Execution cost remains the invisible tax on every FX strategy. Since August 2026 a number of practical developments — faster adoption of machine‑learning routers, vendor pilots for consolidated reporting, and renewed regulatory scrutiny of last‑look and disclosure — have changed the execution landscape. This update explains what’s different in November 2026, why it matters, and what traders (retail and institutional) should do now.

Why this matters now

Two continuing structural forces that motivated the original analysis remain central, but their expression shifted in 2026:

  • Ongoing fragmentation with richer venue types. Displayed ECNs and multi‑bank pools remain core, but more varieties of non‑displayed execution—internal crossing fabrics, broker internalisers with automated hedging, and exchange‑style “lit+dark” venues—are materially larger and more frequently reachable via modern routers.
  • Algorithmic evolution. Smart routers and execution algos now regularly incorporate machine‑learning models trained on venue microstructure and latency characteristics; that makes routing both more powerful and more opaque.

For strategy designers — from scalpers to carry traders — these changes mean the same nominal spread on EUR/USD or USD/exotic does not equal the same tradable cost. How you access a quote, whether your router can reach a hidden pool, and whether a counterparty uses last‑look or FIFO matching will determine realized slippage.

Background: what led here

Market structure evolution accelerated after the pandemic‑era liquidity changes. The BIS Triennial Survey (2022) underlined the scale of FX turnover and the diversity of trading venues; since then, vendors and sell‑side desks have pursued higher connectivity and richer matching protocols. In 2025–26 several factors pushed further change:

  • Cloud adoption and low‑cost connectivity reduced barriers for smaller liquidity providers to offer matching pools.
  • Buy‑side demand for deterministic execution metrics prompted vendors to pilot consolidated post‑trade reporting — not a formal consolidated tape, but richer cross‑venue fills data shared in anonymised pilots.
  • Regulators in major markets increased scrutiny of last‑look and disclosure practices; several brokers updated client documentation and monitoring in 2026.

Data and evidence (what we see in 2026)

Hard public data for FX execution quality remains partial, but several verifiable signals matter:

  • BIS 2022 remains the best public baseline for global turnover; market participants we surveyed in 2026 report similar daily liquidity in core pairs but greater intra‑day fragmentation across more venues.
  • Vendor telemetry and post‑trade pilots launched in H1–H2 2026 show that for mid‑sized orders, routers that incorporate venue latency and hidden liquidity scores reduce average implementation shortfall versus naive SOR by measurable amounts — firms reported improvements in the low‑to‑mid single‑digit basis points for EUR/USD in normal windows.
  • Regulatory filings and disclosures in 2026 reveal more brokers publishing last‑look incidence rates and cancel/requote metrics; traders can now request these as part of due diligence more easily than two years ago.

Venue differences — what’s changed since August 2026

ECNs and multi‑bank pools

Then: tight displayed spreads, transparent time‑priority fills for small/medium sizes.

Now: ECNs still provide the cleanest passive execution for tiny to moderate tickets, but two issues are more prominent:

  • Shallow showed depth at times. More top‑of‑book orders are “stub” quotes placed to attract flow from smart routers; real tradable depth often sits behind hidden or exchange‑internal queues.
  • Latency and co‑location matter more. As routers use microsecond profiling, access tiering (membership, co‑location, cloud proximity) directly affects fill probability for aggressive slices.

Retail market‑makers and internalisers

Then: higher fill rates for small tickets; wider effective spreads.

Now: many retail platforms have upgraded to hybrid models: posted quotes for client display, but automated hedging into pooled liquidity or back to bank partners. Two consequences:

  • Retail clients still enjoy high fill reliability for micro‑lots, but realized slippage on marketable orders is more variable during stress windows when liquidity providers widen or pause hedging.
  • Some platforms now offer transparent per‑trade execution analytics (e.g., midpoint slippage, last‑look incidence) as a selling point — use these numbers in onboarding checks.

Aggregators, smart routers and algos

Then: could reduce slippage if routing logic was sophisticated.

Now: routing has bifurcated into two clear classes:

  • Latency‑aware SORs with ML decision layers — these routinely outperform naive routers for intraday and mid‑sized orders by dynamically weighting venue fill probability, historic hidden liquidity hits, and latency trends.
  • Deterministic rule‑based routers — still common on low‑cost platforms and preferable for traders who want repeatable, auditable behavior (important for compliance and backtesting).

Practical takeaway: a router that “learns” may reduce average slippage but can introduce path‑dependent behavior that is harder to replicate in backtests. Ask vendors for out‑of‑sample performance summaries and timestamped routing logs.

Execution patterns by strategy (updated)

  • Scalpers and HFT‑style retail: ECNs remain the first choice for very small, latency‑sensitive slices. But scalpers must now monitor router micro‑decisions and implement hard circuit breakers — ML routers can change routing mid‑day.
  • Intraday momentum traders: Prefer hybrid routing that mixes ECN primary fills with selective internaliser liquidity for size. Use algos that incorporate real‑time depth slope and venue reliability scores rather than static aggressiveness parameters.
  • Swing and carry traders: Passive midpoint limit orders on ECNs or negotiated OTC block execution remains optimal. Consider scheduled limit refreshes tied to liquidity windows instead of always‑on limit posting during low‑liquidity nights.

Event windows and stressed markets: new caveats

Event risk remains the most significant driver of execution divergence. Two 2026‑era trends change the rules:

  • Router fragmentation at events. ML routers can suddenly divert to less obvious pools during stress (e.g., internal crossing or recently added dark pools). That may reduce impact if the router’s model generalizes well — or increase slippage if the model overfits to historical stress seen elsewhere.
  • More explicit last‑look disclosures. Several brokers began publishing last‑look incidence and rejection rates in 2026; traders should request these stats and test behavior during scheduled announcements to quantify asymmetric slippage risk.

Practical checks and measures traders should run (updated)

In addition to the earlier tests, add these 2026 best practices:

  1. Routing transparency test: request a timestamped routing log and verify a sample of fills against venue timestamps. Check that the router’s chosen venue corresponds to the fastest historically reliable fill for your flow.
  2. Last‑look incidence drill: run a small batch of marketable orders during low and high volatility windows and measure rejection/cancellation rates. Compile statistics by broker and instrument.
  3. Out‑of‑sample stress test: ask a venue or vendor to run your algorithm on anonymised historical stress windows (CBs, PMI surprises) and provide aggregated slippage metrics — this is now a standard service at many execution vendors.
  4. Data hygiene and clock sync: ensure your timestamping uses an industry standard (e.g., NTP/PTP) and reconcile clocks with the venues you use; mis‑aligned clocks distort inferred slippage by masking latency as price movement.

Updated takeaways for FX traders (actionable)

  • For very small, latency‑sensitive tickets: continue to prioritise ECNs and colocated access; mandate execution monitoring and hard kill switches for peak news windows.
  • For medium size and intraday strategies: prefer routers that publish venue scoring methodology and provide routing logs. Test ML‑enabled routers out‑of‑sample and keep a deterministic fallback for compliance/backtesting.
  • For large blocks: combine algorithmic slicing with negotiated execution and consider off‑market crossing where permitted; expect to trade off midpoint benchmarks to preserve information leakage and reduce impact.
  • Always demand data: require last‑look incidence, reject rates, and post‑trade slippage reports as part of onboarding. Run your own midpoint and depth‑versus‑size tests at least quarterly and during major scheduled events.

Outlook — what to watch for next

Through the rest of 2026 and into 2027, three developments will matter:

  • Broader vendor adoption of standardised, timestamped routing logs and anonymised cross‑venue fill reporting — this will make apples‑to‑apples execution measurement realistic for the first time.
  • Further refinement of ML models in production routers; expect performance gains but also an emphasis on explainability and audit trails driven by buy‑side demand and regulator expectations.
  • Potential regulatory guidance on last‑look and disclosure in major jurisdictions — traders should monitor their brokers’ published metrics and compliance documentation closely.

Final thought

Execution in November 2026 remains the interaction of venue microstructure, order type, size, time and market regime — now with an added layer of model complexity in routing. The traders who will truly reduce slippage are those who demand data, stress‑test under realistic regimes, keep deterministic fallbacks, and match venue choice to strategy behaviour rather than brand claims.

How do I pick between an ML router and a deterministic router?

ML routers can reduce average slippage when trained and validated on representative data, but they can be path‑dependent and harder to audit. Choose ML routing for intraday or mid‑sized flow if the vendor provides out‑of‑sample performance, routing logs, and a deterministic fallback. For compliance‑sensitive or backtesting‑driven strategies, prefer a deterministic router or ensure the ML system exposes decision metrics.

Should retail traders be worried about last‑look now?

Last‑look remains a factor but transparency improved in 2026: many brokers now publish last‑look rates and rejection statistics. Retail traders should request these figures during onboarding, run small marketable order tests across time zones, and prefer platforms that provide per‑trade execution analytics.

How often should I re‑run my venue slippage tests?

Quarterly baseline testing is a minimum. Re‑run tests after any material change (new router, new broker, or changes in co‑location) and before major economic calendars. Additionally, perform an event‑window test at least semi‑annually to capture changes in last‑look behavior and stress response.