Bridge Dummy

Summary

The (unfinished) 'Bridge Dummy' platform computes bridge partnership/player performance from actual game play.
The Bridge Deals Analysis website provides successful prototype results.

Introduction

At its core, the innovative 'Bridge Dummy' platform performs automated doubledummy analysis, revealing 'perfect' (i.e. open cards) bidding and declarer/defender play. It compares actual play with this data, quantifying all par deviations. This results in skill/luck performance statistics for all involved partnerships/players, accumulated over all played boards. The generated information can be used for retrieval (boards history), learning (guided singledummy analysis), training (bidding/playing precomputed boards), partnership/player assessment (performance statistics), partner selection (profile matching), etc.

As a recurring caveat: (computable) doubledummy decisions are strictly inferior to (unfortunately incomputable) singledummy decisions. Therefore, this platform by definition generates a presumed small amount of false positives (i.e. wrong par deviations/results), but never false negatives (i.e. missed par deviations/results). False positives might be eliminated via manual singledummy analysis/annotation, i.e. human intervention.

Irrespective of unavoidable theoretical flaws, this platform definitely enables bridge partnership/player skills improvement (as already 'proven' by below prototype usage), ultimately increasing the overall enjoyment of the game.

Early Prototype

As an early prototype, the automatically generated Bridge Deals Analysis website covers all OKbridge deals played by 'ericbaes' (early 2001 to early 2012). This prototype is fairly self-explaining, but nevertheless provides 'Help' links to support correct results interpretation. The 'full platform' idea (i.e. using 'duplicate play' data to enable a generic service towards the bridge community at large, adding to the personal postmortem analysis support) emerged years later.

Platform Design

The following paragraphs summarize some early platform design, which was put on hold indefinitely (due to a lack of further resources, mainly time). A significant part of below listed topics has already been deeply fleshed out in the early prototype. Anybody wanting to take a shot at full platform implementation might want to contact the author for further details and/or additional inspiration.

Functionality

System Components

Data Model

Algorithms