Categories of poker bot software

Poker bot software divides into four primary categories defined by where the decision and execution layers reside. The split governs deployment topology, audit surface, and the operator's exposure to host-side detection. This page describes each category in turn, with notes on typical use, components, and trade-offs.

Standalone client agents

A standalone client agent runs on the same host as the poker client and contains both the policy and execution layers within a single process or process group. State is acquired through screen capture, accessibility APIs, or memory inspection of the host client. The agent submits actions through the operating system's input subsystem.

The category is the oldest in the field and remains common among single-account operators and small fleets. Its advantages are deployment simplicity and a short feedback loop between perception and action. Its principal disadvantage is that the entire decision surface is exposed on the same host as the player account, which raises both the technical detection risk and the recovery cost when a host is lost.

Server-side decision engines

A server-side decision engine runs on a central host and exposes an interface, typically a message queue or RPC endpoint, that consumes table state and returns recommended actions. A separate execution layer, often a thin client on the player host, translates the recommendation into a click or a network action against the poker client.

The category dominates larger operator deployments. Its advantages are central updateability, shared model state across accounts, and a smaller per-host footprint. Its disadvantages include sensitivity to network latency, the need for a robust kill-switch on the message path, and operational dependence on the central host's availability.

Hybrid split-policy systems

A hybrid split-policy system places fast-path decisions, such as preflop folds and continuation bets, on the player host while routing complex postflop decisions to a server-side engine. The split is defined by a budget — usually a target round-trip time — that determines which decisions the host may resolve locally.

The category emerged in response to latency and detection trade-offs in the two pure categories. Its advantage is that median decision latency drops to local-process speed while complex hands still benefit from a centrally maintained model. Its disadvantage is operational complexity, since two policy surfaces must be kept consistent and audited as a single system.

Table-overlay analytics

Table-overlay analytics software presents aggregated opponent statistics to a human player as a transparent layer over the poker client. The software does not select or submit actions and therefore sits at the boundary of the bot software category. Overlays are included here because they share perception components — table state extraction, opponent identity tracking — with bot agents, and because some integrator vendors deliver overlays and agents as a single product line.

The advantages of overlays are low regulatory exposure relative to autonomous agents and a clear human-in-the-loop audit trail. The disadvantages are that decision quality depends on the operator and that automation benefits do not scale with seat count.

Comparison

Category labels are descriptive and do not imply endorsement of any specific product or vendor. Boundaries between hybrid and server-side systems are particularly soft in practice, and operators often run mixed fleets.