Opinions on this combine correlation level wanted.

[Ben E Lou]

I've set up a couple of pages that might help us form opinions on this. But to remove any bias, I'm not going to reveal what combine correlation I used until people have weighed in on this.

I created a draft class of 1,000 QBs, all of whom had 525 (60) in every possible rating. (Sense Rush, Read Defense, and Two-Minute Offense are not available to be imported, and as such you'll see some wide variety in those bars.)

The scouts with which I did this are very good. Most importantly, the OC is a 93. None of the others are below 60.

A few data points regarding the FOF7 constraints. (http://www.operationsports.com/fofc/showthread.php?t=89290)
29 generated less than 28 on their Solecismic tests.
52 had less than 10 on the Bench Press
41 were above 7.80 on the agility drill.Overall, 112 out of 1000 60/60 QBs are misses for one or more of the constraints.

Extrapolating that out over real-life draft classes, so far I'm seeing roughly 1.3 60/60 or better QBs per draft class, so at this particular correlation, we might expect to see .1456 guys at that level who are below the constraints--or roughly one every 7 years.

I think what you'll see if you take a hard look at the data, for the most part when the combines are bad, the bars are bad as well.

Below is a link to the full draft class. Click on any player and you can see his bars. I'd love some feedback on if this is too much variability, not enough, or just right. (And again, I'm not revealing the combine correlation number I used, so as not to bias that opinion.)

http://ccfl.fof-belco.com/draftclass_correlationtest.php

[Ben E Lou]

Ultimately, the question is this:

Is the amount of variability shown here too much, too little, or just right for an FOF MP league with mostly experienced players?

[A-Husker-4-Life]

What combine accuracy are you using? 50 or 39?

[Ben E Lou]

What combine accuracy are you using? 50 or 39?Dude...But to remove any bias, I'm not going to reveal what combine correlation I used until people have weighed in on this.(And again, I'm not revealing the combine correlation number I used, so as not to bias that opinion.)

[A-Husker-4-Life]

maybe I should learn to read better, sorry Ben I'm an Idiot.

[Ben E Lou]

FYI on the whole 7.2 39 thing, here's a snippet that I posted at the CCFL recently.

There's no FOF7.2-style "39" dropoff where the constraints just go away entirely, even at a combine correlation of 0. It is 100% clear that zero does *not* mean "no correlation." It means "The lowest correlation that the developer thinks we should be allowed to use." Similarly, 100 does not mean "perfect." Even with great scouts and correlation of 100, I still get a few guys who are known to be good, but who have bad bars and combines. It's now the sliding scale that one would expect.

[QuikSand]

I'm not fully certain I follow what you're looking for fro this exercise, but here are a couple observations that may be useful regardless:

-Is QB agility on a whole new scale now? I don't recall guys posting scores down under 7.00 before, but you have a handful here (just from memory, I may have missed something from the latest FOF7 patch I guess).

-Curious that nobody ran an eye-popping dash under 4.4. In my fiddling-around SP career, I had a guy run a 4.36 (he turned out to be worthless as a passer).

The obvious observation is that for guys who came into the grinder identical, there are tons of guys who bear virtually no resemblance to one another in this pot.

[Ben E Lou]

I'm not fully certain I follow what you're looking for fro this exerciseEvery single one of these guys is rated 60 at every bar other then the three that can't be imported, but some of them look different. I want to know if at this combine accuracy setting, the class looks too random, too predictable, or just right. Basically on this "sliding scale" that I mentioned, it works like, say...

(WARNING, COMPLETELY MADE-UP NUMBERS ON THE WAY)
100 correlation: 5% of 60/60 players look like duds
50 correlation: 15% of 60/60 players look like duds
0 correlation: 25% of 60/60 players look like duds
(WARNING, COMPLETELY MADE-UP NUMBERS ABOVE)

So I've picked a number between 0 and 100. Some of these 60/60 players look like duds. Is it the right number, too many, or too few.

-Is QB agility on a whole new scale now?According to the Help file, it correlates to different bars than in the pasts, so quite possibly yes. Quarterbacks - throwing short passes and screens. (Used to be Sense Rush and screen.)

-Curious that nobody ran an eye-popping dash under 4.4. In my fiddling-around SP career, I had a guy run a 4.36 (he turned out to be worthless as a passer).All of them are 60 scramble frequency, so I wouldn't expect them to be that fast.

The obvious observation is that for guys who came into the grinder identical, there are tons of guys who bear virtually no resemblance to one another in this pot.Yeah, that's always been the case with these 1000-player draft files. There are always crazy outliers.

[gstelmack]

First comment: I hope constraints have died a horrible death and we never see any sort of plateau like that again.

20 out of 1000 have bureau grades below 4. 211 out of 1000 have bureau grades below 5. 295 at 6+, 36 at 7+. I think I like this - there should be variability with QBs, especially if with combines and heights there are other factors in QB performance. And some may work well in a system and poorly in others. It just feels like you never really know in the NFL what you have with a QB until you get them in and playing, and how many QBs have been wrecked by an unstable coaching staff or horrible O-Lines so we don't know how good they might have been? I think I'm happy with this.

[garion333]

Two thoughts on Bench that are:

(1) High combine, 7 bench is either a tell he's junk or the number the game uses when it wants to lie to you.

(2) 17 is the high end. Constraint of some sort?

The guys with 7 bench look weird as hell up toward the top. I just don't see how they'd ever quite be so athletic and supposedly throw the ball so well yet be so damn weak. Anyway, either the score is a lie for those guys or the rest is a lie, bars included.

Greg's on point though (shocking, I know), seems like the spread is fairly decent. Once in a blue moon a 2.9 Bureau Rating will be 60 rated! But you want higher rated guys for a much better chance at a legit QB.

[aston217]

Two thoughts:

~10% is *not* trivial. However, knowing that it's only 10% does suggest it's important to still have all the numbers memorized. 67, 7.80, 5.27, 37, 4.59, 4.52, 5.07, the works.

~1% is almost trivial. It's still not, but again, if only 1% of players have more than one violation of the combine constraints it further points to the importance of the numbers. Knowing that combine violations are somewhat rare, but that anything more than a point violation is miniscule in probability, is valuable information.

This does not sound like "all traces of the problem have been eliminated" to me. It strikes me as too little variability in the context of "Isn't FOF2k7 over now?" with the caveat that it's completely possible this is a game engine limitation. For example, perhaps the numbers should simply be embraced, if even at 0 they present themselves in some meaningful way.

I mean, it just looks to me that constraints have not died a death in the game engine, which I thought is what we were promised in 7.2, so I'm truly puzzled by this. Variability, as I'm understanding it, refers explicitly to adherence to silly numbers like 28 and 7.80.

[Ben E Lou]

This isn't about using this thread to make a determination on how well or poorly FOF8 handles this. For all anyone other than me knows, this could be how the constraints look at a 100 correlation (remember, they weren't removed anywhere over 39 in 7.2.) Or it could be how they look at 50. Or 0.

I didn't reveal that information to remove all of this sort of extraneous discussion and deal with the only question that is relevant to me at the moment, and I say it again for emphasis:

Is the amount of variability shown here too much, too little, or just right for an FOF MP league with mostly experienced players?

[gstelmack]

My post was answering that as "just right".

[garion333]

Right, and I was agreeing.

[aston217]

Does "mostly experienced" mean "have memorized all the numbers?"

I would say it's fine if we're expecting people to know the numbers. (Maybe 15% would be better than 10, which is a tad on the low side). It's far too little variability, IMO, if the idea is that constraints shouldn't exist.

[Ben E Lou]

My post was answering that as "just right".
Definitely understood that from your post.

[Sharkn20]

If the managers are experienced I would refer the numbers of variability in regards of finding "rough" diamonds in later rounds at QB, aka Tom Brady, Russell Wilson, even Kirk Cousins :popcorn:

So one every seven years sounds good to me too.

[corbes]

It's an area where there may be a distinction between what's real and what's enjoyable. In a fast-paced sim environment, I think there's a benefit in having the outliers appear somewhat less often than in real life. It helps narrow the pool and focuses the attention on the better candidates and somewhat alleviates the tedium of putting "eyes on" hundreds and hundreds of pieces of hay in search of needles--desirable when you're flying through seasons in SP or cranking through drafts every two weeks in multiple MP leagues. At that pace, realistic variability can begin to feel like randomness.

[TAFIV]

actually the schedule we're looking at for our league is about a month from training camp to the Cereal Bowl (name of our super bowl) then we have Early FA, Draft and Late FA so i think its probably going to end up at about 5 weeks per in-game year

[corbes]

Even in one league it's a lot of work to scout a draft. But my point was that players in multiple leagues are often drafting (in one league or another) every week or every other week.

[Firefly]

I'd aim for 15%, too

[zbuckley]

So once every 7 years there's some starter quality QB like this? http://ccfl.fof-belco.com/testdraftee.php?playerid=3217

I think I could make a case for that being too high and also too low. Because so few QB's are actually worth drafting compared to prior versions I think having anything more than 10% would get pretty frustrating for some players. On the flip side 10% seems pretty low for the let's make FOF look like the real NFL drafting experience.

[Ben E Lou]

So once every 7 years there's some starter quality QB like this? http://ccfl.fof-belco.com/testdraftee.php?playerid=3217Not necessarily like that guy. But one who just "doesn't fit."

I think I could make a case for that being too high and also too low. Because so few QB's are actually worth drafting compared to prior versions I think having anything more than 10% would get pretty frustrating for some players. On the flip side 10% seems pretty low for the let's make FOF look like the real NFL drafting experience.I suspect that if FOF (or any computer game) looked like the real NFL drafting experience, it would be extremely annoying and feel "too random" for a computer game.

I've got some better data posted now. Might want to check this thread: Another way to look at combine accuracy - Front Office Football Central (http://www.operationsports.com/fofc/showthread.php?t=92223)

I should probably close this one.

[garion333]

I suspect that if FOF (or any computer game) looked like the real NFL drafting experience, it would be extremely annoying and feel "too random" for a computer game.

Fwiw, I've seen that exact complaint about FOF7 on 39 combine correlation.

[ezlee2]

Personally, I think moving to the 39 setting was one of the things that made me enjoy this game even more. It became way to predictable to rely solely on combines.

[aston217]

Fwiw, I've seen that exact complaint about FOF7 on 39 combine correlation.

I think my biggest annoyance with the draft is there are 850 players, and a ridiculous number of them look like this:

https://s30.postimg.org/wno2k358x/dechandler.png

A thought would be first, whatever changes are needed to make a much lower class size possible. I never look forward to studying a full draft class; haven't for ages. And I know I'm spending half an hour just going through and marking five or six hundred players out, with a sizable percentage of those being obvious. Why the chore? And it just stops me, period, from even thinking of doing that in SP. The MP draft is at least a competitive venture against human owners.

Second, maybe if we had a sane number of players per position that everyone had any information on, and then the rest would be a grab bag of random. Like, hey, study these 10 QBs everybody and try to figure them out, and then there's these other 10-20 that didn't go to the combine and are just a total moonshot. So draft them if you will, but maybe they're Tom Brady or (much more often) they wash out, in a way that models the behavior of NFL late round/UDFA QBs appropriately enough.

Similarly, the 10 combine QBs would model top NFL draft prospects appropriately enough. Sometimes they're obviously flawed and sometimes they're hidden gems; your prospects for a Russell Wilson are reasonably strong but so too might the guy be Gio Carmazzi. A consensus top prospect has a solid chance of being Cam Newton but a nontrivial chance of being Jamarcus Russell. And so on. Use your knowledge of game mechanics as well as your interviews to try and tease out which risk bucket a player falls in.

Would that be too random? I suppose it might be, but I think it can be balanced enough. In FOF7 and 2k7, drafting was *really* fun but also heavily tilted towards those who could both really crack the code and spend infinite time studying a large number of prospects, which really separates the draft landscape between those who have time to invest and those who don't.