.g., 500 ms) or a maximum orders-per-minute cap. Crosses the limit and orders are queued or rejected.
Detection trigger
Not “detection” in the same sense — usually applied as a blanket policy to all accounts on a specific server, or as a “second strike” measure after another plugin has flagged the account.
Punishment mechanic
Strategies that depend on rapid sequential orders (latency arbitrage, news trading on multiple symbols) are mathematically broken because the throttle introduces a hard floor on action speed.
Common deployment
Brokers with limited engineering capacity that cannot deploy AI detection — a throttle is the cheapest blanket defence. Also common at prop firms that want to “permit automation” while making fast strategies unprofitable.
One broker, multiple plugins
Modern retail brokers commonly stack 2–4 plugins simultaneously. The standard combination in 2026 is: AI Pattern Detection (the trigger) + Slippage Engine (the punishment) + Dynamic Spread Widener (the kill). VDP and ALP remain on legacy infrastructure as fallbacks.

Statement signs each plugin leaves

Plugins differ in how they degrade execution, but each one leaves a measurable footprint in your trade history. The table below shows the most reliable single-metric signal for each plugin. None of these signs are conclusive on their own — but together they triangulate which plugin is active.

Plugin Strongest statement signal Test window
1. Virtual Dealer Slippage skewed against profitable trades; “neutral” on losing trades. Asymmetry > 0.3 pips. 50+ trades
2. Anti-Latency Execution time histogram with hard floor at the broker’s hold window (e.g., spike at 200–400 ms). 100+ orders
3. Slippage Engine Slippage variance grows over time as account “ages”; rolling 7-day average slippage worsens. 2–4 weeks
4. Last Look Rejection rate on market orders > 5% in calm markets, > 25% during news. Concentrated on profitable fills. 1 week + 1 news event
5. AI Pattern Detection Step-change in execution quality (slippage, fills, spread) at a specific date — usually 5–14 days after first profit run. 3–4 weeks
6. Dynamic Spread Widener Spread visible at order placement > spread on broker’s quote feed. Compare logged spread vs charted spread. 1 week of logging
7. Account Throttle Order placement times cluster at exact intervals (e.g., always > 500 ms apart). Hard floor in inter-order timing. 50+ rapid orders
How to log this data
For brokers that support FIX API, log every order’s send-timestamp, server-receive-timestamp, fill-timestamp, fill-price, and quoted spread at send time. For standard retail platforms, the exported account history contains the minimum required fields, but tick-level spread comparison requires a separate market-data recorder. SharpTrader Pro logs all of this automatically per-account.

How to test a broker for plugins

The right testing protocol catches plugins before they cost you serious capital. Six-step protocol used internally by BJF Trading Group when onboarding a new broker:

1

Open the account at minimum depositNever deposit full strategy capital before validating execution. The minimum deposit lets you run the test protocol without serious downside; if the broker is hostile, you abandon it for the cost of a coffee.

2

Run your strategy at minimum lot for 14 daysThe first two weeks are the AI Pattern Detection observation window. Most modern detection systems need ~7 days of consistent automation behaviour before they classify an account as arbitrage-like. Running for 14 days at the smallest viable lot lets you see the post-detection deterioration cleanly.

3

Log every order with full metadataSend-timestamp, fill-timestamp, requested price, fill price, slippage, spread at send, current tick direction, and (where available) order ID. This is your ground truth. Without per-order metadata you cannot tell which plugin is firing — only that performance is degrading.

4

Force a news event into the testSchedule the test to span at least one Tier-1 news release (NFP, CPI, FOMC). Last Look and Slippage Engine plugins behave dramatically differently in volatility. A broker that looks fine in calm markets often shows 30–50% rejection rates during NFP — the signature of aggressive Last Look.

5

Compare profitable vs losing trade executionBuild two distributions: slippage on profitable trades, slippage on losing trades. If they are statistically identical (Kolmogorov-Smirnov test, p > 0.1), no asymmetric punishment is happening. If they differ significantly, you have evidence of VDP, Slippage Engine or AI-routed degradation.

6

Verdict matrix and decisionBased on the data, classify the broker as Clean (proceed to scale), Limited (some plugins active but tolerable for masked strategies), or Hostile (multiple plugins, abandon). BJF clients running SharpTrader Pro receive an automated verdict report after the 14-day test based on this exact protocol.

Why this protocol works
Brokers tune plugins to be invisible to a casual trader. The detection thresholds, observation windows, and punishment magnitudes are set assuming the trader is not measuring. A trader who logs the right metadata and runs a structured 14-day test reverse-engineers the broker’s plugin configuration before the account is large enough to matter.

Mitigation playbook

You cannot remove plugins from a broker’s server — only avoid, mask, or out-route them. Six tactics, ordered from quickest to most structural.

🎭

Masking strategies

Phantom Drift and BrightDuo (built into SharpTrader Pro) generate trade patterns that defeat AI Pattern Detection: variable hold times, asymmetric lot sizing, decoy losing trades, and split-execution across two accounts.

Quickest fix
Built into product
🔀

Multi-broker portfolio

Run the same strategy on 2–3 brokers simultaneously, each with smaller lot sizes. When one broker flags the account and degrades execution, the others continue. The trade-off is operational complexity and lower per-broker scaling.

Diversification
🔒

Lock arbitrage variants

Lock CL2 and BrightTrio Plus pre-establish opposing positions across accounts and only “fire” by closing the profitable leg. To the broker’s plugin, the closes look like ordinary position management — not arbitrage entries. Detection rates drop to almost zero.

Plugin-invisible
🏛️

Move to FIX API

Prime-of-prime FIX accounts typically don’t run the seven plugins above (the LP imposes Last Look but not the rest). Capital threshold is higher ($10k–$50k+) but the plugin landscape is dramatically cleaner. See our FIX API forex guide.

Capital required
$10k+
🐢

Slow down deliberately

For some strategies (statistical arb, pair trading) speed is not the edge. Running on hour-to-day timeframes makes ALP and VDP irrelevant — the plugins don’t fire on slow flow. Match strategy speed to what the broker tolerates.

Easy
🧪

Continuous A/B testing

Maintain a “control” account — same strategy, smaller lot — that you only start when you suspect a flag has fired on the main account. The control isolates what is plugin-related vs market-related. Without it you cannot tell whether a profit drop is the broker or the market.

Operational

SharpTrader Pro includes plugin-detection logging

Per-trade execution analytics, automated 14-day broker verdict reports, and the masking strategies (Phantom Drift, BrightDuo, Lock CL2) that defeat each of the seven plugins above. Built and maintained by BJF Trading Group since 2008.

Explore SharpTrader Pro →
Read the broker guide

FAQ

Are anti-arbitrage plugins legal?
In most jurisdictions, yes. No regulator in the major retail-forex jurisdictions (CFTC, FCA, ESMA, ASIC) classifies plugin-based execution adjustment as illegal — it is treated as a contractual matter between broker and client. The legality is asymmetric, however: the broker’s terms of service usually permit plugin operation, while the trader has limited recourse beyond switching brokers. See our Is Forex Arbitrage Legal? guide for the regulatory landscape.
Do all retail brokers run anti-arbitrage plugins?
In our 2026 audit of 200+ retail forex brokers, approximately 85% run at least one of the seven plugins. The remaining 15% are either prime-of-prime FIX-only operators (which rely on LP-side Last Look rather than broker-side plugins) or very small brokers without the engineering capacity to operate them. “Plugin-free retail broker” is essentially a marketing claim — not a category that exists in practice.
Can I detect anti-arbitrage plugins in my trading statement?
Yes, with the right logging. Each plugin leaves a measurable footprint — see the statement-signs table above. The simplest detection is comparing slippage on profitable trades versus losing trades. If they differ statistically (Kolmogorov-Smirnov test, p < 0.05) the broker is running asymmetric punishment. Tick-level logging of spread, RTT, and fill metadata catches the more subtle plugins.
Does FIX API skip the plugins?
Mostly yes, with one exception. Prime-of-prime FIX accounts typically don’t run VDP, ALP, Slippage Engine, AI Pattern Detection, Spread Widener or Throttle — those are retail-platform-side modules. However, the LP layer behind a FIX session almost always implements Last Look. The total plugin exposure on FIX is therefore one (Last Look) versus three to four on a typical retail account. See our FIX API forex guide for details on how to negotiate hold times with the LP.
Which plugin is most common in 2026?
AI Pattern Detection (#5) became the dominant primary detection layer between 2023 and 2026. Five years ago, VDP was the standard primary plugin; today, AI detection has replaced it on tier-1 retail brokers, with VDP and ALP retained as legacy fallbacks. Slippage Engine (#3) remains universal — almost every broker runs it as a punishment layer downstream of the AI detection.
Can SharpTrader Pro bypass these plugins?
Not “bypass” — that would mean modifying the broker’s server, which is impossible from the client side. SharpTrader Pro defeats plugins by changing what the broker observes: Phantom Drift generates trading patterns that AI detection classifies as retail; Lock CL2 hides arbitrage entries inside ordinary position-close events; BrightDuo splits execution across two accounts so neither account exhibits the full arbitrage signature. The plugins still run — they just don’t trigger.
Why don’t brokers publicly disclose their plugins?
Three reasons. First, disclosure makes detection trivial — published thresholds and rules can be reverse-engineered immediately. Second, plugins sit in a regulatory grey area; brokers prefer not to draw the attention of “best-execution” enforcement. Third, plugins contradict the marketing narrative many brokers depend on (“STP”, “no dealing desk”, “raw spread”) — explicit disclosure would invalidate the marketing claim.
Is “best execution” regulation enforced against plugins?
Inconsistently. ESMA and the FCA require best execution but interpret it as “in line with the broker’s published execution policy” — and execution policies are written broadly enough to permit Slippage Engines and Last Look. Active enforcement against asymmetric slippage exists in equities markets but is rare in retail FX. The CFTC has taken action against specific brokers for “Virtual Dealer abuse” historically, but the bar for action is high. Practically, plugins are a market reality and regulators address them post-hoc rather than preventatively.
What do I do if I detect a plugin on my broker?
Three options, in escalating cost. First, switch strategy — move from latency arbitrage (highly plugin-sensitive) to lock or statistical arbitrage (much less so) on the same broker. Second, add masking — Phantom Drift, BrightDuo or Lock CL2 to defeat the specific plugin you’ve identified. Third, move broker — accept the friction of opening a new account and run your test protocol again. Most BJF clients use options one and two before resorting to option three.
Are anti-arbitrage plugins the same as anti-bot filters?
Overlapping but not identical. “Anti-bot” filters target any automation — they fire on hold times shorter than a human reaction window regardless of strategy. “Anti-arbitrage” plugins are more specific: they target trading patterns correlated with profitable arbitrage flow, which is a subset of automation. A broker can permit automation while still running anti-arbitrage detection (and many do). The seven plugins above all qualify as anti-arbitrage; some, like the Throttle, cross over into general anti-bot territory.