FUMBBL :: Online Blood Bowl League

Hogshine

Joined: Apr 04, 2007

Posted: Nov 17, 2008 - 21:30

westerner wrote:

I don't agree with with your characterization of Blackbox as being designed solely to eliminate "pickiness". I believe Blackbox is about automating fair matchups. What is fair? What the market says it is. There should be tons of [R] data available to base this on to show that TS is not the whole story. Like I said, Christer has at least considered this. so please don't start with "if it served your purposes" business.

With TS, the "average" matchup is going to be with equal TS teams... you do see this, right?

And no, it's not normal for teams to play down ~10 TS most games, and down CR nearly every game. If [B]lackbox gave me your last 10 or so matches from [R], I'd feel a little hard done by. I don't enjoy playing down TS all the time, similarly for CR.

TS is not the whole story, we can agree there, but racial strengths are already taken into account. I would hazard a guess that with most teams, a combination of the TS and the racial strength ratios would give fair matches that the majority of coaches would be happy to play as either team.

Oh, and your next point about Jan bullying you? Seriously, no. Just no.

funnyfingers

Joined: Nov 13, 2005

Posted: Nov 17, 2008 - 21:31

Is BBR per team or coach?

Hogshine

Joined: Apr 04, 2007

Posted: Nov 17, 2008 - 21:35

funnyfingers wrote:

Is BBR per team or coach?

My understanding is it's per coach.

Edit: Yes:

Christer wrote:

All B coaches have a Blackbox Ranking (BR), and a Blackbox Bash Ranking (BBR)

Buur

Joined: Apr 29, 2004

Posted: Nov 17, 2008 - 21:40

As for making suggestion to the scheduler system:
I think it would be cool if there where some way incorporate the race rankings, as they emerge for the bbdiv.

-Buur

_________________

For most people, reason is nothing but their own believes.

westerner

Joined: Jul 02, 2008

Posted: Nov 17, 2008 - 22:05

Mr_Foulscumm wrote:

westerner wrote:

Economics suggests that if you aggregate all those agreed-to matches, you can determine the "going rate" for playing vs a bashier, higher TS, or better coach.

westerner, this is pretty close to describing cherry picking.

As soon as you start choosing the best games for you (not the most even team match up) then you're stumbling damn close to being a picker. And yes, it is a smart strategy to use and has worked for all coaches here.

But, [B] is supposed to be a way of getting past this aspect of games on Fumbbl. So any system that you introduce to limit which coaches you get to play immediately reintroduces aspects into the equation that [B] was designed to avoid.

Suppose we define cherrypicking as offering a match you yourself would decline.

Incorporating the "going rate" into the matching algorithm drives out cherrypicking if it is based on the what most coaches would likely accept in [R].

Example A: 10TS premium for bashers
"My 150TS Bashers vs your 140TS elfballers?" / "My 140TS elfballers vs your 150TS Bashers?"
Coach A: No/Yes
Coach B: No/Yes
Coach C: No/Yes
Coach D: No/Yes

Example B:
"My 150TS Bashers vs your 150TS elfballers"? / "My 150TS elfballers vs your 150TS Bashers?"
Coach A: No/Yes
Coach B: No/Yes
Coach C: No/No
Coach D: Yes/Yes

Example C
"My 140TS Bashers vs your 150TS elfballers"? / "My 150TS elfballers vs your 140TS Bashers?"
Coach A: Yes/Yes
Coach B: Yes/Yes
Coach C: No/No
Coach D: Yes/Yes

In this example, the "market" has determined that more matches will be played in scenario C. That is the "going rate". The largest number of games will be played there.

You're likely much more experienced than I am at fumbbl, however, the market is likely to do a better job than either of us in determining what is a "fair" match. My suggestion is, capture that by analyzing the mounds of data on [R], and build that into the matching algorithm.

_________________
\x/es

Mr_Foulscumm

Joined: Mar 05, 2005

Posted: Nov 17, 2008 - 22:17

The problem your market runs into is that people cherry pick. The "going rate" is based on people picking. Hence why it's a bad basis for [B]... you can't drive something out if you're building it into the structure.

_________________
Everybody's favorite coach on FUMBBL

westerner

Joined: Jul 02, 2008

Posted: Nov 17, 2008 - 22:22

CircularLogic wrote:

westerner wrote:

Let's back up a bit. Why don't you tell me your definition of cherrypicking, and we can go from there.

Cherrypicking is taking favorable matchups over balanced ones.

On page 2 CircularLogic wrote:

For me a balanced game is a match, where the odds of winning are not significantly changed, if the opponents would switch their teams.

I like "Offering a match that you yourself would refuse". But by any of these definitions, I am a cherrypicker in [R]. I my have been cherrypicked my
first few games without knowing it, and would have continued blissfully unaware and just trying to win the next game, until I saw some dude throwing around the football in a match without trying to score and explain that he played for team development, and advised me to "refuse all matches vs ogres and khemri". I caught on fast, I guess.

CircularLogic wrote:

westerner wrote:

I don't agree with with your characterization of Blackbox as being designed solely to eliminate "pickiness". I believe Blackbox is about automating fair matchups. What is fair? What the market says it is. There should be tons of [R] data available to base this on to show that TS is not the whole story. Like I said, Christer has at least considered this. so please don't start with "if it served your purposes" business.

So if I persuade total noobs into playing my stronger team (like some coaches do) then it is a fair matchup? And those should should be incorporated into [B] because that`s what I can get out of the 'market'? You might think that because those are the rules you are picking your R matches by, that everyone does that. But I claim, that the opposite is more likely. A new/weaker coach is more likely to accept a game where he is at a disadvantage from the start. Those who are picking (in it`s worst form) are looking for those kind of unsuspecting coaches to stomp them. That`s the market.

That should all wash out of the R data, since the market TS differential for (say) bashiness should not be affected. i.e., when cherrypicking a weaker coach, you are just as likely to do it with bashers as elfballers.

_________________
\x/es

westerner

Joined: Jul 02, 2008

Posted: Nov 17, 2008 - 22:34

Mr_Foulscumm wrote:

The problem your market runs into is that people cherry pick. The "going rate" is based on people picking. Hence why it's a bad basis for [B]... you can't drive something out if you're building it into the structure.

But look what's happened. You've driven out Example A (Bashers receiving a TS boost over Elfballers) b/c it's not as competitive as Example C (Elfballers receiving a TS boost over Bashers).

Building in "the going rate" automatically drives the worst of the cherrypicking out of the system.

Yes, you're still incorporating some "picking" relative to pure TS, but who's to say TS is the untouchable standard? There's ways to game TS too.

_________________
\x/es

westerner

Joined: Jul 02, 2008

Posted: Nov 17, 2008 - 22:46

Hogshine wrote:

With TS, the "average" matchup is going to be with equal TS teams... you do see this, right?

Not necessarily. Suppose, for the sake of argument, the bashy premium is 5 TS in B, i.e. Bashy teams play an average of 5 TS up vs non-bashers. Then, the average match may have a TS differential to the extent that bashers get paired against non bashers. Does that make sense?

Hogshine wrote:

And no, it's not normal for teams to play down ~10 TS most games, and down CR nearly every game. If [B]lackbox gave me your last 10 or so matches from [R], I'd feel a little hard done by.

I never suggested that my matches in R be used as any guideline for B. Jan brought that up to disparage me. Even though he admits to having himself cherrypicked earlier in his career. And what about you?

Hogshine wrote:

TS is not the whole story, we can agree there, but racial strengths are already taken into account. I would hazard a guess that with most teams, a combination of the TS and the racial strength ratios would give fair matches that the majority of coaches would be happy to play as either team.

I agree, and I believe the guiding principle for B should be matches that a majority of coaches would be happy to play as either team.

_________________
\x/es

Mr_Foulscumm

Joined: Mar 05, 2005

Posted: Nov 17, 2008 - 22:47

westerner wrote:

Mr_Foulscumm wrote:

The problem your market runs into is that people cherry pick. The "going rate" is based on people picking. Hence why it's a bad basis for [B]... you can't drive something out if you're building it into the structure.

But look what's happened. You've driven out Example A (Bashers receiving a TS boost over Elfballers) b/c it's not as competitive as Example C (Elfballers receiving a TS boost over Bashers).

Building in "the going rate" automatically drives the worst of the cherrypicking out of the system.

Yes, you're still incorporating some "picking" relative to pure TS, but who's to say TS is the untouchable standard? There's ways to game TS too.

Sorry I disagree with this. The fact that people pick Example C isn't because it's the most competitive, it's because most "ballers" are wuzzy, pixel hugging, wimps.

So the fact that the "market" says this is the most common match up doesn't mean it's the fairest, because, people pick games where the potential gains are bigger then the dangers.

EDIT: What you are suggesting is that it's ok for elfballers to cherry pick games but not ok for bashers? oh, and that and actual even game is out of the question?

_________________
Everybody's favorite coach on FUMBBL

Last edited by Mr_Foulscumm on %b %17, %2008 - %22:%Nov; edited 1 time in total

shadow46x2

Joined: Nov 22, 2003

Posted: Nov 17, 2008 - 22:49

westerner...

i'm curious....

if you were coaching say...109 TS humans...would you take a matchup against 136 TS undead?

--j

_________________

origami wrote:

There is no god but Nuffle, and Shadow is his prophet.

JanMattys

Joined: Feb 29, 2004

Posted: Nov 17, 2008 - 22:58

westerner, plain and simple:

Your reasoning would make sense if all people had the same attitude, goals and information. (I would still disagree on principle, but it would make sense).

Thing is, in the "cherrypicking" thing, one coach is often very determined to have the upper hand in the matchup, while the other just doesn't know/notice/care.

So, if you take R as an example, you get that the most played games are not necessarily the most fair ones.

TS is by far a better indicator than games played in a division with free picking options and different levels of understanding of the game mechanincs.

_________________

Gatts

Joined: Jun 18, 2004

Posted: Nov 17, 2008 - 22:58

id like to see the B box as a wake up call to the online bb community, a way to find the way back to fun games feauturing any side vs any side.

I dont want more of what i would get in R. If we let R statistics decide what kind of games the box should offer you take away my reason for playing the box. I think/hope that B will give rise to a new generation of brave coaches who does not play for "team development" but rather plays their hearts out trying to win the game that's placed before them.

I dont really mind exactly how good the scheduler gets, it doesnt really mater if it always find the best possible match. If I have to fight uphill games cause my BR increases, I can deal with that aswell, as long as people use the box. Dont use it cause I tell you too use it because its good for you.

Use it because it feels good to beat orcs, dwarves, hemri and all the other teams you dont normally play in ranked, use it cause its fun to play as these teams without having to wait around for days getting a game. Use it with clawridden chaos teams and with dodgy woodelfs. If you get beaten down use the blackbox again and take the opportunity to have plorg teach you a lesson on how a halfling team can devastate an outnumbered team Smile

.

This is probably not the rioght place for this kind of rant, but the current talk about using economic models on data from ranked to try to help the blackbox schedule games goes against the very essence of what id like the box to be.

_________________
Players die, touchdowns are forever!

Hogshine

Joined: Apr 04, 2007

Posted: Nov 17, 2008 - 23:18

westerner wrote:

Hogshine wrote:

With TS, the "average" matchup is going to be with equal TS teams... you do see this, right?

Not necessarily. Suppose, for the sake of argument, the bashy premium is 5 TS in B, i.e. Bashy teams play an average of 5 TS up vs non-bashers. Then, the average match may have a TS differential to the extent that bashers get paired against non bashers. Does that make sense?

No. If I play a game where I have a 5TS advantage, my opponent is playing a game with a 5TS disadvantage. So the average of the two game situations taking place is 0TS.

Edit: Just to clarify, I don't care for the difference between bashers and non-bashers. Racial strengths are taken care of, as we agreed already. I see no good reason why non-bashy teams should be somehow rewarded. Elves tend to win more against bashers.

westerner wrote:

Hogshine wrote:

And no, it's not normal for teams to play down ~10 TS most games, and down CR nearly every game. If [B]lackbox gave me your last 10 or so matches from [R], I'd feel a little hard done by.

I never suggested that my matches in R be used as any guideline for B. Jan brought that up to disparage me. Even though he admits to having himself cherrypicked earlier in his career. And what about you?

Go for it. Search through my past matches and find a period when I was here and picking.

I'm not saying I've never played a match where I've had an unfair advantage, or that I've never played a match which I wouldn't play the other way round. But I've never done it consistently. I can't remember every case, but I would guess they were all either fun matches with friends, other people challenging me, or just the result of the closer TS matchups turning me down.

This happened for me mostly by chance. When I first joined, I didn't really know much about Bloodbowl, and so nearly every coach was better than me. I tended to wait until someone else challenged me, and then just accept it. Then I started to talk to people like SillySod etc, who while teaching me how to play the game, also kinda instilled in me that cherry picking (and elfballing) was/is lame. So as soon as I learned how to cherry pick, I also learned not to. (this isn't to say I'm better than anyone else, it's just how it happened. Cherry pick if you want, I don't mind. I just reserve the right to turn down any game against you.)

westerner wrote:

Hogshine wrote:

TS is not the whole story, we can agree there, but racial strengths are already taken into account. I would hazard a guess that with most teams, a combination of the TS and the racial strength ratios would give fair matches that the majority of coaches would be happy to play as either team.

I agree, and I believe the guiding principle for B should be matches that a majority of coaches would be happy to play as either team.

[/quote]

As Shadow points out, you would play 109TS humans against 136TS undead?

Reisender

Joined: Sep 29, 2007

Posted: Nov 17, 2008 - 23:46

westerner wrote:

]
You're likely much more experienced than I am at fumbbl, however, the market is likely to do a better job than either of us in determining what is a "fair" match. My suggestion is, capture that by analyzing the mounds of data on [R], and build that into the matching algorithm.

yeah the market knows it. the market knows all. this is why pure capitalism is so great, has always worked and will always..... wait..... Shocked