Object matching method, model training method and server

Abstract

The invention provides an object matching method, device and equipment and a medium. The method comprises the following steps of using a server for acquiring an to-be-matched object set, wherein the to-be-matched object set comprises at least two to-be-matched objects, and each to-be-matched object comprises at least one kind of attribute information; generating M prediction combinations accordingto the to-be-matched object set, wherein each prediction combination comprises at least two queues, and each queue at least comprises one to-be-matched object; acquiring matching probability information corresponding to each prediction combination through matching prediction models, wherein the matching prediction models comprise an outcome prediction model, the outcome prediction model takes theattribute information corresponding to each object in the queues as input and takes the defeat probability as output; determining a target matching result according to the matching probability information corresponding to each prediction combination. With the adoption of the method, the divergence caused by a large strength gap between players can be effectively avoided, and a matching result ofobjects for a group battle can be optimized.

Description

technical field [0001] The present application relates to the technical field of online games, in particular to an object matching method, a model training method, a server and a computer-readable storage medium. Background technique [0002] In recent years, with the rapid development of the e-sports industry, players are paying more and more attention to the fairness of competitive games. Among them, Multiplayer Online Battle Arena (MOBA) games are widely favored by players all over the world. Different from traditional RTS (real-time strategy, real-time strategy) games, MOBA simplifies the roles that players need to operate from buildings, multiple arms, etc. to a single arm - hero, and thus derives a game that focuses on the faction hero arena battle. Popular MOBA games such as Dota, League of Legends, Three Kingdoms, Glory of Kings, etc. [0003] The basic principle of competitive online games is to match players of the same level together, and try to avoid a crushing ...

AI Technical Summary

Problems solved by technology

[0004] Since the ELO matching system based on the ELO algorithm only uses ELO points to describe the real strength of the player from a single dimension, the algorithm error of the ELO algorithm is relatively large, resulting in a discrepancy between the player's points and the player's strength. The real strength of the team is too different

For example, a matching result that often occurs is that the sum of the ELO points of the two teams is similar, but in fact the strength of the two teams is very different. One team is the trumpet of a relatively strong player, while the other team It is the first time for the players to reach the score, which leads to the crushing of the opponent by the small team, which is a typical crushing round

[0005] Therefore, the matching system based on the ELO score algorithm only locates the player's level based on the ELO score, but its positioning is not accurate enough.

Images