Signest · Strategy modelling & simulation
Design the model, set the rules, watch it run in simulation
Signest is a strategy modelling and simulation platform for markets worldwide, running entirely on your own machines — data, modelling, computation and simulation stay in one local loop. It parses the feed you connect into your own database, lets you define factors, instrument-pool models and position rules without writing code, and runs them in simulated accounts against that live feed. We keep a deliberate neutrality: we ship development tools and model templates, never strategies, signals or recommendations. The views, and the decisions, remain yours.
Signest supplies no strategies, no stock recommendations, no signals and no return guarantee, gives no investment advice and never holds client funds — see the Disclaimer.
Illustrative view of derived indicators · not a live quote
Three ideas the whole thing is built on
Complexity belongs to the system
The hard parts are absorbed by the platform, so what stands in front of you is simple enough to master quickly — and you become fluent in strategy design rather than in tooling.
Built from components
A component-based architecture keeps strategies rigorous, rules flexible and computation efficient. Open and extensible by construction, so a new scenario means new blocks, not a new start.
One closed loop
Data modelling, strategy definition and simulated execution form a single loop that runs in real time — which is what makes a genuinely complex strategy something you can hold in your hands.
The core technology
Six things that are genuinely ours, and the reason the rest of the platform can be as simple as it is.
Purpose-built algorithms
The strategy engine (TacticsEngine), rule engine (RuleEngine), execution engine (TradeEngine) and task engine (TaskEngine) all rest on a substantial body of mathematical and statistical work, which is what removes both the technical barrier and the efficiency ceiling — and lets a definition change as often as your thinking does.
Component-based modular design
Every capability is componentised and standardised into reusable building blocks, so an entire modelling architecture is assembled the way you assemble bricks. Open and extensible by construction, which is what lets it fit scenarios we never anticipated.
Plug in and go
Market-data and strategy databases are linked with a single step, with no code to write, and the result is a factor library ready to use. Standardised data interfaces make connecting to systems you already run a matter of configuration rather than integration work.
Serious computational performance
A memory-management scheme of our own, with computation held entirely in memory, lifts throughput by orders of magnitude. High availability for long unattended runs, microsecond-level execution, memory footprint held under 10 MB, and throughput that does not fall over on a busy day.
No programming language to learn
Strategy development is code-free. The metadata layer is open and extensible: define once, use across many scenarios. Advanced users who know SQL can go further, but nobody has to.
Model building that stays simple
Models are built visually. In the great majority of cases, selecting indicators and setting rules and conditions is the whole job — and what comes out is a professional model, not a toy.
An enterprise-grade strategy system. Written in C++ on a microservice architecture, with execution efficiency tens of times that of platforms built in Java or Python. Works with Oracle, SQL Server, MySQL and PostgreSQL out of the box. The system itself is tiny, carries no dependency packages, and runs straight out of the archive.
Four engines, one closed loop
Research, modelling, strategy definition and simulated execution are parts of one system rather than four disconnected tools — which is what keeps a definition from drifting between the place it is designed and the place it is run.
Turns definitions into a strategy set
Factors, stock pools, index linkage and account-level settings combine into strategy sets. Entry-side and exit-side logic are defined separately, so exits are never an afterthought. Many strategy sets run side by side on one machine.
The cockpit over everything
Position coefficients, behaviour switches, weighting and priority, frequency limits, order-book rules and per-account constraints. Parameters can be changed while strategies keep running.
Rule output becomes simulated fills
Inside a simulated account, rule output is turned into simulated orders and matched against live market data — including batching, cancellation and sizing informed by the visible order book — so you see what a rule would actually have done, tick by tick.
Everything that runs on a schedule
Stock-pool refreshes intraday and post-close, factor recalculation, market-data collection, quality checks and backfills, report generation — each a scheduled job with its own execution record.
You define it, the platform computes it
Nothing below is a black box handed to you: every object is something you create, inspect and change.
Factor library
Any fundamental field, any bar period, any indicator and any rule can become a factor. Index-level and stock-level libraries are evaluated together, entry and exit factor sets stay separate, and factor libraries you maintain elsewhere can be layered on top.
Stock-pool models
Freeze your selection logic into a model and let the task engine refresh the resulting pool intraday or after the close. Per-name weights and priorities belong to the pool, so a change in the model propagates by itself.
Position and risk rules
Position coefficients that respond to index state, automatic scaling in and out, per-account limits, frequency caps, and switches that stop one strategy without stopping the system.
Simulated accounts
Create an account with the starting capital you choose and run it on live market data with the same strategy sets, factor libraries and computation model you designed. A model earns confidence by being watched, not by being believed.
Engineered for the trading day
Figures are design targets measured on our own test benches; your hardware, database and strategy complexity decide what you actually see.
The data layer underneath
A platform is only as good as the data it computes on. Signest does not supply data — it parses the feed you connect, stores it properly, and derives the quantities a terminal screen will never show you.
Snapshots with the depth book, minute bars, unadjusted daily bars with adjustment factors, ticks with trade direction, the full pre-open auction path, and sector and industry definitions — written into your own database and turned into derived indicators such as sector relative strength, market breadth, price-limit state and active buy/sell net flow. See the data layer in detail →
What Signest does not do
Worth stating plainly, because this is the part people most often assume wrongly.
No strategies, no picks, no signals
Signest ships model templates and an empty factor library. It supplies no proprietary strategy, recommends no securities, publishes no trading signals and does not tell you what to buy or when.
No advice, no money, no promises
We are not a licensed corporation and not a broker. We give no investment advice, manage no accounts for anyone, never receive or hold client funds, and make no claim about returns. Simulation results are not a forecast — markets do not repeat themselves to order.
How it is delivered
Server, desktop client and data API are included in one subscription.
Server
The engines, the data collection and the scheduled jobs. Runs on your own machine or cloud host, writing to your own database. No dependency packages — unzip and run.
Desktop client
Where models, factors, pools and rules are designed and monitored, and where the cockpit lives. Connects to your own server instance.
Data API
Reads market data and derived indicators as JSON, for your own research scripts, backtests and dashboards.
Subscriptions are HK$899 per month or HK$8,990 per year, billed in Hong Kong dollars. After payment we email the software and a licence key, normally within 24 hours.