Posted by hlbart on 11/19/2010 9:21:00 PM (view original):
This is the best thread going. Thanks yatzr. Once they finish tweaking the system, I'd appreciate a periodic update!

I'll be running these stats pretty much every night to see if things ever change.  Today's numbers were pretty much exactly the same, even though they said they made changes for the 2nd half.

Just looked at the NCAA FBS real world stats.  Below info was done by adding the team totals (not all 45491 individual rushes).

45491 rushes for 197341 yards for a 4.3 yard rush average. The standard deviation for the team averages is .9 a rush. 

The real world stats for passing are:

37261 attempts, 22292 completions (60%), 267489 yards which gives 7.18 yards per attempt.

Posted by yatzr on 11/19/2010 9:20:00 AM (view original):

Posted by tommy_cian on 11/19/2010 7:04:00 AM (view original):
One last request, and I'm not sure if its even possible (but if it is, that would be awesome!)

Do you have a distribution of running distances?  So, set a counter for how many plays of <-4 yards, -3 to -1 yards, 0 yards, 1-3 yards, 4-10 yards, 11+ yards?

I wonder...

Sure.  This is for designed run plays in Yost on 11/18/10
Yds Gained : # of occurances
-3 : 128
-2 : 833
-1 : 1060
0 : 1336
1 : 2720
2 : 4653
3 : 2527
4 : 1423
5 : 1088
6 : 505
7 : 319
8 : 244
9 : 224
10 : 215
11 : 177
12 : 192
13 : 223
14 : 202
15 : 178
16 : 149
17 : 49
18 : 26
19 : 24
20 : 21
21 : 22
22 : 14
23 : 14
24 : 13
25 : 14
26 : 19
27 : 11
28 : 11
29 : 14
30 : 10
31 : 11
32 : 5
33 : 8
34 : 13
35 : 11
36 : 7
37 : 8
38 : 8
39 : 12
40 : 8
41 : 9
42 : 15
43 : 8
44 : 10
45 : 5
46 : 3
47 : 3
48 : 6
49 : 5
50 : 4
51 : 7
52 : 1
53 : 4
54 : 6
55 : 7
56 : 5
57 : 3
58 : 7
59 : 5
60 : 2
61 : 2
62 : 3
63 : 3
64 : 2
65 : 10
66 : 2
67 : 3
68 : 6
69 : 4
70 : 3
71 : 4
72 : 1
73 : 3
74 : 5
78 : 1
79 : 1
80 : 3
82 : 1
83 : 1
91 : 1
92 : 2
96 : 1
98 : 2

We need something to compare this to, other than anecdotal feelings.  I can not seem to find this for the NCAA ... I did find something for the NFL for:2008:

Comparing Running Performance

This is a chart comparing what I could read (might be inaccurate) from that NFL picture and Yaztr's numbers:

Yards	WIS	NFL
-3	0.68%	1.80%
-2	4.40%	3.50%
-1	5.60%	6.00%
0	7.06%	10.00%
1	14.37%	13.00%
2	24.58%	12.50%
3	13.35%	10.50%
4	7.52%	8.00%
5	5.75%	6.00%
6	2.67%	4.00%
7	1.69%	3.00%
8	1.29%	3.00%
9	1.18%	2.00%
10	1.14%	1.80%
11	0.94%	1.70%

Good link, hughesjr.
http://www.advancednflstats.com/2009/08/comparing-running-performance.html

From my reading on the chart at that link, WIS has the percentages about right. I had no idea that 2 yards was the mode.

It seems that in real life good teams don't turn the ball over and have good QBs. In the past, no one had much incentive to get a good QB and stocked up on RBs. Now, the importance of those positions have flipped overnight, and people have talent in the wrong places. That combined with a turnover problem that JConte is looking into, and you have a situation with upsets happening left and right.

I have to say Yatzr, you do some damm good work. I wish I had 1/10 of your skill level. Being a master at google is not quite the same.

Thanks again for the data yatz.  The averall numbers do not look way off to me, but there are still some questions.

is it also possible that since we had "super rushing teams" that most of our defenses are set for "heavy or always run", which is stopping the run more than it should, but once we start passing more, it will open up the rushing game?  This will take a bit for people to adjust their offenses, and then their defenses, but in short time it will happen... 

When a run occurs, does the engine first decide whether it was a "big play" based on a probability or do big runs happen simply as a result of the position and tackling of the defenders?

Because it seems like the engine may be doing some like Run ---> Choose:  Loss (10% chance), Average (80% chance), Big Play (10% chance); with a defined value range for each category.

Posted by hlbart on 11/20/2010 8:10:00 PM (view original):
Good link, hughesjr.
http://www.advancednflstats.com/2009/08/comparing-running-performance.html

From my reading on the chart at that link, WIS has the percentages about right. I had no idea that 2 yards was the mode.

Again I feel compelled to point out the NFL has far more parity than either GD or college football in general.

Posted by sslodus on 11/21/2010 3:43:00 PM (view original):
When a run occurs, does the engine first decide whether it was a "big play" based on a probability or do big runs happen simply as a result of the position and tackling of the defenders?

Because it seems like the engine may be doing some like Run ---> Choose:  Loss (10% chance), Average (80% chance), Big Play (10% chance); with a defined value range for each category.

That is probably what is happening.  The results from SIMs and humans are way too close together.

 I think that is what I'm seeing. It does no good anymore to sub heavy. Your best RB for example will eventually break 1 or 2 big ones a game.

NCAA D1 as of today:  24 teams averaging 5.0 YPC or greater.
Dobie as of today:  1 team averaging 5.4 YPC.  

All Dobie games under the new engine.  That's 1 team for the entire world.  D1A, D1AA, D2 and D3.  

Posted by dravz on 11/21/2010 3:47:00 PM (view original):

Posted by hlbart on 11/20/2010 8:10:00 PM (view original):
Good link, hughesjr.
http://www.advancednflstats.com/2009/08/comparing-running-performance.html

From my reading on the chart at that link, WIS has the percentages about right. I had no idea that 2 yards was the mode.

Again I feel compelled to point out the NFL has far more parity than either GD or college football in general.

I'd disagree.  I think it is just easier to recognize because of the smaller number of teams.  Instead of the 15-20 team blob in the middle of the NFL, I'm comfortable saying there's a 60-75ish team blob in FCS.  Instead of a handful of great top and bottom end teams, there are 3-4 times as many in college.  But with larger numbers comes more shading between them.

As for the run averages- they need to be higher.   The length of individual possessions needs to average out a little longer as well, there's far too much punting in this engine. But, if the last 20 years of football teach anything it's that it should take a freakish, once in a blue moon alignment of talent to be a title contender running or passing more than 80% of the time.  So there needs to be some sort of built in point of diminishing returns to attempting a highly unbalanced offense.

Of course, all of this needs to be tied to talent, which means they need to work on the distribution of talent.  More shades of gray, fewer standouts, more minor flaws in standouts.  I could even see running a "correction" on player ratings in SIM teams to bring them a little closer to the pack.