For operators and technical teams

We treat gambling as an engineering problem

Models and automation for poker and betting. Specialist data. Private compute for your AI workloads.

Cost modelIllustrative example

When costs erase the edge.

Net expected value

−0.08%

of stake
per trial

Win probability
52%
Win commission
4%
Per-trial cost
2%

Assumed inputs; no pushes; net win payout 1 × stake. Commission applies to win profit; the per-trial cost is a share of stake. This is a calculation, not trading performance.

Change the assumptions
Quantitative engineeringModels Data Compute

Capabilities

Built around the problem.

From a game decision to the data and compute behind it. Start with the output you need.

01 / Models & automation

Quantitative systems

We turn game rules, hand histories and prices into decision models, simulators and automation for the agreed environment.

Explore quantitative systems
Execution / worked example

A quoted margin depends on the fill.

Both bets accepted
+£10Either outcome
One bet partly accepted
−£56If outcome B wins

Synthetic example: £100 on each of two exhaustive outcomes at 2.10, with no fees. If B accepts only £40, the position changes.

Follow the partial fill

02 / Data

Specialist data

Datasets, licensing and processing with field definitions, source rights and quality checks attached to delivery.

Synthetic price observation · selected fields

selection
A
decimal_odds
2.10
available_stake_gbp
100.00
Inspect the sample data

03 / Compute

Private AI infrastructure

Compute rental for your models, with hardware, runtime and access arrangements scoped to the workload.

Illustrative access path

  1. Client
  2. Authenticated endpoint
  3. Runtime & storage

Administrator access and external connections are part of the specification.

Inspect the access boundaries

Private work.
Visible methods.

We keep client identities and project details out of public case studies. Our public examples explain the assumptions, calculations and limits.

About Graywater

From question to usable output.

How we work
  1. Define the decision

    Set the outcome, available inputs and operating limits.

  2. Test the assumptions

    Compare a baseline, costs and the cases that could fail.

  3. Deliver for use

    Agree acceptance checks, handover and operating responsibilities.

Technical notes

Inspect the reasoning.

Worked examples of costs, execution and private model hosting. Assumptions included.

Arbitrage execution

A synthetic £10 quoted margin becomes a £56 loss in one outcome after a partial fill. Follow the accepted stakes and hedge.

Expected value after costs

The calculator’s win, push and loss model: fee definitions, break-even probability and the limits of independent-trial risk.

Private model hosting

Trace a sample request through the endpoint, administrator access, storage, logs and outbound connections.

Start with the outcome

Tell us the problem.

You do not need a finished specification. Describe the decision or output you need, and what you have to work with.

Discuss a project hello@graywater.uk