JeffHCross Project #1: How is OVR determined?

[JeffHCross]

This is #1 in a series of, for lack of a better description, projects that I want to do before NCAA 12 is released. Hopefully long before NCAA 12 is released. I'm creating these threads to hopefully increase my own incentive to work on the projects.

Think of it as incentive by potential public embarrassment.

Now, PocketScout posted a list of the most important ratings that influence OVR. It was a very good list, and initially I didn't see much reason to expand on it. But his list was, or at least appeared to be, limited to the most important. I want them all.

Between NCAA 10 and 11, I did a lot of work on trying to create Historical Players using statistics. Ideally, I'd like to be able to provide the ability for anyone to create any player, and all the ratings, solely from Height, Weight, 40 time, and a moderate amount of statistics. It wouldn't be perfect, but it'd be light years beyond the method I used.

Anyway, after creating an Excel file of every player's ratings, I should be able to, fairly easily, determine the relationship between individual ratings and OVR.

[steelerfan]

Interesting.

Now hurry up!

[morsdraconis]

Truthfully, I can't believe this hasn't been done yet. I'm so used to PC games like WoW and Diablo 2 where everything is broken down into the smallest amount of statistical data (mainly because they have access to scource code and such) where everything can be calculated to the 3rd decimal.

It just amazes me that someone hasn't taken the time to do it for the Ncaa series (or even Madden).

Let me know if you need any hard data on how many points it takes of a certain stat to increase the overrall as I'll gladly do the leg work for that (as it's always been a curiosity of mine)

[JeffHCross]

I'm so used to PC games like WoW and Diablo 2 where everything is broken down into the smallest amount of statistical data (mainly because they have access to scource code and such) where everything can be calculated to the 3rd decimal.

I'm sure you meant something other than source code ... because Blizzard has never released source code for either of those titles. Now, being that they are on the PC and have modding abilities, there are a lot of aspects of both titles that you can access that you can't see on a console. But, for most games like that, since they are RPG games, there's just a lot of detail given that's not present in sports games. Like you see "20.41 damage" or "Critical Hit" or, depending on the level of detail, you may even see stuff like "Rolled 6, no Hit". That stuff is all being done internally on NCAA and we have no way of seeing it.

As to whether or not someone has done it ... I'm sure PocketScout has. The amount of data he's generated over the last year or so -- I'm sure he's got exactly what I'm trying to do. But I'm not going to ask him for all his numbers. I wouldn't expect him to give it to me, and I wouldn't blame him for not doing so. It's easy enough for me to do it myself, and that way, at the end of the day, I have a confidence in those numbers.

One reason you probably haven't seen it done before is that the game keeps changing up on us. A LB that was 87 OVR on NCAA 2004 would be a 78 OVR on NCAA 07 with the same ratings.

[steelerfan]

Think of it as incentive by potential public embarrassment.

Well.....?

[psusnoop]

Jeff is currently not available at this time.

[JeffHCross]

Well.....?

Still working on it.

[ram29jackson]

Interesting.

Now hurry up!

Well.....?

and I thought I was annoying ?

[steelerfan]

and I thought I was annoying ?

You are.

I'm just messing with Jeff because his plate is full with being outed as a Community Event guy.

[ram29jackson]

You are.

I'm just messing with Jeff because his plate is full with being outed as a Community Event guy.

I know, I know,I am. I just try to be thorough with questions and see what answers I can get. its just hard at times trying to type a dialogue/question as opposed to asking someone in person.

[souljahbill]

Jeff, did you think to ask this question at the event?

[JeffHCross]

Jeff, did you think to ask this question at the event?

Of sorts. I didn't get a direct answer (and honestly I wouldn't feel comfortable asking for a list or anything like that), but he didn't dodge my question either. More or less, I asked if this was even a worthwhile effort, and he didn't disagree. I did get some insight into how OVR is computed, but it really just confirmed what I already thought, based on the results that PocketScout had provided me.

The biggest holdup on this is that Microsoft Excel is only capable of doing a Regression on limited amounts of variables. So I'm going to have to use a more powerful program to pull this off, and I just haven't had time to set it up.

[Rudy]

I've got a roster file from Pocket Scout Jeff. I may take a look at this on Tuesday at work (I'm off until then). I'll try running a simple regression using SPSS first although I suspect there will be economies of scale. I can run a non linear regression on it as well and that will probably be the only way to determine the true results. I'm assuming I will have to do this on a position by position basis to get any meaningful results.

[JeffHCross]

Yeah, definitely position by position, Rudy. OVR isn't determined the same was for QBs as it is for WRs, I know that for fact.

[Daywalker86]

I'm pretty excited to see how this breaks down.

[Rudy]

I uploaded all of the formulas by position to google docs: https://spreadsheets.google.com/ccc?...mc&hl=en#gid=0

These are based on a pocket scout recruit export file. I get most positions accurate to 97 or 98%. By this I mean I predict the exact overall rating for 97 to 98% of the players. The other 2-3% that are off are only off by one point on the overall so this is highly reliable imo. The dev team might 3 or even 4 decimal places in their formulas. The rounding might also be slightly different or done in each stage to account for the very small discrepancy. Most positions start out with a fairly large negative to start. It's around -55 to -75. They probably use this to adjust the ratings so the average player comes in where they want. From there you just add so many points based on the ratings each position uses.

Some interesting things. I believe Russ Kiniry said awareness is barely used any more in the calculations on the field other than QB. They tend to use the newer ratings for some positions but AWR is still present in every position's overall rating. It's obviously biggest for the QB position. Weight is NEVER used at all.

QB - accuracy and the throwing power are the biggest. Awareness is next and everything else much less. If someone has a 90 accuracy vs. 70 accuracy it will result in a 12 point OVR difference.
HB - break tackle and speed are both equally important. The next level is carry and acceleration. Blocking is ignored in the ratings.
FB - run blocking footwork and strength are easily the two biggest variables
WR - Catch and speed are #1. Then acceleration and then route running and agility.
TE - No position uses more variables to come up with an overall rating - 26 in total! None of the categories dominate but run blocking, speed and catching are at the top. Route running is next. Interesting to note that catch in traffic adjust the TE rating more than the WR rating.
OL - all are slightly different with run and pass blocking categories dominating the top tier. Both guards and centers give greater weight to run blocking while tackles are equally weighted between pass and run blocking.

DL - DTs are done a bit different than DE. Zone coverage matters for DE but not for the DT. The DT has three main categories - strength, block shedding and tackle. While both have a heavy emphasis on tackling, DE downplays block shedding and strength and focuses a bit more on speed and acceleration.
LB - MLB and OLB are similar although there is more emphasis on tackling for the MLB and more emphasis on speed for the OLB. Block shedding and strength are also important. No other position uses PRC more than MLB (I don't know what PRC is).
CB - CB is different than safety. Speed is very important and #1. Man and zone coverage are also important with acceleration being #4.
S - Tackling is #1 for both but more important for SS. Speed and zone coverage are also important. Man coverage is more important for the FS while HPW (?) and block shedding is more important for the SS.

P and K - both very similar with kick power and accuracy making up the entire rating almost. Awareness is the only other category that counts and it's minimal. K values accuracy a little more than power. P values power over accuracy by a touch.

[JeffHCross]

Rudy, that's awesome. Exactly what I was planning to do, but hadn't had the time. Thanks a hell of a lot, man.

[morsdraconis]

LB - MLB and OLB are similar although there is more emphasis on tackling for the MLB and more emphasis on speed for the OLB. Block shedding and strength are also important. No other position uses PRC more than MLB (I don't know what PRC is).

S - Tackling is #1 for both but more important for SS. Speed and zone coverage are also important. Man coverage is more important for the FS while HPW (?) and block shedding is more important for the SS.

PRC = Play Recognition
HPW = Hit Power

[Rudy]

Rudy, that's awesome. Exactly what I was planning to do, but hadn't had the time. Thanks a hell of a lot, man.

No problem.

PRC = Play Recognition
HPW = Hit Power

Thanks. That makes more sense now. I couldn't figure out what those meant at work (it's been very slow lately, this game me some interesting stuff to look at).

[morsdraconis]

Thanks. That makes more sense now. I couldn't figure out what those meant at work (it's been very slow lately, this game me some interesting stuff to look at).

No problem man. Anything that can be done to help figure out the nuts and bolts to how a game figures out something is always interesting to me. I'm all about getting down to the nitty gritty of how something is figured out on the back end so that I can better understand how it works.

Unfortunately, for NCAA Football, figuring out how the overall rating is calculated doesn't really help much in game.