SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

Developer builds ML-powered Mancala bot to solve game's player-one bias problem

0
·2 views

A developer has built a machine learning bot for Mancala as part of a personal traditional games website aimed at older players, a project previously shelved due to time and skill constraints. The site uses a custom variant of Mancala with randomised stone placement to reduce the game's inherent player-one bias, which caused roughly 75% of tournament matches to be decided before play began. To train the bot, a Python-based game simulator using NumPy was built, allowing a neural network to repeatedly play against itself and improve over time. A 15% randomiser was incorporated into training so the bot occasionally makes non-optimal moves, exposing it to a wider variety of board positions. The developer credits agentic coding tools with finally making the project feasible by combining improved personal skills with AI-assisted development.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

.NET 10 Rejects Dual-Size InlineArray Types, Triggering TypeLoadException at Runtime

A compatibility change in .NET 10 causes a TypeLoadException when the runtime loads a value type that combines InlineArrayAttribute with StructLayoutAttribute.Size, even if the code compiled without errors. The failure can appear late — triggered by reflection, interop, or serialization — rather than at compile time, making it easy to miss during initial testing. This happens because the two attributes create competing, ambiguous descriptions of the type's memory layout, which .NET 10 now explicitly rejects. Earlier runtimes permitted this combination through implementation-specific behavior, but Microsoft has documented the stricter enforcement as a binary compatibility change in .NET 10. The recommended fix is not to blindly remove the explicit size, but to understand its original interop intent and migrate the layout definition to an unambiguous wrapper struct.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Markdown Gatekeeper tool prevents AI agents from acting on outdated project docs

A new open-source tool called Markdown Gatekeeper aims to solve a common problem in long-running software repositories where outdated documentation can mislead AI coding agents. When tools like Codex or Claude operate across multiple sessions, old architecture notes can be mistakenly treated as current plans. The tool enforces a single authoritative revision per topic, keeping competing documents marked as proposals and superseded files still accessible for reference. Markdown Gatekeeper works locally using plain Markdown and Git, requiring no cloud account or external service. It is available to install via npm from the nanlogic GitHub repository.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Simple Failure Classifier After 48 Hours of Misdiagnosing AI Errors

A developer running a 48-hour experiment with a free AI model endpoint discovered that misclassifying failures was a bigger problem than the failures themselves. Over two days, they logged 21 failure events — defined as final attempts that failed after three retries — from a Python worker making timed calls to a free model endpoint. Manual notes proved unreliable, with the same symptom labeled differently within a single day, prompting the developer to build a deterministic script to sort failures into three categories: code issues, model or API problems, and server-side transients. The classifier revealed that nearly half the events had been mislabeled in the original notes, with seven being harmless server blips already handled by the retry loop. The key takeaway is that establishing fault before attempting a fix can prevent wasted effort on code that was never broken.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Out-of-Order WebSocket Stock Ticks Can Break Trading Logic — Here Is How to Fix It

Developers building real-time US stock market applications over WebSocket connections often encounter out-of-order tick data in live environments, even when local testing appears flawless. The root cause is typically cross-border network jitter, variable latency, and client-side processing pressure rather than a faulty API provider. Relying on local message-arrival time instead of the embedded market-event timestamp is a key mistake that leads to broken charts, false trading signals, and corrupted datasets. Engineers are advised to implement a short client-side buffer that collects incoming tick events, sorts them by their original event timestamp, and only then forwards them for processing. The appropriate correction strategy depends on the use case — lightweight buffering suits UI dashboards, while stricter chronological validation and filtering are essential for quantitative trading engines.

0 comments Read more at DEV Community