跳到正文
DEV Community· Ahmad Alsulaiman·· 20 小時前AI 評分27

Project500:105 組排名賽揭示目標優先控制器的表現

Project500: what 105 ranked sets showed about an objective-first controller

AI 導讀

Project500 在 November Cup 的 105 組排名賽快照中錄得 46 勝、44 負、15 和,評分為 1239。其公開排行榜評分為 1334,涵蓋 61 組賽事;兩個排行榜分開計算,數字不能合併成單一勝率。v1-objectives 使用地形流場規劃路線,並派遣 APC 爭奪目標。

正文

Ahmad Alsulaiman

Disclosure: This post was written with Codex and is intended for the Battle.training Bot Bounty, which may pay up to $15 if accepted.

Project500 is a JavaScript controller entered in the November Cup. Its first version, v1-objectives, uses terrain flow fields for routing and assigns the APC to contest objectives.

What the ladder showed

At the 105-set snapshot, Project500 recorded 46 wins, 44 losses, and 15 draws in the Cup, with a rating of 1239. Its public-ladder rating at that snapshot was 1334 across 61 sets. These are separate leaderboards, so the numbers should not be combined into a single win rate.

A replay to inspect

In this ranked set, Project500 won 2–0. The first game ended with the hill locked. The replay makes it possible to inspect which units reached the objective and when, while the standings show how the controller performed across opponents. For the next iteration, I would compare objective arrival times in wins and losses before changing the routing weights. One replay can generate a hypothesis, but it cannot establish which part of the controller caused a win.

Public bot record and ranked-set data.

來源:DEV Community · dev.to