ADVANCED COMPUTE PROTOCOL // EST. MARCH 2010

High-frequency execution,
rebuilt for the autonomous era.

Hotsino is a proprietary trading firm. Our traders deploy firm's capital across US equities and ETFs on short intraday horizons. Positions are held for minutes or seconds, never overnight. We trade using quantitative microstructure research, optimal-execution theory and modern machine learning to turn raw order flow into disciplined, risk-managed edge.

ACCESS EMPIRICAL MODELS INITIALIZE COMPUTE LAB

MARKETS US EQUITIES & ETFs DATA 1-MIN BARS · L2 BOOK UNIVERSE 500 ACTIVE SYMBOLS HORIZON MINUTES → SECONDS · NO OVERNIGHT

YEARS TRADING
US EQUITIES & ETFs

ACTIVE SYMBOLS
MODELLED DAILY

TRADING-DAY
RESEARCH WINDOW

TECHNICAL INDICATORS
GENETICALLY OPTIMIZED

01 / CONVERGENCE CHRONOLOGY

Sixteen years scaling structural efficiency.

We trade the firm's own capital on short intraday horizons, with positions formed, hedged and unwound end-to-end by our systems — and never held overnight. Risk is evaluated continuously through the session, and every strategy is re-optimized against the most recent market data.

March 2010 · Genesis

ETF Arbitrage Foundation

Founded in the early ETF era — our AMEX "ETF Guru" roots — building analytics for exchange-traded products as the asset class itself was taking shape.

2013 · High-Frequency Deployment

Black-Box Automation

Deployed fully automated, black-box high-frequency trading — commanding a leading share of volume across multiple ETF sectors.

Modern Epoch · Autonomous Stack

AI-Native Agentic Frameworks

Embraced large language models and autonomous agents end-to-end, rebuilding research and execution on a modern, AI-native stack.

The Ecosystem Core

Infrastructure Contribution

We give back to the wider quantitative community through open-source projects such as Yuclaw — sharing tooling and research that benefit the whole industry.

Direct Venue Routing

Direct market access and fast automated execution across liquid US equities and ETFs — orders routed and managed with no manual latency.

Agentic AI Toolkit

A deep, in-house toolkit of LLM- and agent-based models spanning research, signal generation, execution and risk — built and shipped continuously.

Regime Survival

Built to compound through stress. Our traders posted a career-best month in March 2020 — while broad markets hit multi-year lows — thriving where static tools break down.

02 / PHILOSOPHICAL DRIVER

Deep conviction behind every computation.

Beyond processing high-frequency data matrices, we design frameworks intended to leave standard industrial assumptions obsolete.

// CORE MISSION STATEMENT

Technology as a Civilization-Scale Asset.

We do not capture market spreads simply to accumulate capital. We believe that advanced quantitative infrastructure should serve human advancement, driving systemic efficiency and creating open, accessible technologies. We dedicate internal engineering cycles to open-source systems, validating our belief that engineering excellence must elevate the global developer collective.

// LOGISTICAL VISION

Decoding the Pulse of Dynamic Chaos.

The market is a hyper-dimensional, non-linear living organism. Our absolute objective is to interface with this complexity seamlessly, unifying human spatial insight with raw accelerated machine learning. We do not attempt to safely predict the future; we construct a modern engineering fabric capable of dancing within the chaos in real time.

// 01 Autonomy

Self-improving loops running continuously on bare-metal systems, removing localized processing limitations.

// 02 Absolute Proof

No assumptions survive without structural validation. Every predictive matrix is stress-tested back-to-back against granular order-flow history.

// 03 Open Ecosystems

Returning core innovation to the community through open-source tools and shared research.

// 04 Perpetual Search

We reject comfortable steady states. We constantly rebuild our trading layers to preserve high capacity and speed.

03 / MICROSTRUCTURE QUANTIZATION

Isolating order book mechanics at the microscopic bound.

Alpha is not found in broad charts; it exists in sub-second order imbalance, dynamic cancellations, and cross-venue latency mechanics.

QUEUE DYNAMICSSTOIKOV MODIFICATION

The Microprice Estimator

Standard mid-market assessments fail under asymmetric size distribution. We weigh real-time depth vectors to discover latent velocity directions.

$$P_{\text{micro}}=\frac{P^{b}Q^{a}+P^{a}Q^{b}}{Q^{a}+Q^{b}}$$

Application: Eliminates adverse fill vulnerability by adjusting quote entry points dynamic to order book queue skew.

FLOW VELOCITYCONT–KUKANOV

Order-Flow Imbalance (OFI)

Quantifying net structural pressure on the best bid and ask tiers. Tracks high-speed institutional accumulation before price transitions occur.

$$\Delta P_k \approx \beta\,\mathrm{OFI}_k,\qquad \beta\approx\frac{1}{2\,\bar D}$$

Application: Yields immediate direction validation during microsecond order matching bottlenecks.

PRICE IMPACTEMPIRICAL √-LAW

Square-Root Impact Law

Across markets and decades, the cost of working a metaorder scales with the square root of size relative to volume — not linearly.

$$\frac{\Delta P}{P}\approx Y\,\sigma_D\sqrt{\frac{Q}{V_D}}$$

Application: sets realistic slippage budgets and caps order participation. Test it live in the Compute Terminal below.

FLOW TOXICITYEASLEY–LdP–O'HARA

VPIN — Toxicity Gauge

Volume-synchronized probability of informed trading. Rising VPIN preceded the 2010 Flash Crash by hours — an early read on adverse selection.

$$\mathrm{VPIN}=\frac{\sum_{\tau=1}^{n}\lvert V^{S}_{\tau}-V^{B}_{\tau}\rvert}{n\,V}$$

Application: high toxicity triggers wider quotes, reduced size, or stepping aside.

SPREAD & DEPTHROLL · KYLE

Decomposing the Spread

Roll backs out the effective spread from return autocovariance; Kyle's lambda measures depth — price moved per unit of signed flow.

$$S=2\sqrt{-\operatorname{Cov}(\Delta p_t,\Delta p_{t-1})},\quad \Delta P=\lambda\,Q$$

Application: splitting cost into liquidity vs. information drives both quoting and sizing.

RISK SCALINGREALIZED VARIANCE

High-Frequency Volatility

Summing squared intraday returns gives a model-free estimate of integrated variance — the live heartbeat every position is scaled against.

$$\mathrm{RV}=\sum_{i=1}^{n} r_i^{2}\;\xrightarrow{n\to\infty}\;\int_0^{T}\sigma_t^{2}\,dt$$

Application: real-time variance feeds execution scheduling and inventory limits.

04 / STOCHASTIC EXECUTION ARCHITECTURES

Fully formalized mathematical execution arrays.

Our algorithms frame risk mitigation and liquidation sequences as explicit optimization paths solved programmatically against current market thickness.

STOCHASTIC CONTROLALMGREN–CHRISS PROTOCOL

Implementation Shortfall Optimization

Balances the friction of market impact penalties against the inventory decay of prolonged exposure. Creates a crisp exponential trajectory.

$$x_j = X\,\frac{\sinh\!\big(\kappa(T-t_j)\big)}{\sinh(\kappa T)},\qquad \kappa=\sqrt{\frac{\lambda\sigma^{2}}{\eta}}$$

Parameters: $X$ total scale · $\lambda$ systemic risk aversion · $\sigma$ immediate volatility matrix · $\eta$ immediate impact penalty.

DYNAMIC LIQUIDITYAVELLANEDA–STOIKOV ASYMMETRY

Inventory-Aware Quoting Arrays

Quotes track a reservation metric shifting away from market midpoints to automatically counter adverse position accumulation.

$$r(s,t)=s-q\,\gamma\sigma^{2}(T-t)$$

Application: Forces market-making nodes to offset directional toxicity while collecting constant spread premium.

STATISTICAL ARBITRAGEORNSTEIN–UHLENBECK

Mean-Reversion on Spreads

Residual spreads between cointegrated symbols are modelled as mean-reverting; entries trigger when the spread's z-score breaches an optimized band.

$$dX_t=\theta(\mu-X_t)\,dt+\sigma\,dW_t$$

Application: market-neutral carry, largely uncorrelated to broad beta; half-life = ln2 / $\theta$.

ORDER CLUSTERINGHAWKES PROCESS

Self-Exciting Arrivals

Trades and quotes cluster. Modelling the conditional intensity as self-exciting forecasts short-horizon volatility bursts and toxic flow.

$$\lambda(t)=\mu+\sum_{t_i<t}\alpha\,e^{-\beta(t-t_i)}$$

Application: branching ratio $\alpha/\beta \to 1$ flags reflexive, high-impact regimes.

ADAPTIVE HEDGINGKALMAN FILTER

Time-Varying Hedge Ratios

Relationships drift. A Kalman filter tracks the hedge ratio as a hidden state, updating online with each tick instead of refitting a static regression.

$$\beta_t=\beta_{t-1}+w_t,\qquad y_t=\beta_t x_t+v_t$$

Application: hedges stay accurate through regime change, with no lookahead bias.

CAPITAL ALLOCATIONKELLY CRITERION

Growth-Optimal Sizing

Each signal's edge and odds map to a fraction of capital maximizing long-run compounded growth — then de-leveraged for drawdown control.

$$f^{*}=p-\frac{1-p}{b}$$

Application: we deploy a fractional $\tfrac{1}{2}f^{*}$ in production to cap variance.

05 / THE HOTSINO COMPUTE KERNEL

Live mathematical synthesis. Zero interpolation.

Interact with live client-side instances of our operational engines. Adjust hardware stress conditions and infrastructure limits to observe structural responses. SIMULATED AGENT ENVIRONMENT

VECTOR STREAM ENGINE // RSI

OPTIMAL PATH // ALMGREN-CHRISS

NON-LINEAR IMPACT // SQUARE-ROOT

Vector Stream Explorer

Evaluating technical boundary conditions across high-throughput data streams. Adjust the lookback window to recalculate convergence models.

Compute Buffer Depth 14

—Inference Index

—Pipeline State

SYNTHETIC EQUITIES FREQUENCY MATRIX

REAL-TIME INFERENCE STATE (OVERBOUGHT / OVERSOLD BOUNDARIES)

Execution Curve Dispatcher

Modifying computational urgency coefficients shifts the liquidation trajectory from standard linear TWAP formats into defensive fronts.

Urgency Coefficient (λ) 0.30

Noise Factor (σ) 0.25

Impact Imbalance (η) 0.20

—κ Urgency Matrix

—Calculated Half-Life

HOLDING DECAY TRAJECTORY ($x_j$)

DISCRETE CHILD TRADES MATRIX ($n_j$)

Slippage Curvature Analysis

Verifying institutional slippage metrics against continuous empirical scale distributions.

Block Penetration Ratio 5.0%

Underlying Realized Noise 2.0%

Venue Resilience Const 1.00

—Est. Impact Cost (BPS)

BASIS POINT COST SLIPPAGE FUNCTION (SQUARE-ROOT BOUND)

06 / AUTONOMIC INFRASTRUCTURE STACK

From raw capture layers to direct market execution.

Our pipeline automates deep-learning hyperparameter adjustments continuously, rewriting target parameters dynamically at the end of each session.

// 01

Order-Book Deep Topology Layers

High-density convolutional sequence networks tracking structural imbalances.

DEEP-LOB

// 02

Stochastic Optimization Infrastructure

Evolutionary pipelines running parameters through thousands of generations nightly.

GENETIC-EVO

// 03

High-Throughput Re-Verification

Fault-tolerant verification fabrics checking active models against order histories.

SPARK-CORE

// 04

Reinforcement-Learning Execution

Policy networks that learn to minimize slippage under live market impact.

// 05

LLM News & Filings Agents

Parsing headlines, 10-Ks and transcripts into structured, tradeable catalysts.

LLM

// 06

Regime & Drift Detection

Unsupervised clustering of volatility / liquidity regimes; drift-aware retraining.

REGIME

hotsino-core // internal-pipeline-node-04

# running online optimization grid
$ hotsino --initialize-matrix --window 20d
→ ingesting granular sub-second structures
→ populating microprice imbalance matrices
→ processing evolutionary architecture ... optimized
→ state compiled. writing core memory fields.

$ hotsino --deploy-live-agents
[OK] limit order book encoders synchronized
[OK] neural reinforcement execution armed
[WARN] system detecting cross-venue jitter anomalies
[OK] defensive reservation spread widened autonomously
$

07 / SYSTEM CO-DEVELOPMENT

Scale with our capital, natively.

We collaborate with exceptional analytical minds, providing raw hardware acceleration access and liquid assets to support structural edge discovery.

08 / CONTACT PORTAL

Initialize communication channels.

Connect with our engineering desks for structural layout inquiries, technical verification, or algorithmic alignment.

admin@hotsino.com //

Headquarters:
Calgary, Alberta, Canada

admin@hotsino.com

High-frequency execution,rebuilt for the autonomous era.