When do affinities become diminishing returns?

[Vaevictis_386]

When do affinities become diminishing returns?

Is there any point to having 5 Exceptionals in the receiving corps for example, or is 3 exceptional receviers almost as good?

[Dutch]

If your team is given a ranking (vs. a rating) and you are #1, then additional firepower in that department won't do anymore good.

Cohesion seems to be on the bell curve, so perhaps Chemistry is as well?

[Vaevictis_386]

Sorry, what do you mean by ranking and rating?

[Dutch]

Yeah, I figured that might not make any sense.

What I mean to say is if FOF hard-codes a bonus based on how you rank against other teams, the point of diminishing returns then is when you continue to add chemistry even though you are #1.

If it's not in relation to other teams and is simply an individual effort, at some point FOF is calculating how many affinities you have for your QB and leaders. I was trying to suggest that is a rating.

However, all that being said, with us not even understanding how it works (and not knowing if you are #1 or not), there really is no reason to ever believe you aren't doing enough for your chemistry and there simply is no point of diminishing returns.

[Vaevictis_386]

Got it thanks.

[QuikSand]

there really is no reason to ever believe you aren't doing enough for your chemistry and there simply is no point of diminishing returns.

This is the part I subscribe to.

[QuikSand]

What I mean to say is if FOF hard-codes a bonus based on how you rank against other teams...

Just to put this into context, off the top of my head there's only one area I know of that is strictly based on pure rankings (where all that matters is where you rank, and not how much better/worse you were) and that's the amount allowed for staff hiring arising from the team's bottom line the prior year.

Some people get confused by some of the 100-based ratings in the game -- things like the cohesion and roster ratings. Yes, the game arbitrarily shows us a rating of 100 for whomever is the top team, but the ratings are still relative. If you already have the 100-rated roster, for example, and you add three more stud players... your rating won't change from 100, but the second best team will drop from something like 95 to 88, reflecting how you're pulling even further away from the pack. (I happen to think these 100-based ratings are purely cosmetic, but regardless they still do reward the #1 team getting better or the #32 team getting worse)

[aston217]

That's a great point Quik. I don't think we can know for sure if all that matters is you are #1, but I think that's generally accepted, at least for cohesion. I remember Ben commenting that after a few years in a new league, passing games will start to take off because at league inception nobody has had the chance to accumulate cohesion yet.

I think it's fair to suspect that at some point there is a limit though. Like "this is the maximum bonus you can get out of cohesion or affinity."

I would guess it's some point before this:

http://img843.imageshack.us/img843/4319/diminishingreturns.png

...but it's hard to say. I always put talent over affinities and try to focus on talent & cohesion rather than sacrificing. Go too crazy on affinities and you could be sacrificing talent in a big way, where talent is the base, and probably most important factor. Of course if you can somehow do both....but I've never been a fan of looking up birthdays, it's too tedious.

[aston217]

Another thing to note about QB affinities is I remember Firefly did some limited testing in SP and I think we came to the conclusion that it really only matters for your starters and not the backups. Or at least, it would be tough to say otherwise.

I.e, if your backups have 3x affinities, it won't help all of your offense position groups if they aren't playing.

I don't know if that applies to other positions as well

[Firefly]

Hi, I am Firefly and I do not approve this message.


However, it's true that I couldn't find evidence of a bonus for a triple exceptional affinity backup QB in limited testing.

[merry]

However, it's true that I couldn't find evidence of a bonus for a triple exceptional affinity backup QB in limited testing.

Hi,
Could you find evidence that affinities give a bonus for other positions?
Conclusive or marginal evidence?
Merry

[Firefly]

When I tested I found the evidence was pretty conclusive but I didn't keep any records and I don't know much about scientific statistics.

[QuikSand]

This is brutally hard to test for, period.

What testing there is suggests affinities do matter. I don't think we can say a lot more for sure.

What makes the most sense to me is to recall this is a computer game. Programmed by one guy. Who is human. Given that, I think it's a ton more likely that whatever system there is in the game is just some rote calculation and add-on factor... Rather than some much more complicated system of logarithmically adjusted scale-able results.

Search for AffIndex, used on this site a least a few times I suspect...and I suspect that's a quite effective proxy for how it is actually implemented in game.

I invest more time and effort into FOF chemistry than anyone, I'm rather certain. I don't do it for my health. It matters.

[aston217]

I don't think log-adjusted is that far out of the question. I mean, you probably have a lot more experience with this than almost anybody but we all agree that you can't get arbitrarily high benefits from affinity, right?

So a cap would be on way to do it. Say, a 10% bonus effect is all you can get.

A log scale is another. #1 team gets ___ bonus, #2 team gets ___ and so on. Both would be just a rote calculation and not a complex, multi-step algorithm, right?

Just some function that returns "% bonus".

[QuikSand]

...but we all agree that you can't get arbitrarily high benefits from affinity, right?

Why would we necessarily all agree to that? I think it's simpler and more straightforward to imagine it uncapped than capped.

This developer, despite his many strengths, doesn't have a superb track record of anticipating unusual things in how his game would be played.

[aston217]

...doesn't have a superb track record of anticipating unusual things in how his game would be played.


This is actually my line of thought.

I said that because I don't think Jim designed the affinities system thinking that guys would use them to land entire teams of aff guys. I think he designed it so that it would have a meaningful effect in typical occurrence - a few affinities here, a few conflicts there.

If this effect scaled linearly, then teams that overload in the manner of say, Camden, would get incredibly overpowering bonuses. Or, normal occurrence of affinities would have no meaningful effect at all, which I don't think makes sense.

Just food for thought, I understand we're more or less throwing darts here.

[QuikSand]

So, we agree that Jim likely didn't envision pure affinity freakshow rosters.

To me, that seems to underpin my working conclusion that the game doesn't include curbs to keep that strategy from being too effective. You seem to reach to opposite conclusion, I presume based on not seeing a ton of evidence that these teams become massively dominant, all else equal.

Here's my bottom line. I think Jim plugged in this system so that a team with a token effort in this - no conflicts, picking up affinity guys where it's easy - might get an aggregate bonus of about 1/2 a game a season. Not enough to make it overpowering, but meaningful. Now, a nutty affinity nest like most of my teams...maybe that becomes 1-2 games, all else equal. Bigger deal, but still not overpowering. It won't turn an average team into a 15-1 juggernaut.

What it could do, in those cases, is turn teams that look on paper like 7-9 win fringe players into 10-12 win division winners. And that's pretty much the station of most of my affinity-focused teams in serious MP leagues.

[MRL17]

FWIW, as a programmer (all of which are inherently lazy) it makes a lot more sense to reuse code than to write several completely different ways to rate teams.

It's pretty likely that chemistry bonuses are given on your team's chemistry rating vs all the other teams in the league in a 0-100 scale just like cohesion and roster rating. The only difference is that we don't have a way to see the table.

Of course, this is just speculation and YMMV

[QuikSand]

...and I'd again speculate that you're missing how FOF works.

I believe that the 100-based ratings like cohesion and roster rating are purely cosmetic way to show what's going on without revealing the specifics too closely. When the game "checks" cohesion, I firmly believe it's not checking the phony 100-based number we see in the summary page, it's checking the real, calculated number from which the 100-based stuff is derived.

Again supporting the general theory that these calculations are simpler rather than more complex, including whatever unseen values there are for chemistry effects.

[aston217]

I agree with Quik on those being purely cosmetic.


Not enough to make it overpowering, but meaningful. Now, a nutty affinity nest like most of my teams...maybe that becomes 1-2 games, all else equal.


You might be right, but a nutty affinity nest is going to have maybe 5,6 times the number of affinities as a reasonable but laid-back affinities team.

You seem to reach to opposite conclusion, I presume based on not seeing a ton of evidence that these teams become massively dominant, all else equal.

I think this is my main basis for this. Camden in the OSFL is consistently that team with, well, see above screenshot, but while they are quite successful on a yearly basis, they also fared quite a bit better when they had a collection of stars in their prime.

So even with over 10 exceptional affinities and approaching 30 overall yearly, the effects don't appear to hold a candle to talent. Which makes me think it can't just scale linearly with no bounds. Again though, throwing darts :P

[Ben E Lou]

I wonder if the disconnect here is in that aston doesn't play with the "big boys" when it comes to chemistry. That team is ok, but nowhere near an "affinity nest" compared to many QuikSand teams (heck, or even some of my recent teams). They have only 12 Exceptionals, and they have single-affinity QBs. My GML team, for example, was 23rd in roster strength, but with 18 Exceptionals, a triple-affinity starting QB, and no game planning allowed, that below-average-in-talent squad went 11-5 in the most recent season and was one point away from leading the league in scoring. I've seen teams where every single non-starter was a strong or exceptional affinity (and quite a few of the starters). You've got to work chemistry much harder than what I see in that screenie to see the kinds of results that some of us have seen.

[aston217]

you guys are probably right. I've only ever had the chance to observe one team with any sort of affinity nest, and while they've had 3x QBs in the past, they also had some better talent when they were winning 14 games and going bowling.

I remember Ben saying chemistry and cohesion and so forth might just be say, a 10% modifier, but doing everything right stacks up. That's largely why I've always thought of it as something that's good to do, but doesn't override talent and has some kind of ceiling.

If we could edit player birthdays that might make for a cool test.

[Ben E Lou]

I don't think Quik is saying that it overrides talent. He stated clearly that it won't turn an average team into a juggernaut. Maybe the other disconnect is that it's not and either/or situation. It's very possible to build a very talented team that also has outstanding chemistry.

[aston217]

Hm...I guess my question would be, if even 18 exceptionals and a 3x QB doesn't start overriding talent, then a normal affinity loadout that is maybe 1/10th as good, that barely makes a footprint, right? (all other things being equal)

As in you may as well not stretch your roster to keep that extra affinity or three on your roster unless you are really going to go hard after it. Go hard or go home.

[MalcPow]

I'm generally too lazy (or too prone to miss an export and have the AI shuffle things to crap) for a real commitment to this stuff. But there's no doubt in my mind that the affinity loading approach of others can turn otherwise mediocre teams into division winners.

To paraphrase a tips thread somewhere else, there really is no rational reason for your 32/32 backup whatever to not be an affinity, and doubly so for inactive guys.

[cuervo72]

Yeah, I'm too lazy too. I'm looking at my team now and I just think - "wow, there is just so much of a mix of affinity groups, it would take me years to get where Quik is." Then I figure eh, why bother? (That, and I typically have little taste for FA). Of course that's a reason why 2/3 of the time, Quik's team wins the division.

It's not a screencap, but the color uniformity here shows what we're up against: fof-IHOF.com : Personality (http://www.fof-ihof.com/teampages/personality.php?t=29)

[QuikSand]

I couldn't easily find this reference here, so I'm posting a custom metric that got developed on a MP league forum elsewhere, but basically summarizes my view of things:

AffIndex = This custom affinity index weights the various affinity factors from each team – each affinity or conflict is included, and weighted by strength. It is designed so that a completely random effect would yield a zero – either absolutely no chemistry effects, or else two affinities for every conflict. The actual weightings I used are as follows: Mild Affinity +2, Affinity +3, Strong Affinity +4, Exceptional Affinity +5, and QB Affinity +2 --with corresponding conflicts each twice as strong in the negative. I don’t have a lot to go on that this is “correct” but it certainly captures the overall picture for that team’s roster in a single number reasonably well.

[ZappBrannigan]

AffIndex = This custom affinity index weights the various affinity factors from each team – each affinity or conflict is included, and weighted by strength. It is designed so that a completely random effect would yield a zero – either absolutely no chemistry effects, or else two affinities for every conflict. The actual weightings I used are as follows: Mild Affinity +2, Affinity +3, Strong Affinity +4, Exceptional Affinity +5, and QB Affinity +2 --with corresponding conflicts each twice as strong in the negative. I don’t have a lot to go on that this is “correct” but it certainly captures the overall picture for that team’s roster in a single number reasonably well.

Sorry for the tangent but this is the first time I've heard that conflicts are twice as powerful as affinities. Is this a known quantity?

[aston217]

That wouldn't surprise me, but I think Quik or whoever came up with that did it just as a quick way to quantify "how good" or "how bad" a team's overall affinity situation is at a a glance.

[QuikSand]

Sorry for the tangent but this is the first time I've heard that conflicts are twice as powerful as affinities. Is this a known quantity?

Nothing concrete... but it wouldn't surprise me, either.

The reason I placed it in the calculation that way is, at least in part, for the zero symmetry it creates, which appeals to my common sense as well. So, for a team created completely by dart throwing, you'd expect to have roughly twice as many affinities as conflicts, so the expected result would be zero. That, to me, adds real value to this metric -- a plus is better than random, a minus is worse than random.

[aston217]

Thought I'd ask here, to Quik and the other pros, do you activate your position leaders to get the most out of their affinities? Or does it not matter?

[Pyser]

not positive, but i dont think it matters

[QuikSand]

I don't think active/inactive matters.