Hedge Arbitrage — Two-Broker Latency Strategy Explained

Most retail traders who try arbitrage start with one-leg レイテンシ・アービトラージ at a single broker. They achieve a few weeks of strong results, scale up, and shortly afterwards run into either an account closure or a withdrawal dispute. The cause is almost always the same: at large size, a single-broker trading pattern with 80%+ wins on short holding times is impossible to disguise as discretionary trading.

ヘッジアービトラージ exists to solve this problem. By splitting each trade into two legs across two brokers, the trading pattern at each individual broker looks structurally different — and the strategy can scale up to capital levels where one-leg latency is no longer viable. This guide covers the mechanics, the broker requirements, the detection profile, the capital efficiency, and the cases where hedge arbitrage is the better choice over one-leg.

What is hedge arbitrage

Hedge arbitrage is a two-account latency strategy. Each trade has two legs:

1. Slow leg. A position is opened at the slow-feed broker — the broker whose stale quote provides the arbitrage edge. This is the leg that captures the profit.
2. Fast leg. The opposite position is opened simultaneously at a fast-feed broker — typically a top-tier ECN または STP broker. This leg absorbs market exposure during the holding period.

Net market exposure across the two legs is approximately zero. If the market moves up while the position is open, the slow-leg long gains and the fast-leg short loses by approximately the same amount. The profit is not directional — it comes from the fact that the slow broker filled at a stale price, while the fast broker filled at the current institutional price.

The strategy is sometimes called 2本足のレイテンシ, cross-broker hedge, あるいは locked arbitrage. All three names describe the same mechanism.

Why hedge arbitrage exists

One-leg latency arbitrage is mechanically simpler and produces a higher profit per dollar of margin. So why bother with the operational complexity of running two brokers? Five reasons:

1. Lower per-trade risk

In one-leg, you carry directional market risk for the holding period (typically 50 ms to 5 seconds). Most of the time the market does not move enough to matter. Once in a while — during a news spike or a sudden order-flow imbalance — the market moves against you faster than the broker’s quote catches up, and the trade closes at a loss. In hedge mode the fast leg cancels this risk; profit and loss are decoupled from market direction during the hold.

2. Lower detection signature at the slow broker

In one-leg, your slow broker sees 100% of your trading: every entry, every exit, every win, every loss. Pattern detection is a question of when, not whether. In hedge mode, the slow broker sees only one side of every trade. Without the offsetting hedge it cannot see, individual trades look like ordinary directional positions — sometimes profitable, sometimes not. The detection signature is much weaker than 80%+ wins on short holds at a single broker.

3. Better prop-firm compatibility

Most prop firms (FTMO, FundedNext, The5ers, MyForexFunds replacements) explicitly prohibit “latency exploitation,” “tick scalping,” or “arbitrage” in their rules. The hedge module — particularly the 2-レッグ遅延 3 variant — is designed with longer holding times, configurable randomisation, and a P&L profile that more closely resembles short-term scalping than tick-level arbitrage. Compliance is the trader’s responsibility, but hedge mode is the variant most often configurable to comply with common prop firm rule sets.

4. Better scaling at large capital

At small accounts ($5k–$20k) one-leg arbitrage works for weeks at most brokers. At $100k+ accounts it triggers detection within days. Hedge arbitrage scales meaningfully better — large position sizes still look like ordinary directional bets at the slow broker, and the fast broker (whose feed is fast and whose business model accepts the flow) absorbs the hedge volume cleanly.

5. Operational symmetry across brokers

Many institutional and semi-institutional desks already operate accounts at multiple brokers for redundancy and capital diversification. Hedge arbitrage uses that infrastructure rather than fighting it. If you already have accounts at three or four brokers for unrelated reasons, the marginal cost of running hedge arbitrage across them is low.

How hedge arbitrage works — full pipeline

The pipeline is a coordinated version of the latency arbitrage flow, with two execution paths instead of one:

1. Reference tick arrives. The fast feed (CME futures via CQG または リズミック, LMAX, 積分OCX, あるいは cTrader 生) shows a price move.
2. Compare to slow broker quote. If the slow broker’s quote is stale and the divergence exceeds the configured threshold, a signal is generated.
3. Two orders constructed. A buy order for the slow broker (capturing the stale ask) and a sell order for the fast broker (locking in the current market price as the hedge).
4. Orders fired in sequence. The slow-broker order is sent first or simultaneously, depending on the variant. Inter-leg latency is typically 1–5 ms.
5. Both fills confirmed. Position is now open: long at slow broker, short at fast broker. Net market exposure is approximately zero. P&L equals the price differential between the two fills.
6. Convergence. The slow broker’s quote catches up. The slow leg now shows a profit equal to the original stale-ask gap; the fast leg shows a small loss for the bid-ask spread and any market drift. Net is positive.
7. Both legs closed. The slow leg closes at profit; the fast leg closes at a small loss. Net per trade: positive after spreads and commissions on both sides.

The mathematical edge is the slow-broker gap minus the round-trip cost on both brokers (slow-broker spread plus commission, fast-broker spread plus commission, any small market drift during the hold). Properly configured, the gap exceeds combined costs on a high percentage of triggered signals.

The three 2-legs variants — what’s different about each

The platform ships three variants of 2-legs latency arbitrage. Each was developed in response to a specific operational problem and is suited to a different broker/prop-firm context.

2-legs latency 1 — classic synchronous

Both legs are constructed from the same signal and fired as close to simultaneously as the network allows. Inter-leg delay is minimised. Holding time is short — typically 50 ms to 2 seconds. P&L target is the immediate convergence between slow and fast prices.

Best for: aggressive retail accounts at slow B-book brokers, where speed matters more than signature management. Highest profit per trade among the three variants but also the most obvious signature.

2-legs latency 2 — sequential with confirmation

The slow-broker order is sent first. The fast-broker hedge is fired only after the slow order is confirmed filled. This eliminates the worst-case scenario of variant 1 — a hedge filled but a slow order rejected, leaving the trader with naked directional exposure.

Inter-leg latency is higher (typically 5–15 ms longer hedge delay) and a small market move during the gap may eat into profit. In return, the failure mode is significantly safer.

Best for: brokers with inconsistent fill behaviour or noticeable rejection rates. Recommended baseline for most operators.

2-legs latency 3 — extended-hold prop-firm-aware

The most operationally sophisticated variant. Several differences from variants 1 and 2:

• Longer holding times (configurable, typically 5–60 seconds rather than sub-second)
• Lot-size and timing randomisation to avoid mechanical signature
• Filtered signal selection — the system intentionally skips a portion of high-conviction signals to avoid statistical detection
• P&L throttling — daily profit caps that keep account performance below visibility thresholds at prop firms

Net profit per trade is lower than variants 1 and 2, but operational lifetime at a single account is much longer, and the variant is the one most often configurable to comply with common prop firm rule sets.

Best for: prop firm accounts (FTMO, FundedNext, The5ers), large capital deployments where lifetime matters more than per-trade yield, and any scenario where the trader needs the strategy to look as much like ordinary scalping as possible.

Capital requirements and split

Hedge arbitrage requires capital at two brokers. The split is not 50/50.

At the slow broker, the account holds the position that captures the profit. This is where most of the capital usually sits — sized to support the chosen lot size with broker-specific margin requirements, typically 1:30 to 1:500 leverage on FX and 1:20 to 1:200 on gold.

At the fast broker, the account holds the offsetting hedge. Capital here can be smaller because top-tier ECN brokers typically offer better leverage and faster margin recycling. However, the fast account must always have enough free margin to open the hedge instantly — running the fast account too lean creates the worst failure mode (hedge rejected while slow leg fills).

Practical splits seen in deployment:

• $10,000 total: $7,000 slow / $3,000 fast — minimum viable for variant 2 or 3
• $50,000 total: $35,000 slow / $15,000 fast — comfortable size for prop-firm-style operation
• $200,000+ total: $130,000 slow / $70,000 fast — institutional small-desk size; often split across multiple slow brokers

Below $10,000 total the fixed costs (two broker accounts, fast feed subscription, VPS いいえ LD4 または NY4) eat too much of the gross profit. The strategy starts to be economically interesting at $10,000 and gets meaningfully more profitable per dollar at $50,000+.

Broker selection — slow side

The same criteria as one-leg latency arbitrage apply to the slow broker. You want:

• A measurable feed lag versus the institutional reference (100–300 ms is the sweet spot)
• Reasonable XAUUSD or EURUSD spread
• Permissive execution — no aggressive requoting, no execution-delay speed bumps
• No explicit anti-arbitrage clause in terms of service (or terms loose enough to be worked with)
• A regulatory profile you are comfortable with
• Available platform — エムティーフォー, MT5, cトレーダー, DXTrade, or FIX integration

Typical slow brokers are mid-tier dealing-desk or hybrid A/B-book brokers using third-party aggregator feeds. We maintain an internal list and can confirm broker compatibility via email — see the お問い合わせ.

Broker selection — fast (hedge) side

The fast broker has the opposite profile. You want:

• Top-tier ECN/STP execution — feed within 5–25 ms of institutional
• Tight spreads, transparent commission
• FIX or API FIX connectivity preferred (lowest hedge latency)
• Reliable margin recycling so the hedge account can sustain rapid order flow
• Good fill rate — the hedge needs to be reliable; partial fills create exposure
• Permissive treatment of high-frequency activity — many ECNs are happy with HFT flow because they earn commission on it

Common fast-broker choices in deployment: LMAX direct, cTrader-based ECNs with raw spread accounts, IC Markets / Pepperstone-class STP brokers with FIX accounts, or institutional liquidity providers via prime-of-prime relationships.

Infrastructure — coordinating two brokers

Hedge arbitrage infrastructure is one degree more complex than one-leg. The arb server needs:

• Cross-connect to the fast feed (CME via Rithmic/CQG or LMAX co-ordination point)
• Network path to the slow broker — ideally same datacenter; if not, the lowest-latency available route
• Network path to the fast broker — same datacenter is ideal; many FX ECNs and the fast feed share LD4 or NY4
• Time synchronisation — both broker clocks and the arb server clock should be NTP-synced to better than 1 ms accuracy for proper trade reconciliation

The dominant geometry: one box in LD4 with cross-connects to LMAX, the slow broker, and the fast broker. NY4 is the equivalent if your slow broker is US-routed or you trade against CME futures heavily.

Detection profile — what each broker sees

This is where hedge arbitrage’s main advantage shows up.

What the slow broker sees

Individual buy and sell trades, with mixed outcomes. Some win, some lose, depending on whether the slow leg captured a real edge or got caught by feed noise. The win rate visible to the slow broker on its account is typically 55–75% — high but not implausibly high. プロフィットファクター is in the 1.5–2.5 range. Holding times in variants 2 and 3 are long enough to look like ordinary short-term trading rather than tick scalping.

Detection at the slow broker is still possible, but it requires the broker to either correlate trades with external market events (which most retail brokers do not do at scale) or to flag the account on win-rate or P&L profile alone. Both are feasible but slower than detecting one-leg arbitrage.

What the fast broker sees

The opposite of every slow-broker trade. Mostly small losses (the fast broker absorbs the spread + market drift on the hedge), with a relatively flat or slightly negative P&L over time. From the fast broker’s perspective, the account looks like a high-volume directional trader who is consistently slightly losing — a profitable customer for an STP/ECN broker that earns commission on the volume.

The fast broker has very little incentive to investigate or restrict this account. Volume + commission + small flow loss is the ideal customer profile for an A-book broker.

The main risk — inter-leg slippage

The unique risk of hedge arbitrage is the gap between the two legs. If the slow leg fills at one price and the fast leg fills at a different price than the system expected, the hedge is partially broken and the trade may be a loss instead of a profit.

This happens in three scenarios:

1. Slow leg fills, fast leg rejects. Account is now naked long at the slow broker. Most dangerous failure mode. Variant 2 mitigates this by sequencing fast-leg only after slow-leg confirmation; variant 1 is exposed to it.
2. Fast leg fills, slow leg rejects. Less dangerous because the fast leg is at a top-tier broker with predictable execution; you can typically close it at the next tick with minor cost.
3. Both legs fill but at unexpected prices. Slow leg slipped, or fast leg slipped, or both. Net P&L on the trade is below expectation. Frequent in volatile sessions.

Production systems handle these with: maximum acceptable スリッページ caps (abort the second leg if first-leg slip exceeds threshold), automatic position-flatten logic for orphaned legs, and pre-trade margin checks at the fast broker to reduce rejection rates.

Realistic profitability

Hedge arbitrage produces lower profit per trade than one-leg latency (because you pay broker costs on two sides) but typically higher reliability, lower variance, and longer operational lifetime. Representative ranges:

• 1日あたりの取引数 30–250 depending on volatility regime and which variant is active
• 勝率: 70–88% (higher in variant 1, lower but with fewer losses in variant 3)
• Average net profit per trade after both brokers’ costs: 1.5–10 USD per 0.1 lot
• プロフィットファクター 1.4–3.0
• Monthly return on $10k–$50k combined capital: 4–18% in normal markets, occasionally higher in volatile months

Two of the three live FxBlue-verified reference accounts on the パフォーマンスページ run 2-legs configurations. They are not pure single-instrument tests; they show realistic combined output across multiple instruments and brokers.

Anyone advertising “guaranteed monthly returns” or “set-and-forget” profitability for hedge arbitrage is misrepresenting the strategy. Sustainable hedge arbitrage produces strong but not magical returns and requires active management of broker relationships, feed quality, and capital allocation.

One-leg vs hedge — when to use which

Factor	One-leg better	Hedge better
Capital	Under $10k total	$10k+ total
Account type	Personal retail account	Prop firm account
Risk tolerance	High — accept short-term market risk	Lower — want market-neutral hold
Operational complexity	Simple — one account to monitor	Two-broker coordination
Lifetime per broker	Weeks at small size	Months at moderate size
Profit per trade	Higher (no hedge cost)	Lower
Detection signature	Strong — easy to spot	Weak — looks like ordinary trading

Common mistakes in hedge arbitrage

1. Underfunded fast account. Insufficient margin at the hedge broker means hedge orders get rejected. Always overfund the fast account by 30–50% beyond the minimum.
2. Variant 1 with an unreliable slow broker. If the slow broker has high rejection rates, variant 1 leaves you with naked hedges several times per day. Switch to variant 2 (sequential) for any broker with execution reliability under 99%.
3. Treating the fast broker as a profit centre. The fast broker is a cost; do not optimise it for profit. Optimise it for fill reliability and tight spreads.
4. Ignoring inter-leg time difference. If your fast leg lags the slow leg by 50+ ms during a fast-moving market, the hedge is broken before it arrives. Co-locate.
5. Running a single fast feed for both signal and hedge sizing. If your fast feed disconnects, both legs fire on stale data. Use independent feed sources for signal generation and execution timing where possible.
6. Skipping the news blackout. During FOMC or CPI both feeds move in lockstep across brokers, hedge profitability collapses, and false signals dominate.
7. Not tracking inter-broker P&L correlation. Hedge mode P&L should be slow-broker positive + fast-broker slightly negative. If both brokers are showing similar P&L direction, the hedge is broken — investigate immediately.

よくある質問

Is hedge arbitrage legal?

Yes, in jurisdictions where retail FX/CFD trading is legal. The contractual question is whether your slow broker prohibits “latency exploitation” or “abusive trading practices” — many do, and breach is grounds for account closure. Hedge mode is harder to detect than one-leg, but the underlying activity is the same and the same broker terms apply.

Do I need accounts at two different broker companies?

Yes — that is the entire point. The slow broker provides the arbitrage edge; the fast broker absorbs market risk. Accounts at the same broker (even sub-accounts) cannot be used as the hedge because both accounts share the same feed and execution.

Can the same software run all three variants?

Yes. The platform’s 2-legs latency module supports variants 1, 2, and 3 as configuration options on the same engine. You can switch between them per instrument, per session, or per broker without reinstalling — typically variant 2 for the bulk of trading and variant 3 reserved for prop-firm-routed accounts.

Is hedge arbitrage compatible with FTMO and other prop firms?

The 2-legs latency 3 variant is designed to be configurable for compliance with common prop firm rule sets — longer holds, randomised lot sizing, P&L throttling. Compliance is the trader’s responsibility; rules vary by firm and update frequently. Read your specific firm’s terms and verify your configuration matches before going live.

What happens if the fast broker rejects the hedge?

In variant 2, the slow leg is not opened until the fast hedge is confirmed possible — so a rejected hedge means the trade simply does not happen. In variant 1, a rejected fast hedge after a filled slow leg leaves you naked; the platform’s risk module either closes the slow leg immediately at market or flips the slow leg to flat depending on configuration. Variant 1 should only be used at fast brokers with proven 99%+ fill reliability.

Does hedge arbitrage work on gold and indices?

Yes — and gold (XAUUSD) is one of the most profitable instruments for hedge arbitrage because the slow-broker feed dispersion is structurally larger than for major FX pairs. See the 金裁定 guide for details. Indices (US30, NAS100) work on slow-feed brokers but require careful news-event handling.

How is hedge arbitrage different from a “lock” or “locked position”?

A lock is two opposing positions at the same broker — typically used for risk parking rather than profit extraction. Hedge arbitrage is two opposing positions at different brokers, where the price differential between the two fills is the profit. Mechanically and economically they are different strategies despite both being “long and short at the same time.” See ロックアービトラージ in the glossary for the related distinction.

What if my slow broker freezes quotes during volatility?

This is one of the main failure modes during news events. If the slow broker freezes its quote at a stale price while the market moves significantly, the platform should detect the freeze and abort signal generation — frozen quotes generate false signals on every reference tick. The risk module includes freeze-detection and auto-pause; configuration is broker-specific.

Can I run multiple slow brokers with one fast broker?

Yes — and many institutional deployments do exactly this. One fast broker (typically LMAX or an institutional FIX account) hedges flow from three or four slow brokers in parallel. This concentrates hedge activity in one place where the operational profile is welcomed (high-volume STP customer) while diversifying detection risk across multiple slow accounts.

What is the smallest profitable hedge arbitrage setup?

$10,000 split as $7,000 slow / $3,000 fast is the practical minimum once feed and VPS costs are factored in. Below that, fixed costs eat too much of the gross profit. The strategy gets meaningfully more profitable per dollar at $50,000+ where lot sizes can be increased without breaching detection thresholds at either broker.

続きを読む

• HFTアービトラージ — 完全ガイド — full pillar covering all arbitrage types and the broader strategy landscape
• Latency Arbitrage Explained — the one-leg foundation that hedge arbitrage builds on
• Gold Arbitrage — Strategies and Opportunities — XAUUSD-specific deep dive (often run in hedge mode)
• HFTアービトラージプラットフォームの仕組み — live FxBlue-verified reference accounts including 2-legs configurations
• HFT＆裁定取引用語集 — 55の重要用語解説