Leonidas
11-08-2006, 02:16 PM
The latest in my line of gameplan testing for FoF2K7 is on defense. Some of you may recall my offensive gameplanning, but let me refresh one and all on my conclusions as it plays a big part in how I went about my defensive gameplan testing.
Much like every poker show, now's the time to rehash the rules of the game. I take a team and establish a baseline by running the same team, same season, over and over again for 20 seasons with the AI set to do the gameplanning. This gives me a 320 game sample. I turn the injuries off to ensure the same players are playing every time and eliminate the uncertainty injuries play in the game. Then I tweak the gameplan and run the same team, only with my own gameplan for 20 seasons then compare key data points. I also break down both the baseline and my test runs according to how well the team performs on on those data points when the team wins or loses more games in a season than the average and also when the team scores (or allows scores) above and below the seasonal scoring average. I did that to determine what are the kinds of data that would be indicators for an efficient gameplan.
OK, all that out of the way I need to rehash my offensive findings. I found that changing the running portion of the gameplan had no correlation with improved offensive performance. I've done this many times over (maybe up to 20 test runs now) and not once has a team showed any significant correlation between running the ball better (or worse) and winning (or losing) or scoring more points (or less). In other words, in my study I have not found an instance where any change in the running portion of the gameplan will improve or lessen overall offensive efficiency.
As for passing, I found that completion percentage and yards per attempt both correlate to winning more and scoring more. In fact I have never seen an instance where the yards per completion rate did not support this and have only seen two instances where completion percentage did not, and they were both very close. So from that I concluded to improve the offense you had to come up with the best combination of completing the most passes for the most distance. I came up with an offense that does this very well in every single test run I have done. In other words, my gameplan performs better than the AI gameplan every time I've tested using the lessons learned from this study.
So, based on what I found from offensive gameplanning I figured the way to have a more efficient defensive gameplan was to go for the same things, inhibit completion percentage and reduce completion yardage averages. I also figured since tweaking the running game on offense was pointless, going after it on defense would be equally pointless. Boy was I wrong on both counts.
I decided to reverse engineer the offense I gameplanned and turn it around into a defense. Gave me no significant changes from the AI. I tried tweaking blitz rates and nickel/dime frequencies, nothing. By this time I was pretty flustered so I decided to go for broke. Build an extreme defense and see what kinds of changes it was sure to effect on overall performance and build back from there. Naturally assuming the passing game was the thing to attack I went for an all pass defense. Nothing. Well not nothing. My run defense went to total shit, but the pass stats were static. So I went with an all blitz, all pass defense. Nothing. I went with an all aggressive pass defense. Nothing. I went 50-pass, 50-aggressive pass, nothing. I did absolutely nothing to affect the passing game whatsoever.
Noticing that going whole hog against the pass made the run defense totally suck, I decided what the hell. I can't do anything to the pass, clearly I'm doing something to the run, let's go with it. So I went to radical run defenses, and voila, I had success. I didn't do a straight 100 run or 100 aggressive run. I thought it would be pointless especially on 3rd or 4th and longs, so I do have some settings for pass defense and I do have some settings for nickel and dime packages.
My first successful run was with a fictional Vikings team. The baseline is the first number, my radical run D stats are in parentheses.
rush yards per season: 2056 (1830)
yards per carry: 4.21 (3.85)
pass yards per season: 3175 (3206)
completion %: 60.78 (60.76)
yards per completion: 6.64 (6.58)
points per season: 320.7 (308.1)
wins per season: 7.45 (8.80)
The most striking difference was in wins. It was over a 15% improvement. Nothing in my offensive gameplanning was this dramatic. So I thought maybe this is a fluke. I looked more carefully at the numbers and found a 40% increase in interceptions between winning and losing seasons. Interceptions are usually pretty static on these tests so I thought maybe this was an anomoly where turnovers were getting me more wins, not better defense.
So I had doubts. Therefore I decided to dig out those Cincinnati Bengals I did so much offensive gameplanning with and try this out on their defense. I had the offense set to AI so as not to taint it with my offensive gameplan. Here's what I got.
rush yards per season: 2091 (1807)
yards per carry: 4.29 (3.90)
pass yards per season: 3214 (3157)
completion %: 62.2 (61.34)
yards per completion: 6.53 (6.44)
points per season: 307.15 (276.9)
wins per season: 8.55 (10.6)
The improvements were more pronounced, the win differential even larger. In my offensive tests a really good run eked out maybe half a win a season more. In this defense testing I'm seeing 1-2 full win improvements. Something else I'm noticing is the more efficient seasons have considerably fewer plays run against them. That could be a turnover issue, as in this defense causes more, but I doubt it as I had fewer int's in my gameplan on the Bengals run. So my next theory is that by inhibiting the run I am also controlling the ball more. I intend to do a full on test of my offense and defense together and see if that is the case.
So far (I want to do at least two more runs to kill all my doubts) my preliminary conclusions on defense are 180 out with offensive gameplanning. In the offense the run factors are useless and you have to attack the pass. On defense I could find no effective way to inhibit the pass, but going strong after the run does very well overall for the defense.
Nagging me in the back of my head is the feeling the defensive gameplanning could get totally killed by a pass happy human opponent. I think my offense will do well on any level, but committing so strongly to the run kind of bothers me. I have to think my own pass offense would kill my own run defense. Maybe I'll need to set up my own personal online league and test that out.
Much like every poker show, now's the time to rehash the rules of the game. I take a team and establish a baseline by running the same team, same season, over and over again for 20 seasons with the AI set to do the gameplanning. This gives me a 320 game sample. I turn the injuries off to ensure the same players are playing every time and eliminate the uncertainty injuries play in the game. Then I tweak the gameplan and run the same team, only with my own gameplan for 20 seasons then compare key data points. I also break down both the baseline and my test runs according to how well the team performs on on those data points when the team wins or loses more games in a season than the average and also when the team scores (or allows scores) above and below the seasonal scoring average. I did that to determine what are the kinds of data that would be indicators for an efficient gameplan.
OK, all that out of the way I need to rehash my offensive findings. I found that changing the running portion of the gameplan had no correlation with improved offensive performance. I've done this many times over (maybe up to 20 test runs now) and not once has a team showed any significant correlation between running the ball better (or worse) and winning (or losing) or scoring more points (or less). In other words, in my study I have not found an instance where any change in the running portion of the gameplan will improve or lessen overall offensive efficiency.
As for passing, I found that completion percentage and yards per attempt both correlate to winning more and scoring more. In fact I have never seen an instance where the yards per completion rate did not support this and have only seen two instances where completion percentage did not, and they were both very close. So from that I concluded to improve the offense you had to come up with the best combination of completing the most passes for the most distance. I came up with an offense that does this very well in every single test run I have done. In other words, my gameplan performs better than the AI gameplan every time I've tested using the lessons learned from this study.
So, based on what I found from offensive gameplanning I figured the way to have a more efficient defensive gameplan was to go for the same things, inhibit completion percentage and reduce completion yardage averages. I also figured since tweaking the running game on offense was pointless, going after it on defense would be equally pointless. Boy was I wrong on both counts.
I decided to reverse engineer the offense I gameplanned and turn it around into a defense. Gave me no significant changes from the AI. I tried tweaking blitz rates and nickel/dime frequencies, nothing. By this time I was pretty flustered so I decided to go for broke. Build an extreme defense and see what kinds of changes it was sure to effect on overall performance and build back from there. Naturally assuming the passing game was the thing to attack I went for an all pass defense. Nothing. Well not nothing. My run defense went to total shit, but the pass stats were static. So I went with an all blitz, all pass defense. Nothing. I went with an all aggressive pass defense. Nothing. I went 50-pass, 50-aggressive pass, nothing. I did absolutely nothing to affect the passing game whatsoever.
Noticing that going whole hog against the pass made the run defense totally suck, I decided what the hell. I can't do anything to the pass, clearly I'm doing something to the run, let's go with it. So I went to radical run defenses, and voila, I had success. I didn't do a straight 100 run or 100 aggressive run. I thought it would be pointless especially on 3rd or 4th and longs, so I do have some settings for pass defense and I do have some settings for nickel and dime packages.
My first successful run was with a fictional Vikings team. The baseline is the first number, my radical run D stats are in parentheses.
rush yards per season: 2056 (1830)
yards per carry: 4.21 (3.85)
pass yards per season: 3175 (3206)
completion %: 60.78 (60.76)
yards per completion: 6.64 (6.58)
points per season: 320.7 (308.1)
wins per season: 7.45 (8.80)
The most striking difference was in wins. It was over a 15% improvement. Nothing in my offensive gameplanning was this dramatic. So I thought maybe this is a fluke. I looked more carefully at the numbers and found a 40% increase in interceptions between winning and losing seasons. Interceptions are usually pretty static on these tests so I thought maybe this was an anomoly where turnovers were getting me more wins, not better defense.
So I had doubts. Therefore I decided to dig out those Cincinnati Bengals I did so much offensive gameplanning with and try this out on their defense. I had the offense set to AI so as not to taint it with my offensive gameplan. Here's what I got.
rush yards per season: 2091 (1807)
yards per carry: 4.29 (3.90)
pass yards per season: 3214 (3157)
completion %: 62.2 (61.34)
yards per completion: 6.53 (6.44)
points per season: 307.15 (276.9)
wins per season: 8.55 (10.6)
The improvements were more pronounced, the win differential even larger. In my offensive tests a really good run eked out maybe half a win a season more. In this defense testing I'm seeing 1-2 full win improvements. Something else I'm noticing is the more efficient seasons have considerably fewer plays run against them. That could be a turnover issue, as in this defense causes more, but I doubt it as I had fewer int's in my gameplan on the Bengals run. So my next theory is that by inhibiting the run I am also controlling the ball more. I intend to do a full on test of my offense and defense together and see if that is the case.
So far (I want to do at least two more runs to kill all my doubts) my preliminary conclusions on defense are 180 out with offensive gameplanning. In the offense the run factors are useless and you have to attack the pass. On defense I could find no effective way to inhibit the pass, but going strong after the run does very well overall for the defense.
Nagging me in the back of my head is the feeling the defensive gameplanning could get totally killed by a pass happy human opponent. I think my offense will do well on any level, but committing so strongly to the run kind of bothers me. I have to think my own pass offense would kill my own run defense. Maybe I'll need to set up my own personal online league and test that out.