| Poll |
| Change the Scheduler? |
| yes, encourage mirror matches |
|
65% |
[ 88 ] |
| no, discourage mirror matches |
|
34% |
[ 47 ] |
|
| Total Votes : 135 |
|
Fela
Joined: Dec 27, 2004
|
 Posted:
Apr 21, 2011 - 22:13 |
 
|
| licker wrote: |
We know there is a bias in the selection procedure, we even see the evidence of it presented in the raw data, else the %matchups would match the %teams in [B], which they clearly do not. Hell, we even understand that they are not even supposed to! It's in the formula being used.
|
Wiseguy. There is a miniscule bias in the formula that supposedly is there to counter ANOTHER (inherent) bias in the system. Based on an overinterpretation of said bias you expect to see something in the raw data and when you see an effect in there you claim evidence for your expectation.
Did you realize that there was a study that clearly shows that drinking coke significantly lowers the chance of having complications in child birth? Correlation != Causality.
| Quote: |
As to providing you with expected matchups based on an unbiased scheduler all I have to say is, really?
Really? That's what you think requires graduate level statistics course work?
|
I've got basic statistics covered, therefore I can confidently say that the effect the small sample sizes may have on expectation values is not trivial. Yet you're treating it as if it were trivial. All i'm asking you is to STOP that, come up with a solid theory or at least WAIT until someone else does before you demand changes based on uneducated assumptions.
| Quote: |
But moreover, what do you think would happen? Nothing but mirrors? |
This actually shows that you didn't understand a thing.
What I expect to happen concerning mirror matches alone is: Nothing significant. Because i'm completely unconvinced that the artificial bias of 3% in the formula is responsible for the behaviour we're observing regarding mirror matches for the more active races.
Whichever way we look at it, one thing the raw data clearly showed: the bias AND the other underlying effects causing mirror matches to apparently be less common can EASILY be overcome by a more dominating factor. Maybe you can show how much you understood by explaining, what I mean with that? |
|
|
licker
Joined: Jul 10, 2009
|
| Posted:
Apr 21, 2011 - 22:14 |
|
Macavity has it, I think others get it as well.
The issue really is what is [B] supposed to be?
The answer is, whatever is agreed/decided upon.
The system in place now is 'fair' so long as people understand what the system is. That the system encourages certain racial choices is just part of it. It is. It isn't good, it isn't bad.
It could also be different, and again, it wouldn't be better or worse, it would just be different.
Now I'm speaking entirely generally from a removed position. Clearly it would be 'worse' to change it for team who currently have it 'better', but overall considerations like that should not matter.
Personally, I think the system should be unbiased, selecting matchups on TV only with no 15% rule (or maybe a 50% rule). This just matches up teams starting from the highest TV to the next highest, and so on. Or you can do it from the bottom up. Yes, some consideration given to players activating multiple teams if your concern is maximizing games played, not minimizing TV differentials in matches selected.
Of course this still has some flaws, any system will have some flaws, but it removes race as a consideration in how matches are selected. |
|
|
Fela
Joined: Dec 27, 2004
|
| Posted:
Apr 21, 2011 - 22:30 |
|
Thanks for being a voice of reason, as always. Well, maybe not always, but often enough  |
|
|
licker
Joined: Jul 10, 2009
|
| Posted:
Apr 21, 2011 - 22:31 |
|
PC-
Thank you for the detailed explanation. I think in my prior post (which I wrote before seeing your post) you see that I get it. [B] doesn't have to be anything other than what it is, and that it was agreed (or decided) to be what it is whenever that was is fine. Though there's no harm in discussing if perhaps time has come for it to become something else. That's my only motivation here, I find the discussion interesting. I play [R] and [B] to get game experience, I'm hoping to join a couple leagues because that's the place I think I will get the most enjoyment out of BB, though that's kind of beside the point... just to restate I don't really have an axe to grind about clawpomb or the [B]/[R] differences or what anything should be to anyone.
I'll have to take your word for it that removing the bias from [B] would further bolster the bashy teams case, it does not appear immediately apparent to me why, but perhaps there is just always going to be more bashy teams in any kind of division where teams can play 100s of games. What would be nice is 'enforced' diversity such that there is by design an even number of teams in any group, but that's impossible outside of [L] really.
Fela-
Don't make things up which I never said. I didn't expect to see anything in the data, the data showed what it showed all by itself. Now if you're going to state that not enough data points are available to draw a conclusion about the relative% of mirror matchs vs. expected then by all means do so, but do what you keep on asking everyone else to do, and actually prove it.
And if you think that a 'fix' for the system is to decrease the 3% value then I'm all for it. Just don't decrease it to 0 without taking out the racial suitability scores at the same time. |
|
|
JackassRampant
Joined: Feb 26, 2011
|
| Posted:
Apr 21, 2011 - 22:39 |
|
IMO, the best first answer to the Blackbox bottleneck is to get the Wizard going... that may not be enough, but it'd be a good start. Wizards ruin drives, and the best way to avoid injuries is to ruin the offensive drive. I doubt it'll totally shake things out, but it will have an impact.
Now, for a suggestion to transcend the war over data and statistics and make everybody happy. Pardon my newbieness: I don't know much about your league formats, and it's possible this has already been resolved in L, or will be addressed when L goes to the new client. In that case, my suggestion just goes to "get L working", which should go unsaid. With that caveat, here goes:
Maybe the solution is to have more than one scheduler league? This would give Christer a venue to see what happens when one plays with various elements, in addition to satisfying the most vocal complaints of other coaches. After all, if ClawPOMB is ruining the non-Chaos teams that peak out over 2M, then using games played to dilute TV in the scheduler format might have a big impact; 2M Chaos teams would still run rampant, but at least they'd be eating saws and fireballs on a regular basis. You would also protect the young teams from the worst imbalances and make it impossible to hover at a low TV and farm newer teams. How hard would this be?
If it's feasible, it would be a great testing ground for Blackbox. You'd be able to see the signal of format-created imbalances in the noise of the new ruleset and the basic differences between FUMBBL and tabletop. Then you can evaluate those for your criteria and make sure the Blackbox formula is doing what you want it to do. For instance, certain TV deficits can produce better relative matchups for one kind of team or another, and a floating formula can allow for fun stars and Wizard games. TV alone isn't a great indicator of team effectiveness, once inducements are considered. |
_________________
Lude enixe, obliviscatur timor. |
|
RandomOracle
Joined: Jan 11, 2004
|
| Posted:
Apr 21, 2011 - 22:47 |
|
| JackassRampant wrote: |
Maybe the solution is to have more than one scheduler league? |
The last thing we need is more divisions. Two Blackbox-type divisions would just mean that one or (in the worst case) both would die out. Divisions with a scheduler need a critical mass to stay viable and I don't think FUMBBL has enough active players to support more than one. |
|
|
Fela
Joined: Dec 27, 2004
|
| Posted:
Apr 21, 2011 - 23:13 |
|
| licker wrote: |
Fela-
Don't make things up which I never said. I didn't expect to see anything in the data, the data showed what it showed all by itself.
|
You're interpreting causality into the data without valid justification giving the limited knowledge presented. Sound better?
| Quote: |
Now if you're going to state that not enough data points are available to draw a conclusion about the relative% of mirror matchs vs. expected then by all means do so, but do what you keep on asking everyone else to do, and actually prove it.
|
My position is: let's gather some proof for one side or another before making rash decisions, yet you demand i bring proof to support my position? Sounds a bit weird to me.
| Quote: |
And if you think that a 'fix' for the system is to decrease the 3% value then I'm all for it. Just don't decrease it to 0 without taking out the racial suitability scores at the same time. |
It may make mirror matches more probable, including mirror matches noone really wants to see. I really care a lot less about that minuscule factor than about the way this discussion is held.
What happened to your answer to my question in the end? Too difficult? |
|
|
pythrr
Joined: Mar 07, 2006
|
| Posted:
Apr 21, 2011 - 23:43 |
|
PC - a brilliant post. The rest of us should shut up now. |
_________________
 |
|
Garion
Joined: Aug 19, 2009
|
| Posted:
Apr 21, 2011 - 23:45 |
|
My god is this still going on, the scheduler is fine. It is just completely random who you will be drawn against usually in a draw there are between 4 and 6 coaches active. You get paired against the team with the nearest TV. It will be very very rare that all the teams in the draw have the same TV difference and as a result get scheduled to avoid mirror matches.
I have a wood elf team that has played 12 games and apart from getting dwarves 3 times most of the others have all been different races. Do I think there is a problem with the scheduler because i got dwarves 3 times? NO i got dwarves 3 times because the times i activated my team they were the closest match according to TV.
Stop all the madness please. Box is working perfectly as it is. If you dont like it play ranked. That is the reason there are two divsion.
Some please lock this thing. |
_________________
 |
|
licker
Joined: Jul 10, 2009
|
| Posted:
Apr 21, 2011 - 23:49 |
|
| Fela wrote: | | You're interpreting causality into the data without valid justification giving the limited knowledge presented. Sound better? |
Sounds as wrong as the first time you said it. If it sounds better to you so be it. How many data points are you asking to look at?
| Fela wrote: |
My position is: let's gather some proof for one side or another before making rash decisions, yet you demand i bring proof to support my position? Sounds a bit weird to me. |
lol...
I'm asking you for evidence that there is not enough data right now to use in drawing a conclusion. Surely someone with your apparent background in stats can back that up.
But what 'proof' do you need? We already saw the formula, we all know it has bias in it, we all see that the data reflects this bias (again, it's expected and intended), so what 'proof' are you actually after?
| Fela wrote: | | It may make mirror matches more probable, including mirror matches noone really wants to see. I really care a lot less about that minuscule factor than about the way this discussion is held. |
It may??? Oh my, let me ask you how it wouldn't then. And I love your use of the absolute 'no one' here. Oh my, again.
| Fela wrote: | | What happened to your answer to my question in the end? Too difficult? |
Too pointless actually. I'm against these artificial 'dominating factors' to begin with. In any case, I already prescribed one which would never happen, too dominating to be sure. |
|
|
licker
Joined: Jul 10, 2009
|
| Posted:
Apr 21, 2011 - 23:50 |
|
| pythrr wrote: | | PC - a brilliant post. The rest of us should shut up now. |
You first |
|
|
PorkSol
Joined: Jan 10, 2008
|
| Posted:
Apr 22, 2011 - 00:13 |
|
| PurpleChest wrote: |
Oddly this is a massive change to B people are dressed up as a tweak.
In a 'true random' blackbox the weaker races get screwed. the emphasis moves utterly to maximising the punch for your TW in EVERY sense. In your race choice, in the level the team sits at, in the skills they choose, in the depth of the sqaud.
It will punish diversity. We will see more Orcs. More Dwarves. More Pomb squads.
|
Could you elaborate on the reasoning behind this a little more?
I take it your thinking is that the loss of the racial suitability factors would more than cancel out the popular races having to play themselves more often?
Is there an easy way to see the current impact of the racial suitability matchups? |
|
|
On1
Joined: Jul 12, 2004
|
| Posted:
Apr 22, 2011 - 00:32 |
|
Remove this mirror match stuff....
Dorfs need to face Dorfs as much as other people face Dorfs! |
|
|
Timlagor
Joined: Feb 13, 2009
|
| Posted:
Apr 22, 2011 - 00:39 |
|
Garion: a lot of coaches activate multiple teams at a time making the possible matches bowlbot chooses from a lot more numerous than might be immediately apprent (and thereby reducing the impact of few coaches on the games that get played)
Licker: someone worked out (and listed) that only a small proportion of racial matches were more favoured than mirror matches -removing the racial factor would decrease mirror matches for many races.
PC: I've seen the light!
I hereby propose a new system:
1) Check all activated teams for ClawPOMB and mathc them against each other
2) Check for MB and match them together with any remaining ClawPOMB team
3) Add the AV9 teams
4) Add everyone else (though more subdivisions may prove necessary later).
JackassRampant: multiple Divisions running concurrently would be hopelss. There may be a case for a one month experiment but I don't think Christer is going to do that lightly. |
_________________
Time for a new .sig |
|
Timlagor
Joined: Feb 13, 2009
|
| Posted:
Apr 22, 2011 - 00:41 |
|
| On1 wrote: | Remove this mirror match stuff....
Dorfs need to face Dorfs as much as other people face Dorfs! |
Ahh what we really need is a much stronger bias in favour of mirror matches to counter the improbability of them inherent in the system! |
|
|
|
|
| |
SWL Season 100!
Search
