Frequently Asked Forum Questions | ||||
Search Older Posts on This Forum: Posts on Current Forum | Archived Posts | ||||
Re: So you guys want specifics on game design eh.. | |
Posted By: ThorsHammer <nate@ishness.net> | Date: 11/16/09 5:14 p.m. |
In Response To: Re: So you guys want specifics on game design eh.. (Hawaiian Pig) I don't have an XBL account, and I never have. Thus, I'm really not qualified to talk about this. I'm going to anyway, because armchair philosophizing is fun sometimes and this thread has managed to suck me into wasting time I really, really don't have. : I'm in agreement with much of what you and Wu said. I understand it's
: In any case, the problem that lies here with mismatches is a result of
: The most obvious solution, to me, would be to keep track of a global rank
I wonder if this is because different playlists emphasize different skills, and their ranking engine isn't very granular. That is, to use your terminology, TrueSkill doesn't know anything about a player's ability to execute, strategize, or perceive displayed gameworld state - I don't know anything about the ranking algorithms, but I doubt they're looking at things that complex. A quick scan of b.net's playlists suggest to me that different playlists do select for specific skillsets and strengths, at least somewhat - some are objective-focused, some are FFA, some are team-based, etc. If all you have is a win/loss ratio (or, possibly a little more data, like an 'ownage' variable representing how strongly they won/lost the match, but still no real information on how a player executes, strategizes, or perceives available information), it'd be hard to have any true insight on how a player performs across the board. Their win/loss record in, say, CTF on maps A, B, and C (I don't play H3 multi so I don't know the map names) doesn't tell you much about how well they'd play Team Slayer on maps I, J, and K. I think we all agree that Bungie is trying to optimize the experience for more casual players, rather than 'hardcore' players. That means, above all, they want to avoid frustrating matchups. Thus, a test that yields a false positive response to "Should we increase their rank?" is very, very bad, from their perspective. Pushing someone to the next level incorrectly could frustrate them badly, and make them want to quit. Total speculation, and I've only been thinking about the problem for about fifteen minutes, so it's likely to be very wrong. Still, it might be worth something. : If this system were to be properly implemented, the only problems that would
: I think the question is: is the problem of new accounts significant enough to
: I guess I don't know the answer to that. I think my hunch about false positives applies to the new account problem, too - otherwise, it'd seem like you could get much closer to solving the new account problem by looking at the hypothetical 'degree-of-ownage' variable after each game, and bumping accounts that passed some threshold value for degree-of-ownage to the next level. Actually, 498 out of 500 games won, even if those numbers aren't exact, really makes me think the TrueSkill ranking algorithms are terrified of false positives. Someone with a win rate of over 90% ought to be rocketing up the rankings until their win rate goes down, to my mind. Obviously you'd need to keep track of their win rate at the current skill level independently of overall win rate, so you don't push them too high too quickly. Usual caveats apply: I've no idea how Bungie's ranking algorithms work, this is all speculation, they have actual data and we do not. I'm making a wild guess based on one number that may or may not be real. That said, it seems to me that the 'new account' problem, while not solvable, could be made less problematic than what I'm hearing discussed in this thread. However, you'd probably have to accept a higher frustration rate among the players who aren't trying to abuse the system, and Bungie probably doesn't want to do that. -Nate (might have a response to HP's original post sometime, which he really enjoyed)
|
|
Replies: |
The HBO Forum Archive is maintained with WebBBS 4.33. |