Live Execution · Binance exchange connectivity Active
// Welcome to Fib16 Limited
PRECISION EXECUTION.
DEMOCRATISED.
Fib16 is an algorithmic trading firm. We build and operate our own trading systems, designed to execute with precision across cryptocurrency and futures markets. execution infrastructure, ML-driven signal generation and co-located deployment give sophisticated investors access to technology previously available only to the largest proprietary desks.
Our Products
Crypton · Cryptocurrency
A highly optimised trading system designed to capture micro-alpha through rapid, low-latency execution. ML-driven signal generation, sub-2.5s MicroScalp hold times, velocity-based exits and per-symbol configuration. Targeting the most liquid cryptocurrency instruments.
Live on Binance Production
Binanceexchange connectivityML-Driven<2.5s HoldAWS Tokyo
Terrax · Futures
A futures trading system built on the same platform as Crypton. Like Crypton, connectivity is handled by a thin venue-specific layer — the execution engine, machine learning and risk framework are shared across both products. Currently integrated with Interactive Brokers.
90% Developed
FuturesIBKRAWS US-EastShared Platform
LIVE
Crypton · Binance
FIX
API Protocol
6+2
ML Signal Heads
<2.5s
MicroScalp Hold
90%
Terrax Dev Progress
Ultra-Low
Latency
Under The Hood
Built to Institutional Standard
Both Crypton and Terrax share a common execution philosophy: performance-optimised execution in the critical path, machine learning inference at every decision point, and continuous retraining on live market data. Adding a new market requires only a new connectivity layer — not a new system.
fib16_core · live
// Active systems
system[0] : CRYPTON · BinanceFIX
system[1] : TERRAX · IBKR
// Crypton live state
mode : MicroScalp
ml_signal : 0.847 LONG
gate_ok : true ✓
// Terrax state
instrument : MES
adapter : IBKR · FIX eval
▸ shared platform core active
// About Fib16
Fib16 Limited
We trade using two proprietary systems — Crypton and Terrax — each targeting a specific market and instrument class. Both are built in-house, from execution engine to signal generation.storically priced out of it. Two products, one core architecture, one founding vision.
"Institutional execution quality for a market segment historically priced out of it."
— Fib16
Fib16 Limited
Founding Team
Experience40 years
DomainSystems / ML / HFT
ProductsCrypton · Terrax
Target LiveQ3 2026
The Founder
Fib16 brings forty years of deep technical experience spanning systems architecture, high-performance systems engineering and applied machine learning to the construction of Fib16's product suite. This depth of experience allows the systems to be engineered from first principles rather than assembled from off-the-shelf components.
Both Crypton and Terrax were built to address a simple problem: the infrastructure required for high-frequency trading has historically been accessible only to the largest institutions. Fib16 is the answer to that.
Two Products, One Platform
Crypton trades cryptocurrency markets and is fully live. Terrax trades futures contracts and is in final development ahead of go-live.res via Interactive Brokers, reaching 90% completion and operating on the same core engine and ML pipeline. Both products share the same platform — adding a new venue means only a new connectivity layer. Signal generation, risk management and execution logic remain shared across all markets.
The Vision
Fib16 Trading is the group's dedicated trading entity, targeting go-live in 2026 across both products.
Parent Entity
Fib16 Limited
Trading Entity
Fib16 Trading
Target Go-Live
Q3 2026
Products
Crypton · Terrax
Infrastructure
AWS Tokyo · IBKR
// Technology Platform
One Platform. Two Products.
Crypton and Terrax are built on a shared foundation — the same execution engine, the same machine learning infrastructure, and the same risk framework. A thin, venue-specific adapter layer is the only component that differs between products.
// Shared Infrastructure
01 — EXECUTION ENGINE
Execution Engine
Purpose-built for speed and reliability. No off-the-shelf frameworks, no unnecessary abstraction. A thin, venue-specific adapter layer handles connectivity and order translation — all upstream logic is shared across both products. Config-driven per-symbol execution profiles without code changes.
Every trading decision is driven by machine learning. A six-head MLP ensemble scores directional signals across multiple feature dimensions simultaneously. A TCN processes tick-level sequences for precise entry timing. A dedicated Gate 3 MLP filters signals in real time, blocking low-quality setups before execution. All models are trained per instrument and validated continuously against live data.
6× MLPTCNGate 3 MLPPer-Instrument
03 — MARKET DATA
Market Data Analysis
Market depth, trade flow and price momentum are computed continuously across all active instruments. Order flow imbalance, trade rate and velocity signals feed directly into the ML inference pipeline. Low-activity symbols are flagged and deprioritised automatically.
L2 Order BookTrade FlowTick AnalysisReal-Time
04 — TRAINING INFRASTRUCTURE
Training Infrastructure
Models are trained across a multi-server GPU cluster, each node powered by NVIDIA RTX PRO 5000 Blackwell GPUs — purpose-built for large-scale ML training workloads. Training data spans billions of data points across multiple distinct market regimes, ensuring the models have been tested against real market stress, not just a single market environment.
NVIDIA RTX PRO 5000BlackwellMulti-ServerMulti-Regime
01
Exit Logic
Velocity-Based Exit
Exit parameters are tuned independently per instrument — velocity lookback window, peak profit lookback and drop threshold are all configurable. Drop-from-peak protection triggers as soon as a position begins to give back gains, regardless of current price direction.
02
Risk Gates
Rolling Sharpe/Sortino
20-trade rolling window. Both Sharpe and Sortino must clear 0.0 before any new entry per symbol. Gate failures suspend that symbol independently — no cross-symbol cascade.
03
Infrastructure
Low-Latency Infrastructure
Both products are deployed in cloud regions co-located close to their respective exchanges. NVIDIA RTX PRO 6000 GPU drives TensorRT inference on each node, with dataset processing pipelines optimised for high-volume tick data.
// Execution Profiles
Each instrument trades under a profile tuned to its liquidity and behaviour. Parameters are configured per symbol and adjusted as markets change.
// Risk Controls
Performance is assessed on a rolling basis for each instrument. Underperforming symbols are suspended automatically without affecting others. Dynamic exit logic monitors signal state and closes positions before losses compound.
Crypton · Cryptocurrency
CRYPTON
Cryptocurrency · Live
Live on Binance Production Infrastructure
01 — CONNECTIVITY
Exchange Connectivity
Crypton connects directly to cryptocurrency exchanges via high-performance protocols. Multiple venues are supported simultaneously through the exchange-agnostic adapter layer, with each venue adding only a thin connectivity component on top of the shared execution core.
Targeting the most liquid cryptocurrency instruments across major venues. Instruments are selected and prioritised based on tick cadence, spread characteristics and signal quality. Low-activity symbols are deprioritised dynamically.
SpotPerpetualsLiquid InstrumentsAWS Tokyo
03 — TRAINING DATA
Training Data
Models are trained on over 7 billion data points sourced from live Binance tick data, spanning seven distinct market regimes — including bull trends, ATH conditions, sharp corrections, ranging markets and recovery periods. This breadth of coverage ensures the models have been stress-tested across real market conditions, not just a single environment.
7B+ Data Points7 RegimesLive Binance Data
04 — DEPLOYMENT STATUS
Live in Production
Crypton is fully live on Binance production infrastructure. Crypto.com REST and WebSocket connectivity is operational; FIX integration is in progress. ByBit REST and WebSocket support is active. All order flow runs through the high-performance execution path — no REST polling in the critical path.
Binance · LiveCrypto.com · LiveByBit · LiveAWS Tokyo
Terrax · Futures
TERRAX
Futures · In Development
90% Complete · Go-Live 2026
01 — CONNECTIVITY
Exchange Connectivity
Terrax connects via Interactive Brokers, providing access to CME futures and micro-futures contracts. The system currently operates via TWS; native FIX connectivity is in development, which will align Terrax's execution model with Crypton's direct market access approach.
Targeting liquid futures and micro-futures contracts — instruments that combine institutional-quality execution with accessible notional sizes. The same machine learning signal generation and risk framework as Crypton operates on futures tick data.
FuturesMicro-FuturesCMEAWS US-East
03 — TRAINING DATA
Training Data
Models are trained on extensive historical futures tick data sourced from Databento in native DBN format, covering a broad range of market conditions across CME futures instruments. Training data volume and regime breakdown will be published ahead of go-live.
DatabentoDBN FormatCME HistoricalMulti-Regime
04 — DEVELOPMENT STATUS
90% Complete
Core engine, ML models, risk architecture and IBKR adapter are fully operational. Remaining work focuses on production hardening, live performance evaluation and FIX connectivity integration. Target delivery alongside Crypton investor go-live in 2026.
90% CompleteGo-Live 2026TWS Operational
// Why One Platform?
Building on a shared platform means Fib16 can target new markets without rebuilding from scratch. Each new venue requires only a new connectivity layer — signal generation, risk management and execution logic remain shared across all products.
// News & Updates
Latest from Fib16
Development milestones, technology updates and market commentary from Fib16 and the Fib16 team.
Milestone · May 2026
Crypton exchange connectivity Fully Operational on Binance Production
Fib16's Crypton system has achieved a major development milestone: exchange connectivity 4.2 integration is fully operational on Binance's production trading environment. Order submission, acknowledgement and full lifecycle management are confirmed functional across all target instruments.
Terrax futures system completes core development. The shared platform architecture — used by both Crypton and Terrax — is validated against IBKR integration, with additional venue connectivity under evaluation.
Why Infrastructure Proximity Matters in Algorithmic Trading
Fib16's production deployment targets AWS ap-northeast-1. We outline the latency architecture decisions and why physical proximity to exchange infrastructure is non-negotiable for HFT.
Engineering Efficiency: How We Process Large-Scale Market Data
Complete rewrite of ML training dataset builders using pandas and numpy reduced construction time across 14GB of tick log archives from ~18 hours to under 2 minutes without loss of data fidelity.
Risk Controls: How Fib16 Protects Against Underperforming Markets
Both Crypton and Terrax now require 20-trade rolling Sharpe and Sortino ratios to clear minimum thresholds before any new position. The implementation, rationale and edge cases are detailed here.
How Fib16 Uses Machine Learning to Make Trading Decisions
A technical overview of the Temporal Convolutional Network entry model — architecture choices, receptive field design (SEQ_LEN=20, RF=31) and ONNX export strategy for production inference.
Fib16 is preparing for go-live across both products. Enquiries from qualified counterparties are welcome via the contact page.ied investors, institutional counterparties and technology partners are welcome.
Fib16 Limited and Fib16 Trading are in active development. Nothing on this site constitutes an offer or solicitation to invest. Past backtested performance is not indicative of future results. Algorithmic trading involves significant risk of loss.
// Send Enquiry
By submitting you consent to Fib16 Limited retaining your contact details for the purpose of responding to your enquiry.
