Bukkit

Work/Process

Recent changes: See implementation/changelog

Working on:

• paused / planning 0.3.0

Implemented for 0.3.0

• Few internals.

Planned for Next version: 0.3.0

• Per player files, preparing for database support, making saving single players data lighter + distribute over time.
• Suspicious light level threshold per item id.
• Exclude players from checks (command+file based + maybe by a permission node)
• Maybe: Reworking/extending the browsing functionality to a more generic base allowing for browsing through anything, virtually. If so, maybe also view recently active players, other...
• Maybe: Iff DB support: still support conversions from per player files to/from database + hybrid approach (lazy adaption, either way)
• MaybeMaybe: add suspicious counts for absolute light level, and maybe some abstract facts, and maybe short term count of ratios (adding to abstract facts rather) to track mixing xray use into other stuff.
• MaybeMaybe: Ignore-placed-blocks feature [maybe as an extra plugin or linking to another one, optional]
• MaybeMaybeMaybe: Adding some reset features for unsuspicious or old entries, probably also on placing blocks.

Planned soon:

• testing routines / unit tests for systematic functionality and robustness checks.
• "/dt report -anon" or similar feature to write a file with all data, just names exchanged by "player1, player2, ..." for data exchange.
• Implement another method to determine how long players mine stuff and how much they get from a period of mining. Using time-start+end and number of blocks, then check with some threshold on breaking blocks of that kind. (+configurable +display).
• add documentation for data safety (backup your data file now and then to minimize loss in case of file system trouble).

Planned:

• Detach all data handling completely from the bukkit api, so that it can be testet independently.
• Database support with on demand loading of data portions for individual players.
• (+ other changes/additions from other list)

Interesting:

• Rather than using global average stats, use several basic "stats patterns" which players are assigned to, to distinguish approaches like "cave-mining"
• Support for other plugins like log-block (Though the light level checks won't be in there. But additional on-demand analysis methods could be interesting to find patterns of mining, lateron...).

Methods under review:
(OUTDATED)

• Av. mining speed
  • simple data (n, t-start, t-end), maybe using "blotch"-bounding rects as well.
  • Keep three blotches and unite under certain conditions - prevent: Mining diamond, wait slightly less than time-threshold and then mine one more diamond to get less suspicious stats.
• Discovery tracking: tracking how fast players discvover certain ores (similar to mining speed, just using found ores as reference - real discovery).
• Blotch break/discovery: track how fast groups of diamonds are found (counting diamonds that are not too far apart as a group).
• Interval-partitions: For keeping track of certain types of mined blocks it could be interesting to regard the change per partition rather than the dumb sum of all elements per partition (quite simple). Besides it makes a difference if a block is the only one in the queues or just one of many (only one = the difference would not count concerning the last >0 values for the partitions).
• Simple mining pattern analysis - "shape factor": some average values how often a players is mining with what kind of "shape" (tunnel, vertical horizontal, other).
• Light placement strategy - keep track of players placing light (torch, lava, glowstone, jack-o-lantern, ..) and how it relates to other events like finding ores/tunnel shape
• More specific Analysis how stuff was found: caves, surface, through some tunnels, tunneling straight towards(turning towards?) etc.
• (The most simple and seemingly effective or complementary will be tried.)

Questions:

• It seems that lightlevels get even less interesting with 1.8 not providing total darkness anymore. The ore discovery in darkness might get even easier with altered texture packs (even non transparent). -> Usually the light level is a good indicator for cheats, still. Yet one should not be too quick on judgement, always judge the combination of materials.