PredictionManager

PredictionManager provides states, callbacks, and settings to fine tuning prediction for your game type.

Component Settings

Server are settings which only affect the server.

Drop Excessive Replicates will discard replicate datas received from clients and server after the cached count exceeds a certain value. When false multiple datas will be consumed per tick to clear the cache quicker.

• Maximum Server Replicates is shown when the above value is true. This value indicates the maximum number of replicates which the server will cache from client before dropping old ones. Generally, cached replicates will never significantly exceed Queued Inputs.

Settings are general settings related to prediction.

Queued Inputs are the goal number of inputs kept in queue for server and clients. Higher values will create a larger buffer to reduce predicted inputs, but will add more delay to when inputs are run. Lower values will run the inputs faster at the cost of a lower buffer; lower values are more ideal for fast paced games.

Redundancy Count is the number of past inputs replicates to send to the server. Higher values will reduce the likeliness of data not arriving at the cost of bandwidth.

Allow Predicted Spawning lets prefabs and scene objects be setup to use Predicted Spawning. You can use this setting to enable or disable predicted spawning without having to change the settings for every object. This value may also be set at runtime. If changing at runtime be certain to also change on the client; otherwise they could be kicked for trying to use a disabled feature.

• Reserved Object Ids is the number of ObjectIds to reserve per client for predicted spawning. Clients will start out with the specified number of Ids and receive new ones as the server validates their predicted spawning requests. For example: if this value was set to 15 and a client with a 100ms ping sent 3 predicted spawns in one tick then they would have only 12 predicted spawns left to use until the server responded with 3 new Ids, which would be approximately 50ms later.