Please turn off split command or make it a vote.

Please PLEASE PLEASE either disable split command, or make it a vote.
It is actively killing lobbies constantly.

I loaded up the game today, joined the room which was running.
Started watching some videos in a window over the lobby.
Waited for 30min for the game to end.
Notice that the game has ended and click on the lobby to find I am now in a room with 2 other people.
Switch back to the actual lobby which has now started the game.
Well OK then guess I go watch videos for another HOUR.

This is NOT a good user experience when anyone can split the lobby instantly.
It has been happening every night now with a big lobby that there are splits.
All that happens is everyone leaves one room and joins back to the other.

The other outcome is that the high elo lobby instantly starts to stop low elo getting in. The low elo lobby then plays 1 game where a bunch of not quite high enough elo players have to carry the team in an excruciating match. The game ends and everyone then leaves that lobby. If the two games take a different time to finish then now players might need to wait another 10-40min for a game.

Counter proposal, limit lobby sizes. Then you have more matches going at the same time and you don't have to wait so much in the first place.

I get the distinct impression that the devs/mods only implemented this as an experiment as a means to collect data.

So, an alternative: instead of a vote for splitting the channel, have a vote for splitting OR to remove the roomsize limit for the next match. Giving us the option between the two is a great way to force a specific kind of data collection that's not being done -- the sort that actually answers questions that both devs and players are interested in answering.

quote: I get the distinct impression that the devs/mods only implemented this as an experiment as a means to collect data.

I would advise asking about the intentions of the developers rather than making assumptions about them.

quote: actually answers questions that both devs and players are interested in answering.

I, for one, am very uninterested in finding out the inevitable conclusion that inertia and each individual choosing the least-risk-in-short-term option will lead to a 25v25 that will lag out after ten minutes.

I'll defer that question to GoogleFrog.

Some sort of split seems to be necessary. Without it, we saw games where 45 people were trying to play a 32 player game. The extra room required for these players does not make itself, with good reason. Seeding a new room is very risky as it needs a critical mass for actual games to happen.

The goal is for the single room to be able to smoothly turn into two rooms when the player count swells, then smoothly turn back into one room in the off hours. Otherwise, the peak core teams playerbase will be limited to the size of the teams room, unless some sudden influx gets us past the awkward zone in which seeding a second room is too risky. I just don't think the teams playerbase is going to grow if it clusters into one room. There is some point where people look at the waiting list, and give up.

Various forms of making new rooms were tried in the past, but they all failed in similar ways. The waiting list was added about a year or so ago to let a split system distinguish between true spectators and players who want to play. This meant the system could split when there were 40 potential players, rather than just 32.

The new split seemed to be working for a while. The splits are tracked and I've analysed the majority that happened since the system was added six months ago, and at least half seemed to result in more games. Here is an example from the 23rd of March:

A split occurs, then both rooms persist for multiple decent-sized games. Extra people found a game to play.

Here is one from a month ago:

The first split created a room that persisted for 3 games, then the second split failed.

Overall, the system seems to be on the cusp of working. It has a decent chance of creating a persistent second room, which is more than previous systems could achieve. What I really need is some detailed accounts of what happened during this splits, why they succeed or fail, because all I have is logs.

For example, I suspect split tends to fail more often when it follows a 1-hour game, or when the peak activity time has passed and the games are winding down over the next eight or so hours. So one question I have is when do people split, does it happen right after the game has ended, before the players figure themselves out?

Splits seem to be failing unusually often in the past two weeks or so. Can anyone report on why? I have some hypotheses:

• High ranked players have found a way around the rejoin restriction, and enough of them use it to go back to the old room so the new room dies. This is supported by the fact that someone told me they knew a way around the split, around a month ago, (that they used for when their party is split), and by the fact that the high rank room is the one that has died in recent splits. In previous months, cases of room death were randomly split between high and low rank.
• Something changed on the technical end to remove the secret sauce that kept the system afloat. All I can think of is the release from the 7th of April that fixed the bug that caused high rank spectators to split to the high rank room as a player. I speculated at the time that this bug may well have propped up the life of the high rank room, because suddenly becoming a player forces the previous spectators to re-evaluate whether they would actually like to play the game that is now in front of them. Perhaps that speculation was true.
• This is random noise. Some splits fail, and we get a split every few days, so there are going to be runs of weeks where all the splits fail.

It could be that we're out of the honeymoon period, but usually with things like this people don't like it at the start, then get used to it. So a backlash six months on would be unusual. Another thing to take into account is the cost of failure. The failed splits I've looked into tend to have the same players as the battle before, which indicates that a failed split doesn't leave a whole bunch of people out of playing. One or two might not notice and be lost, but mostly people are making it back to the room.

Anyway, I've increased the 32 player split threshold to 44, because the split failure rate seems a bit too high. We have to accept some failure rate, just as we accept some rate of games being !exited during placement due a problem, because that is just life. But the rate should be kept low enough for most people to trust the system.

From what I remember of the code which implements votes, trying to make multiple votes in one room at the same time work (with no server-breaking bugs ever) is a far from trivial task. As in, I expect it is possible, but it's definitely not low hanging fruit.

It's a technical solution to a tragedy of the commons situation. Solving such situations culturally, for a "general public", is difficult because everyone's local incentives point towards continuing the problem.

I personally often strongly prefer large games over small games.
Since a split game in unlikely to be 16v16, ending up in the smaller lobby is effectively "you are randomly banned from playing now", and I either join an open coop game, or quit entirely (and I'd much rather have to spectate for a game when joining a full lobby, than be forced into another lobby where I don't want to play).

I'd like a system where players/waitlisters can volunteer to join a new lobby (somehow displayed next to the player list), and once enough players volunteer they are moved to the new room - that should serve the purpose of finding a game for "surplus" people to play, while also preserving the agency of the players to choose what kind of game they are interested in.

I'm not sure that is how it would go in practise. There will be some range of sizes where the waiting list is the difference between people making a new room, or just sitting in this one. But below that, and the new game will have trouble getting off the ground, while above that, I expect people to eventually make a new room. What is to stop people spectating the running game, then joining it in a frenzy when it ends?

By the timing of your post, it sounds like you were part of the split that happened two hours ago. I would like to know what happened, what actions did people take and what caused them to move around. Why did the split fail? Have people figured out a way around it, and enough people implement these workarounds to bring the system below the coordination threshold?