When seeding a server, how much would you like those rounds to affect your hive skill?
moultano
Creator of ns_shiva. Join Date: 2002-12-14 Member: 10806Members, NS1 Playtester, Contributor, Constellation, NS2 Playtester, Squad Five Blue, Reinforced - Shadow, WC 2013 - Gold, NS2 Community Developer, Pistachionauts
Comments
So option B-and-a-half
Also it would be great to have a separate vote for "reset seeding game" (for the games that started without full teams) which could be started also later in the game ("reset" vote can only be started for some short time in the beginning of the round, right?).
Option C is assuming an Option A starting point but veering into relevancy later on.
It seems like the polling questions aren't accounting for the overlap/confusion, and aren't specific enough.
I think that Option A makes sense - if there's not enough players for a real game, then why treat it like one?
Option C can work only if a specific time is mentioned - are there 12 players by the 2 minute mark o the 15 minute mark? There's quite a difference there.. and its why Option A is going to win imo.
Ideally the best option would be to have two seperate hive skills.. One for whitelisted servers, and a seperate one for all other servers.
This lets people who play on non-whitelisted servers still benefit from hive and shuffle, without affecting whitelisted servers.
Otherwise you get some people mad that some servers are whitelisted while theirs aren't and it starts the whole "uwe specifically hates our server" nonsense up again.
So basically like ranked/unranked.. Except both would have their own ranking so call it something like normal/arcade instead.
weighted_time = a^( - join_time) - a^( - quit_time) / a^( - round_start) - a^( - round_end).
Where a is a parameter that controls the balance between the beginning and the end of the round, the same parameter that is used in the rest of the skill system.
We sum the weighted times for the human players on each team, and then scale the weighting of the round by
(min( sum_weighted_times, 12 ) / 12) ^ 4
So if there are 10 human players for the whole match, the round will get scaled down by (10/12)^4 = 0.48. If there are 8 human players for the whole match, the round will get scaled down by 0.20. (6/12)^4 = 0.062 and so forth.
The reason for doing this with a smooth function to scale it down is that there are a lot of edge cases otherwise. Consider the case where a full server of players is screwing around in the ready room. 4 players join teams and start the game, then 30 seconds later or so, the rest of the server joins up. We don't want to handle that with some arbitrary threshold.
Any match that wasn't started with 12 players should be not be rated, in which case the server should spit a message saying just that.
If there was the case where 4 players started a match by themselves with the rest screwing around in the ready room, then this match should not be rated, until elo shuffle or restart is voted.
You don't need some arbitrary threshhold for joining, because anything that starts with less than 12 players is already bullshit, no matter if it's 30, 20 or 5 seconds into the game until you have 12 players joined up.
There has to be some grace period at the beginning of the game right? Otherwise we'd be discarding almost every round that starts right after map change as half the players load slower and join late. I'm thinking of making the exponent 5 instead of 4 so 5v5s get ~1/3 credit instead of ~1/2. Does that sound better?
No. People won't rush into the comm chair to start the game. Are we even playing the same game?
I see people do that all the time. Expecting people to coordinate to keep something bad from happening is just asking for unnecessary rage. NS2 really doesn't need more reasons for people to yell at the guy who did something reasonable but wrong. It also doesn't need more reasons for people to delay getting in the command chair.
In my expririence, this scenario of 4 players starting a match that gets played out almost never happens anyway. Not only do you rarely get two commanders right after the map change, but the players will also vote shuffle 10 times before an actual match starts.
Just let the players know, when the game will not be rated. Then they will vote shuffle, which resets the round, anyway.
Whatever. I don't care. Contrary to your imagination, this is a very small subset of games anyway. Its the same few people seeding and people won't just start the game.
And your thinking that games should just start as soon as possible is just sad and contrary to the objective of the skill system. What is the point of having a skill value if you don't even try to start games by it? To whine and mope about late joiners ruining the game? To complain about stacked teams that you can do nothing about? To stroke your ego?
If it's a small subset of games, then this decision doesn't matter either way. I don't think games should start ASAP. Ideally people would wait for enough players to join and vote for even teams. The more games we get though, the better hive works, so it's important to be able to do something reasonable with less than ideal games, so long as they aren't total noise.
A ) matches that start and end with 10 players are not representative for your skill at all. Not by a fraction of 1/2, 1/3 or less, NO!, it's absolutele 0 representative, because this game does not work with anything below 6vs6 AT ALL.
B ) matches that start with <12 players and then fill up are also not representative, because you need 5 field marines to properly build, lane block and push from the very start. Otherwise the aliens will always get a head start.
So lets assume that this is all true and try to encode it as policy. If there's a 5v5 and two more players join at the 1 minute mark, you are saying we should still discard that game. But presumably, if the game starts as a 5v5 and two players join immediately, it's identical to a 6v6 so we should keep that game. Somewhere between those two points, there's a crossover where we should count the game or not. Can you describe where you would want that crossover point to be? Are you ok with it being continous instead of binary?
The crossover point is the third to first person zoom in at the start. So, no I'm not okay with it being continous. You either start the match with a supported player count or you don't get rated for it.
Moreover, since we're talking about continuous weighing, why not have hive impact lessen as player count increases from 16, since 8v8 is supposedly what devs "truly" balance for? Instead of an "abitrary" cutoff at 24?
edit: This may be outside the scope of new hive formula, but I'll mention it. Can hive decide to accept or reject rounds, each round, accounting for mods? For example, instead of the blanket server whitelist we have now, have a mod whitelist. Mods on this list do not significantly affect gameplay and so rounds with them can be recorded by hive. This way you can temporarily put comp mod or beta or 20 hydras per gorge on a server (for example), and rightfully discriminate against those rounds, but not throw out perfectly respectable modless data. (Conceivably, whitelisting mods would be easier than whitelisting servers, since presumably whitelisting a server already means checking out each mod it uses.)
That's actually how it works roughly. When the game is smaller it is assumed that each player has a larger impact on the game so the updates to your skill are correspondingly higher. Without the change we're discussing in this thread, the update size is inversely proportional to the number of players.