pac
Joined: Oct 03, 2005
|
 Posted:
Feb 28, 2007 - 03:58 |
 
|
I've been discussing with Craftnburn in IRC the question of whether or not a second open division - in addition to Ranked - can be viable. What it seems to come down to is: will enough coaches move? Will it be possible to go into Gamefinder and be confident of finding a match fairly quickly in a division that is not Ranked?
It is a chicken-and-egg situation. Coaches don't want to move because they aren't sure they'll be able to find games, and coaches aren't able to find games because not enough move.
So - in my opinion, I don't know whether Craftnburn actually agrees! - it comes down to a matter of numbers. It's also fine to make general appeals to those who want casual games - as brownrob has done in the first thread here - but here I want to hammer out the maths of it all.
So, the question is: how many coaches would have to switch over to playing predominantly Unranked in order for it to be viable as an open division?
I started trying to calculate this myself. I figured out some of the assumptions I needed to make, but haven't figured out how to put them together to get a value for the figure I'm after. And so I turn to sk8b … I mean, to the boards in general. 
Here are some assumptions (feel free to criticise them - these are Craftnburn's and my best guesses so far):
To be confident of finding a game, the average coach needs to see at least 18 teams on Gamefinder. (We'll call this c.)
When he lfgs, the average coach puts 3 teams onto Gamefinder. (g)
The average match lasts 75 minutes. (m)
An acceptable length of time (for the average coach) to wait before finding a game is 15 minutes. (w - am I allowed to use w?)
The average coach plays one match per day. (d)
<i>We need assumptions regarding coach distribution over time(-zones).</i> [Hitonagashi]
Do we still need more assumptions? Probably. I'm quite tired - too tired to figure this out, and I wouldn't be able to trust my conclusions anyway.
Remember, we're trying to improve these assumptions, and refine the model (once we have a model!), so that we can come up with a reasonable estimate of how many coaches need to switch over to Unranked to make it viable as an open division (this will be x!).
Once we have this figure, we can then start a thread saying: 'Sign up here saying that you would prefer to play in an unranked, non-time-limited, casual play, open division. You don't have to switch yet. Should we get x names, we have calculated that that will be enough to sustain a division. Then, and only then, we will all switch over, and no one need put their teams "at risk".'
Of course, all this assumes that - at some point - Christer will make it possible for Ranked teams to be transferred again. However, I'm quite confident that he will do so. What's more, if we can figure this out, and get x coaches together, we can give Christer the confidence that he can allow such transfers without well-known teams ending up 'stuck' if things go wrong (as has happened in the past).
I hope you will contribute, sk … er, anyone!  |
_________________
Join us in building Blood Bowl Sixth Edition.
In other news, the Hittites are back. Join us in #fumbbl.hi
Last edited by pac on %b %28, %2007 - %04:%Feb; edited 3 times in total |
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 04:05 |
|
One further aspect before I get some needed sleep.
Naturally, coaches might not want to move over totally. The whole point of having two divisions is that they would have different functions. A coach might prefer to play most of their games in Unranked, but sometimes play competitively in Ranked in official tournaments, say. Or a coach might prefer to play most of their games in Ranked, but occasionally play a few in a more relaxed way in Unranked.
The easiest way to handle this would be to allow coaches to sign up as, say, 2/3 of a coach. That is, 'I would not switch over completely, but would play 2/3 of my open games in [U]'. Or any other proportion.
This works well for the (lack of) model, but may be hard for those signing up to understand … |
_________________
Join us in building Blood Bowl Sixth Edition.
In other news, the Hittites are back. Join us in #fumbbl.hi |
|
Hitonagashi
Joined: Apr 09, 2006
|
| Posted:
Feb 28, 2007 - 04:08 |
|
Mmm, you also need a balance of the timezones for a realistic probability of the chances of finding a game at a particular point in time.
For example, say that x people sign up, but x/4 are on Australian time, x/3 on European, x/3 on American and the rest awkward peeps. At no point in time will you have all x online, but you will have your total as x. Either double your requirements for x once it is calculated, or work out a separate x for each time zone...I think that the former would be easier. |
_________________
http://www.calculateyour.tv - an easy way to work out specific team builds.
 |
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 04:14 |
|
| Hitonagashi wrote: | | Mmm, you also need a balance of the timezones for a realistic probability of the chances of finding a game at a particular point in time. |
Noted.
Edited this point into the top post. It may develop as we go until we have a formula of some kind (with lots of credits!). |
|
|
ClayInfinity
Joined: Aug 15, 2003
|
| Posted:
Feb 28, 2007 - 04:57 |
|
There is a number of users online number that I see on my fumbbl page. It fluctuates from low numbers in the 40's (now, which is 3pm in Sydney, 4am in Europe) and when I am on at night in prime time Aussie timezones which is middle day in Europe its around 140. In the wee small hours in Sydney, I am not sure what the number is but it would be higher (say 200).
So I think you need to do a distribution of how many players are online to play at any given time from a number from 40 to 200. That distribution obviously would be skewed to peak times (8pm Europe) but you would get your basis using an assumed Normal Distribution graph.
So rather than absolutes, you need to think of distributions of variables...
Any stat wizzes out there?? |
|
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 10:36 |
|
| ClayInfinity wrote: | | So I think you need to do a distribution of how many players are online to play at any given time from a number from 40 to 200. That distribution obviously would be skewed to peak times (8pm Europe) but you would get your basis using an assumed Normal Distribution graph. |
Yes, it occurred to me last night that aiming at 'the average coach' was no good. If a division is just viable for 'the average coach' then for almost half the coaches it's not viable - they will have to wait too long to get games, so they will give up, so it will become viable for fewer people, etc …
Essentially this means we need more conservative assumptions, so that the division would be viable for a larger chunk of that normal distribution curve.
| Quote: | | Any stat wizzes out there?? |
I've sent for one. |
_________________
Join us in building Blood Bowl Sixth Edition.
In other news, the Hittites are back. Join us in #fumbbl.hi |
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 17:04 |
|
Where are all our resident maths people?
I don't know, when it's pseudo-randomness or the luck stat we can't get rid of you, but when we need you for something useful …? 
I'm sure there are some minds out there who can help put this together!  |
_________________
Join us in building Blood Bowl Sixth Edition.
In other news, the Hittites are back. Join us in #fumbbl.hi |
|
PeteW
Joined: Aug 05, 2005
|
| Posted:
Feb 28, 2007 - 17:41 |
|
Throwing your numbers together I reckon you'd need 576 coaches to move over to make 18 games on the lfg page.
However, if you can convince Emphasy to move then it reduces to about 12. |
_________________
"Jesus loves me this I know, 'cos my Bible tells me so." MrMojo - where did you go? |
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 17:44 |
|
| PeteW wrote: | | Throwing your numbers together I reckon you'd need 576 coaches to move over to make 18 games on the lfg page. |
Not that I don't trust you, but could you show your working so that we can then test the model with different assumptions?
| Quote: | | However, if you can convince Emphasy to move then it reduces to about 12. |
No chance of that - why would he fear a time limit!? |
|
|
PeteW
Joined: Aug 05, 2005
|
| Posted:
Feb 28, 2007 - 17:49 |
|
My rather rude model was
c = xgwd / 1440
The reason being that to get a certain number of games you need to have an expected number of people with a good number of teams willing to wait to play d games in the course of a day - hence the 1440 (24 * 60)
It could be made much neater but this is all very rough. |
_________________
"Jesus loves me this I know, 'cos my Bible tells me so." MrMojo - where did you go? |
|
Rijssiej
Joined: Jan 04, 2005
|
| Posted:
Feb 28, 2007 - 17:50 |
|
my calculations come to 576 coaches aswell. I must have used the same formula PeteW used. |
|
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 18:00 |
|
My instinct is that this first estimate must be too high. We know that Ranked is a viable open division, but do we really have many more than 500 active Ranked coaches?
Certainly the assumptions, and probably the model, need refinement.
As a starting point, the hours of, say, 0100 to 0900 bb time are a pretty dead time on gamefinder (Ranked included). You can find games then, but I don't think most coaches expect to. (I realise this excludes antipodeans somewhat, but they do already have their own (SWL-linked) methods for finding games.)
So that allows us to discount a third of the day, which in turns cuts our required number of coaches by a third to under 400. |
_________________
Join us in building Blood Bowl Sixth Edition.
In other news, the Hittites are back. Join us in #fumbbl.hi |
|
PeteW
Joined: Aug 05, 2005
|
| Posted:
Feb 28, 2007 - 18:08 |
|
Also I reckon the average number of games played my be higher and that many people are prepared to wait longer. |
_________________
"Jesus loves me this I know, 'cos my Bible tells me so." MrMojo - where did you go? |
|
pac
Joined: Oct 03, 2005
|
| Posted:
Feb 28, 2007 - 18:13 |
|
Looking at this from another perspective, we know that Ranked is a viable open division, so let's take what we know about Ranked.
http://fumbbl.com/FUMBBL.php?page=stats&op=games
As you can see, around 450 matches are played in Ranked each day. That would mean (according to the present model's assumption of one game per coach per day) that there are 900 active Ranked coaches …
That doesn't sound right: average number of games played per coach should be higher, you're right. |
_________________
Join us in building Blood Bowl Sixth Edition.
In other news, the Hittites are back. Join us in #fumbbl.hi |
|
Markus
Joined: Aug 26, 2003
|
| Posted:
Feb 28, 2007 - 18:24 |
|
Christer should have statistical data about the number of ranked games per day, nr of active coaches per day, nr of teams in gamefinder per day, maybe even splitted into divisions. If he shares his data it would be much easier to make assumptions.
You can find games from 0800 bbtime to 1900 bbtime during the week (mo-fr) in ranked (my own experience). As long as you're not to picky with the choice of opponents and/or teams.
You need less coaches for u/l than for ranked, as u/l isn't as restricted concerning back to back games as is ranked. If you like the way your opponent plays one game you can get the next right after the very game. No need for as many teams in gamefinder as in ranked. |
_________________
"When the gods wish to punish us they answer our prayers."- Oscar Wilde
"The opposite of a correct statement is a false statement. But the opposite of a profound truth may well be another profound truth."- Niels Bohr |
|
|
|
| |
Designer's Comm...
Search
