How to fix autofill.

[{quoted}](name=Dog woof woof,realm=EUNE,application-id=3ErqAdtq,discussion-id=ZuXEEWw4,comment-id=00000000,timestamp=2020-01-03T21:15:23.480+0000)

Do you mind elaborating, how it makes queue times longer for autofill willing. I genuinely dont understand

I won't make it longer, but it will increase the frequency that someone willing to autofill will get autofilled.

If I get autofilled every 5th game on average, if people opt out, I will get filled every 4th game. Also, there will be so many in queue for popular roles that if I want to play mid, the only actually get it will be to be unwilling to autofill.

Autofill is needed because there is a shortage of players in certain roles. Allowing people to opt out does not remove the shortage... it reduces the available players for that role making the problem worse.

The way to get rid of autofill is to remove the need for it by getting the same number of people maining each role. Figure out how to make that happen and autofill is no longer needed.

Do you mind elaborating, how it makes queue times longer for autofill willing.

Sure!

By having the option to not autofill, a lot of potential support players are removed from the queue. There are multiple possible options, which I will elaborate upon using an example. Let's simplify the following example by assuming that about 10% of players queue for support and every role except support is equally popular.

As it stands now, an unknown amount of players queue up at the same time. 10% of them queue for support, the other 90% queue for various different roles. Roughly 11% of the non-support players get autofilled to support, resulting in a perfect distribution of 20% of players per role. Low queue times for everyone. Now let's assume 50% of all players are using the no-autofill queue. Let's also assume that all support players stayed in the normal queue, since they have no reason to switch.

The first options is to continue autofilling player in the normal queue at the normal rate. If we continue to autofill the remaining 40% of non-support players at the same rate of roughly 11%, we can get a maximum of 14.5% of players in the support role. That would result in 27.5% of players not being able to find a match, because their team would be lacking a support. If players continue to queue up with roughly the same role distribution, the average wait time for any role except support would be about half a standard game length, or 15-20 minutes.

The second option is to increase the rate at which players are autofilled to deal with the lack of support players. In our example, the average autofill rate would need to rise to 25% to provide enough support players. Queue times would stay low at first, but getting autofilled 25% of the time won't sit well with many players. That will drive them to use the no-autofill queue instead, thus removing more potential support players from the queue and increasing the rate at which players need to get autofilled. This can easily cause a feedback loop, until there are only support and fill players left in the normal queue and almost everybody else is using the no-autofill queue. At that point, autofill might as well not exist. 10% of players still queue for support, but nobody is left to fill for the missing 10%, resulting in an average wait time for non-support players of over one standard game length, or ~35 minutes.

As you can see both options so far result in long queue times. The third option is to separate the queues from each other completely. However, this would fail almost immediately, since no support or fill player has any reason at all to use the no-autofill queue. They are always autofill protected anyway. The normal queue would be fine, but the no-autofill queue wouldn't have a single player willing to play support in it, resulting in infinite queue times.

In conclusion, every possible outcome either increases queue times for non-support players (or whatever role is the least popular) or is simply not viable. Mixing the options is possible (for example giving the normal queue preferred treatment), but will always result in a mixture of the negative effects.

The only possibility I can see that wouldn't increase queue times, is to mix the second option with the third. Increase the rate at which players get autofilled, so the normal queue has a surplus of support players. Only use this surplus to queue with the no-autofill queue. Normal queue times for the normal queue, but with increased autofill rates. Long queue times for the other queue, but with no autofill. However, this is an incredibly delicate balance and could even be impossible. The matchmaking must not increase autofill rates too much, otherwise it will cause the feedback loop I mentioned earlier. Once you go past that point there's no going back. But if autofill rates are too low, then there's no surplus of supports for the no-autofill queue, basically killing it.

I hope I could explain this in an understandable manner.

Edit: And one last thing to consider; Even if this type of queue could work, would it be fair to increase the rate at which you're autofilled to accommodate people who refuse to get autofilled at all?