A look at 4th quarter scoring and other keys to victory

[ibdb]

It's been widely accepted in previous posts that there are more points scored in the 4th Quarter than any other. I was curious whether this was a streetlight effect, or whether there was really anything to it. Since I'm a stats geek (one of the things that first attracted me to this game) with some time to kill, here's what I found.

I setup a test season using a player distribution draft. The human team set the computer to manage everything and accepted all defaults. One complete season was simulated, and point distribution was recorded for each team (I'll assume that a 248 game sample is large enough not to need multiple iterations). The percentages are reported below:


Point Distribution

/ 1st Q/ 2nd Q/ 3rd Q/ 4th Q

Median/ 20.00%/ 26.16%/ 19.25%/ 31.68%
Min./ 13.21%/ 11.74%/ 12.50%/ 21.16%
Max./ 30.99%/ 39.43%/ 28.62%/ 44.13%


Teams did score a larger percentage of their points in the 4th Quarter than any other. The team with the highest 4th Quarter scoring percentage scored 44.13% of their points on the season in 4th Quarter rallies, and the league median was 31.68% of points scored in the 4th.

The next question is, does it matter? Offensive point distribution did not significantly correlate with winning percentage, no matter where those points were distributed. Teams that scored a lot early were no more likely to win than teams that scored a lot late.

What did correlate with winning percentage?


Offensive Performance

/ Pass Yds/ Pass Avg/ Rush Yds

Win %/ 0.38/ 0.51/ 0.32

(A correlation of +/- .30 or greater is statistically significant at the 5% level. Positive results indicate a positive relationship -- more passing yards correlate with a higher winning percentage.)


Not surprisingly, the ability to pass effectively mattered most. The strongest relationship with winning percentage was average yards per pass. Total yards passing was also significant, and total rushing yards just passed the significance test. Average yards per carry was not significant.

Defensive point distribution did correlate significantly with winning percentage, but not in the expected direction. The strongest correlation was between a high percentage of 4th Quarter points allowed and a high winning percentage.


Defensive Performance

/ 4th Q/ Pass Yds/ Pass Avg

Win %/ 0.47/ -0.42/ -0.39

(A correlation of +/- .30 or greater is statistically significant at the 5% level. Negative results indicate a negative relationship -- more passing yards correlate with a lower winning percentage.)


Of special note here is that there was no significant relationship between rushing defense and winning percentage. In FOF2K1, the passing offense is king. If you can throw effectively, you can win. If you can't stop your opponent from throwing effectively, you can't. A running game only matters in so far as it helps you succeed with the pass.

------------------
The city of Houston was named after Sam, not Dean.

[This message has been edited by ibdb (edited 02-08-2001).]

[This message has been edited by ibdb (edited 02-08-2001).]

[Morgado]

Originally posted by ibdb:
Offensive Performance

/ Pass Yds/ Pass Avg/ Rush Yds

Win %/ 0.38/ 0.51/ 0.32
(A correlation of +/- .30 or greater is statistically significant at the 5% level. Positive results indicate a positive relationship -- more passing yards correlate with a higher winning percentage.)

Not surprisingly, the ability to pass effectively mattered most. The strongest relationship with winning percentage was average yards per pass. Total yards passing was also significant, and total rushing yards just passed the significance test. Average yards per carry was not significant.

Okay, i'm not sure I quite understand what your regression looked like and what you mean by 0.3 as a 5% critical value. What was the R-square on the regression? I'm guessing you're doing something like:

Wins = a + b(PassY) + c(PassAvg) + d(RushY) + e

Where e is an iid random disturbance. What i'm not so sure about is the choice of instruments. Did you use more regressors i'm leaving out?

I think what I find strange is that all of this totally contradicts everything i've been doing in WLC and other FOF careers while developing the 75 Offense. Most years when I won the NFC Championship or the Super Bowl, my team was near the top in rushing, run defense and turnover margin but very low in pass defense and pass offense. I'll go along with the YPA as being a key coefficient - I totally believe that passing efficiency is the real deal as opposed to passing yardage. As Quik and others pointed out in another thread, even though my pass offense didn't generate much yardage it had great YPA and starting QB passer efficiency rating (mostly due to high completion rate).

Defensive point distribution did correlate significantly with winning percentage, but not in the expected direction. The strongest correlation was between a high percentage of 4th Quarter points allowed and a high winning percentage.

Actually, I would argue that this is the expected sign of the coefficient because a team that is behind will likely be throwing often and long in hopes of scoring. CPU coaches in particular seem to have a penchant for getting desperate fast and launching the bomb in the fourth quarter. This leads to big plays when they get lucky and INTs when they're not. If a team has a dominant QB, then at least a reasonable percentage of these bombs end up completed - and it only takes one to score. On the other hand, when the thing ends up as an INT, it's no worse than a punt if the DB doesn't run it back. Then the CPU coaches' other 4th quarter tendency takes over (run the clock when ahead). Since the losing CPU coach knows this, it will stack the line and sometimes force 3 and outs - leading to more bomb attempts.

Add in the fact that by the fourth quarter at least one starting DB and at least one starting LB is typically injured (even if it's a phantom 0-1 week injury) and that completion likelihood seems to be more QB dependent than anything else. Plus too, the winning CPU coach will often call Nickel, Dime, and Prevent packages that take the defenses' best players off the field (maxed out LBs) and puts in extremely bad third string DBs instead. It makes sense that the more frequent bomb attempts against possibly a thinner defense will result in higher point generation for teams that are behind.

Edit: Just to clarify here, what i'm guessing is that yes the team that's behind is scoring a lot of points, but usually this is in the face of an insurmountable lead (like down 35-10 with 5 minutes to go). Since the team that is ahead isn't looking to score, they try to prolong the drive and eat up clock. Like the Hidden Game of Football authors say... there's three factors in judging the situation of the game: clock, scoreboard, and field position. If one team (behind) is going after scoreboard while the other team's goal in the fourth quarter is clock, then I don't find it too surprising that the scoreboard pursuing team wins in that department. What would really gong my line is if the teams that are behind are also dominating in fourth quarter time of possession.

I think Lombardi said something like they never lost games, but just ran out of time.

Defensive Performance

/ 4th Q/ Pass Yds/ Pass Avg

Win %/ 0.47/ -0.42/ -0.39
(A correlation of +/- .30 or greater is statistically significant at the 5% level. Negative results indicate a negative relationship -- more passing yards correlate with a lower winning percentage.)

Of special note here is that there was no significant relationship between rushing defense and winning percentage. In FOF2K1, the passing offense is king. If you can throw effectively, you can win. If you can't stop your opponent from throwing effectively, you can't. A running game only matters in so far as it helps you succeed with the pass.

I think this is by far the most surprising of your results. Usually I read low rushing attempts and yardage totals as an indicator that a team is behind (so the opposing defense gets good run defense stats). In fact, my team was usually ranked in the lower half of the league in passing yards allowed per game due to the "throw from behind" effect. Sometime in the 3rd quarter, teams usually just gave up trying to establish the run and threw every down.

Certainly we've seen Midwest 35 type offenses that dominate the league... especially with a dominant QB like say Daimyo's QB Summers, but i'm not sure i'm entirely prepared to throw out all the experiences Subby and I have compiled on "line of scrimmage" focused teams. About the best I can come up with is that maybe there's misspecification in the regressions somewhere... heh. Definitely some good points in here i'll be thinking about.

------------------
The 64 Dollar Question: What *is* The FOF Journal?

[This message has been edited by Morgado (edited 02-08-2001).]

[Daimyo]

I would have to say I agree with the conclusions on first glance. Heading into the WLC I always assumed you needed a rushing YPC>4 to dominate on offense. In my early years of the Super Bowl runs my rushing offense and passing offense were both good and my rushing and passing defense were also good. However, in the last few years my rushing YPC has dropped heavily on offense and risen on defense and yet I didn't see any difference at all in wins or game-to-game stats. The one constant on my offense during the run has been that the passing offense has produced about 8.5 YPA and the passing defense has generally allowed <6.5 YPA. I think this is an interesting topic and I might run some "tests" as well to see what I can come up with.

[Daimyo]

So far what I'm finding is that offensive pass completion percentage is having the strongest correlation with wins. I too am finding very small correlations between rushing YPC and wins.

[FishFan]

shameless wigfecta attempt...

[sooner333]

And sooner333 goes in to make FishFan's wigfecta short lived. Wigfecta block enabled.

[MIJB#19]

Originally posted by FishFan:
shameless wigfecta attempt...

Funny to know that Wignasty was NOT the first person to complete a wigfecta.
I wonder which Dutch person was the first at this forum...

[QuikSand]

Getting back on topic for a moment...

Even if we take these correlations at face value, recall these are based on the successes of the CPU teams in this game.

I know that every time I have occasion to peruse a roster of any CPU-managed team. I find lots of players being used actively who I would never consider for my team. I'm assuming this is generally true, even within a universe of hand-crafted NFL players.

If that's so, I wonder if there is a disconnect between "what works" for CPU teams and "what works" for human-managed teams in a league of CPU opposition? Does looking at the entre range of possible personnel decisions (end oter decisions), even including things that a sensible human manger would never consider, have a distorting effect on the utility of this data for our own purposes?

[Morgado]

Originally posted by QuikSand:
Getting back on topic for a moment...

Even if we take these correlations at face value, recall these are based on the successes of the CPU teams in this game.

Took the words right of my mouth, Quik. I did some thinking on this while playing Armada and came to pretty much the same conclusion.

------------------
The 64 Dollar Question: What *is* The FOF Journal?

[This message has been edited by Morgado (edited 02-09-2001).]

[ibdb]

On the methodology:

The numbers reported are simple correlations. I was more curious to see if there was anything there worth further consideration than I was in fitting the data to a curve. All I was looking for was a simple, if X increases, does the incidence of Y increase at a corresponding rate. The .3 or greater/5% can be read as saying that if the correlation coefficient is .3 or larger, there's less than a 5% chance that the correlation observed would occur at random.

On the computer/human issue:

I agree that we use players and develop strategies in ways that differ significantly from the AI. I'd started out thinking about this more in terms of what I could gain offensively, but it may be better to read this as what I should setup a defense to face. If computer success is based on the pass, what defensive strategy best combats that?

There's lots of other ways I can look at this data still, so maybe something else will show up on further study that will help make sense of this.

On the other hand, I am assuming that the game I'm studying would create fairly normal results, and not have enough random variance to make this all an attempt to explain away the grassy knoll.

------------------
The city of Houston was named after Sam, not Dean.

[QuikSand]

I wonder if there would be enough data from, say, the assembled sets of WLC results to draw some conclusions about "keys to success" among human-managed teams?

With most everyone using essentially the same rules, and various strategies, luck, and skill being variable-- we'd probably see a fair representation of the reasonable range of human-controlled teams.

Do any of you stat junkies see anything worthwhile in there? Would the data be reasonably easy to assemble?

[Morgado]

Originally posted by QuikSand:
I wonder if there would be enough data from, say, the assembled sets of WLC results to draw some conclusions about "keys to success" among human-managed teams?

I think a fair statement to make about the WLC teams in general is that the human players are probably placing a much greater emphasis on defense than the computer AI does. The key factor being the holistic approach to defense that the human players are taking - every unit of the defense is kept in fairly good shape. You don't ever see someone who gets the all-universe secondary but allows his defensive line to go to the crapper. When you scan computer controlled rosters, you can often see distinct "strengths" and "weaknesses" in their defensive lineup. Often times i'll take a peek at a computer roster and see two fabulous LBs with maybe 3 decent ones and a rookie to make a great LB corps and maybe a dominant safety with two great shutdown corners. BUT that same team might have one great DE and the rest of the DL across the board is total garbage.

Part of the problem, naturally, is the fact that the CPU teams overpay their undertalented veteran backups, leaving very little room for depth, much less simply respectable starters at all places. The result is that their maybe 4 or 5 good players on defense are totally overpaid (especially DEs, CBs, and DTs), leaving little in the defense budget to flesh out the rest of the 11. That creates serious problems when you consider that there's no corresponding "dominant quarterbacking" factor that covers for holes in the lineup.

On offense, CPU teams can often be devastating without good starters everywhere because a dominant quarterback paired with one stupid money receiver and one stupid money RB will usually cover for a lot of deficiencies across the offensive line or lack of depth in the receiver corps.

Just suppose rushing offense is really based off of key blocking like how I think it is. Then suppose that key run blocking is really determined off of your offensive linemen and blocking backs (FB, TE) only. Then if you have really crappy linemen, the chances of getting KRB are greatly diminished unless the offense is doing a Larry Allen type deal (pull him and run behind him on every play). Granted, this could happen, but it isn't likely a "realistic" model hardcoded into FOF's engine.

Now if the line has say 2 or 3 really really good linemen but 2 really awful ones, then you'll probably see higher KRB per player for the 3 great ones but way less for the two junk ones. On balance i'm guessing this will probably turn out on the short end of the stick if all our BS about balance in the offense is true (balance in run direction... etc). Since RBs almost never break huge runs unless there is a KRB (possible, but not likely), chances are the CPU run teams with "holey" offensive lines aren't going to produce as good a rushing total as a human team with a solid line, given the same running back.

For the passing game, i'm thinking that the offensive line and blocking matters, but only in determining pressures and sacks. If a QB's ability figures in somehow (Marino's knack for avoiding the sack), then the dominant QB might even be masking for deficiencies in them. I don't know if i'm right or wrong, but my best guess is that completions are two part deals. The more impartant part is an accurate throw, which is at the QB end. The second part is that given the ball gets there, does the WR catch, WR drop, defender bat away, or defender intercept? This second part is probably a function of WR 3rd down, WR catching, defender's INT, and defender's coverage skill dependent on coverage scheme called. Modified then, by the offensive set, pass route, etc... etc...

But if there's a great QB who can consistently deliver the ball, then chances are the receiver will in fact beat the defender. This is especially true if the QB is checking off from the double coverage (which would be some function of the dominant QB). As a debatable opinion, I think there's more quality receivers in the game than there are quality pass defenders - especially on a per team basis. Since you have to consider that the FB, RB, WR, and TE talent pools are all contributing into the "receiver" pool but S, CB, and LB are the only ones contributing into the "defender" pool, I think this makes sense (4 positions vs. 3). One might argue that RBs and Fbs aren't particularly weighted heavily by the AI for catching, but neither are LBs for pass defese so I think it's a wash.

n the end, the type of roster schemes that the AI will come up, probably gives itself more to making a passing offense work than a running offense, and also to a spotty defense. I'm not sure how the AI ranks positions by importance, but the draft tendencies of AI teams should be a fairly good indicator. Notice that the AI tends to draft QB, RB, DE, T extremely high. Dominant quarterbacking i've already talked about, but regarding RB and DE.... haven't we had a lot of discussions on how these aren't the crucial elements of a running game (offensive line) or a good defense (S, DT) anyway? As for T, they don't get nearly as many KRO as C, so they're not keying a running game. They may however, be more important to pass blocking since the DE is primary pass rusher and the T is the one head up on the end. So this would tend to make the CPU teams drafting/signing T with high priority better suited to pass blocking than run blocking.

On the other hand, most of the human players are reporting All-Pro C and G on a regular basis. Also, I think we've managed to convert everyone to the DT and S theory. If you look at the human players, I think you'll find aggressive talent hunting in areas that the AI really does not pursue as hard (S, DT, FB, TE, G, C, OLB) as maybe it should.

All that rolled up with the salary bug likely leads to all the successful CPU teams looking pretty much alike: they throw and need to stop the throw since only the human player can really run the ball anyway. CPU teams might have amazing rushing, but only when they have a dominant RB and by coincidence of overlapping contracts have an across the board solid OL.

Or not. I could be totally wrong...

Heh.... anyone else have a theory?

------------------
The 64 Dollar Question: What *is* The FOF Journal?

[This message has been edited by Morgado (edited 02-09-2001).]

[ez]

Originally posted by Morgado:
Heh.... anyone else have a theory?

cut & paste from your "variation notes thread":

--snip--
the simplest approach to offensive success in FOF is to hire a coach with VG/EXC playcalling, and then go balanced across the board (game plan and talent). with no imbalances to discover or exploit, your guys will simply beat their guys...
--snip--

i'll extend that to include defensive success, and stand by it.

[Morgado]

Originally posted by ez:
...i'll extend that to include defensive success, and stand by it.

Hrm, I really need a shave. I just know I left Occam's Razor lying around here somewhere...

Combining Quik's note on human rosters vs. CPU rosters and that line of thinking looks pretty good. I think I confuse myself sometimes by speculating about Jim's engine...



------------------
The 64 Dollar Question: What *is* The FOF Journal?

[MartinD]

Originally posted by Morgado:
The key factor being the holistic approach to defense that the human players are taking - every unit of the defense is kept in fairly good shape. You don't ever see someone who gets the all-universe secondary but allows his defensive line to go to the crapper.

Have you looked at the Thunder Walkers' roster recently? Great secondary and linebackers, reasonable defensive tackles, but no quality to speak of at defensive end!

Martin

[Morgado]

Originally posted by MartinD:
Have you looked at the Thunder Walkers' roster recently? Great secondary and linebackers, reasonable defensive tackles, but no quality to speak of at defensive end!

Martin

Generally speaking, i've always felt DEs were strictly pass rushers and don't contribute much to run defense. So if you aren't stressing sack-mongering, then DEs aren't all that important. Most of the careers I play, DE is really an afterthought and the DL is always held down eclusively by quality at the DE position. The years I do manage to somehow find better-than-average DEs is when the defense goes from spectacular to legendary, so I guess they have some value, but not really that much. It's the DTs that really count.

------------------
The 64 Dollar Question: What *is* The FOF Journal?

[Daimyo]

I was bored today at work so I simmed through abotu 15 seasons letting the computer handle EVERYTHING for my team. At the end of each season I pasted the general team statistics into excel and then looked for any correlations to wins.

I know there was some discussion on the validity of this type of research, but the way I see it, human teams in the first few years of expansion generally resemble computer teams talent wise. Sure, human teams are much more loaded after a few years, but I think any strategy will work when all the players are talented. I'm much more interested in how to get performance with the minimum of talent and so I think studying the computer teams is a decent place to start.

Anyway, I compared the following team statistic to wins:

OffRush (# of rushing attempts on offense)
OffYards (rushing yards)
OffYPC (rushing yards per carry)
OffAtt (pass attempts)
OffComp (completions)
OffComp% (completion %)
OffYards (passing yards)
OffYPA (passing yards per atempt)
DefRush (number of rushes by opponent)
DefYards (rushing yards allowed)
DefYPC (rushing yards allowed per carry)
DefAtt (passing attempts by opponent)
DefComp (completions by opp.)
DefYards (passing yards allowed)
DefYPA (passing yards allowed per attempt)
Stadium (stadium value)
Roster (roster value)
Support
Economy
Franchise Value
PF (points scored)
PA (points allowed)

Of these items, the ones I found to correlate the best (R^2>0.20) were
OffRush (r^2=0.32)
OffComp% (0.24)
DefRush (0.43)
DefYards rushing (0.21)
PF (0.52)
PA (0.42)

The ones that seemed to make no difference (r^2<0.05> were:
OffYPC (r^2=0.014)
OffAtt (0.010)
OffComp (0.034)
OffYards passing (0.045)
DefYPC (0.016)
DefAtt (0.044)
DefYards passing (0.0088)
Stadium (0.010)
Economy (0.0071)
Franchise Value (0.026)

Finally, the items that seemed to have a slight effect:
OffYards rushing (r^2=0.17)
OffYPA (0.12)
DefComp (0.072)
DefYPA (0.060)
Roster (0.14)
Support (0.089)


Conclusions:
It seems the syle of offense favored by the computer seems to be similar to Morgado's 75 offense. Lots of rushes, high percentage passing, and stopping the opponents run seem to lead to more wins. However since it is already known that the team with the lead late in the game will rush more than usual, this could account for the correlation between rushes and wins so this research in no way meant to imply causation.

[QuikSand]

You saved me time with your last sentence.

I'd have to think that a good deal of any corelation between team wins and things like: # of rushes, # of opponents rushes, and (to a lesser degree) # of opponents rushing yards would HAVE to be largely driven by the fact that teams who are trailing tend to throw more often.

[QuikSand]

All that said (top to bottom)... wouldn't we rather live in a world where there was a high correlation of wins with your yards per carry and those of your opponent?

Am I the only one distressed that these stats seem to almost wash out, and end up being only marginally predictive? Again, it could the the fluid nature of in-game situations (you're a great team, you get a 3 TD lead, you pound the ball into 8-man frints the entire second half to eat clock) but I'd really feel that a team that could post a 4.5 ypc for the season ought to be ripping people up, and same for the defense who holds opponents to 3.0 ypc.

Alas.

Oh, I forgot where I was...

*shurg*

When in Rome...

[Daimyo]

I was very surprised by the fact that so many of the statistics I thought to be key had little effect. I also kind of expected the "roster" value to have a much higher correlation.

[Daimyo]

What might be interesting (although beyong my abilities) would be if someone would make a simple program to examine the game logs and only record statistics from situations in which the game was still on the line. Perhaps this could be defined as the score being within 14 points in the first three quarters and within 7 pts in the fourth quarter.

Not sure if the usefulness would outweigh the difficulty (I have little to no programming experience past BASIC myself), but it might give an insight into what is effective in "normal" situations...

[Celeval]

Anecdotal evidence, but how much of the 4th quarter high scoring might be caused by:

Team A is down 10 points, goes for it 4th and 10 from their own 15. Fails.

Team B scores a TD.

This seems to happen on a semi-regular basis in my games, could the tendency to 'go for the kill' rather than running the clock out have something to do with it?

Kevin

[Morgado]

As much as it sounds like i'm continually pimping this book, I really think everyone on FOF Central ought to go out and find a copy of The Hidden Game of Football by Carroll, Palmer, and Thorn. Chapters 5 and 6 are of particular relevance here, when the authors break down what's going on according to the three part situation: clock, scoreboard, and field position. The whole thing about running the clock and throwing from behind makes a lot more sense when you look at situational data as opposed to aggregated boxscore data (as anyone who watches their scoreboard simulated games will attest to).

In chapter 6, they check a game that really highlights the noncorrelation thing that's bothering Quik (and myself, might I add)... Kansas City at the Pittsburgh Steelers, 1986 right before the playoffs. Total yards offense 515 to 171, advantage PIT. First downs, 28-8 advantage PIT. Time of possession 34:58 PIT.

KCY 24, PIT 19.

WTF?!?!?!?!

Kickoff return yardage (each team had 5 KO returns): 179-59 advantage KCY. PIT kicking: 4-5 FG.

So what's the deal here? The problem was that PIT was able to chew up yards like mad (total offense) but couldn't punch it in for TDs (see FG kicking). Field position was way out of whack, as KCY typically had to drive far fewer yards for scores than PIT, explaining at least some of the low yardage totals. What killed PIT was special teams (KCY had a KO return for TD, a blocked punt, and a blocked FG return for TD) and red zone efficiency. While PIT was getting 3s, KCY was getting 7s and in less time (compare a 80 yard offensive march to a block FG return).

Sometimes when you watch the scoreboard simulations, you can actually see this red zone effect come into play. Your defense gives up a 10 yard run here and a 15 yard pass there, but the opponent stalls near your 30 yard line with all kinds of problems. When a team has to settle for 3 (or even punt!) after chewing up yards, it skews the correlation of win probability and yardage production (since PF and PA has a high correlation with wins).

So maybe it isn't so surprising that all these things seem to wash out in the end. After all, we're trying to fit something that is decidedly nonlinear (how performance leads to winning) into a linear regression form. Toss in all kinds of misspecification, omitted variables, and pretest problems and you have a huge mess. Granted, it is the best we can do though, and it's somewhat disappointing we can't get anything meaningful. Heck, most empirical economic academic papers rarely get R-square values greater than 0.40 and are hailed as landmark discoveries (sheyah hehe).

Such are the difficulties of analyzing football, where everything interacts with everything else...

------------------
The 64 Dollar Question: What *is* The FOF Journal?

[ez]

good points, as usual... i went looking through my box of old sports books, 'cause i thought i had the hidden game of football, but only turned up the baseball book by the same guys.

anyway, i did find "inside football 1984", which i suspect was actually only published once. they address the point vs. yard thing immediately, and tout their "points per yard" stat (actually, they calculate it as points per 100 yards - P100Y, which can be calculated for both offenses - OP100Y - and defenses - DP100Y).

the conclusion is that P100Y correlates very well with wins, as it combines "opportunity" with "achievement" (turnovers, red-zone failures, etc.).

anybody got some FOF data they could crunch?

while this stat will generally correlate with success, i wonder how useful it is in planning/coaching...

[Alf]

Thanks guys for all these analysis !

I'm glad I've found this forum :-)

------------------

Alf

[MIJB#19]

My recent experience with FOF2 says that turnovers and special teams are the key to success.
Also, I have to agree with Margado that stopping the opponents run is the second key to success.
I'm still astonished by 3 straight division titles (twice being the AFL #1 team) with the Cleveland Browns in FOF2 in 1999-2001 seasons.

I'd gladly discuss more about "How to win in FOF", but I'm not aware of the difference between FOF2 and FOF2k1 in how the games calculate the outcome of games.

[TheDawgsAreOut]

So, if people think that turnovers are important, why weren't they included in the study?
I think I may undertake a similar study with turnovers and the other things included under team statistics (3rd down %, penalties, etc.). Another thing I want to test is qb rating, though I'm not sure if Quiksand may have addressed this in a previous qb importance thread.

------------------
The best damn announcer the fake FOFL has ever seen! Check out my cd "Who cares who let them out, I know they're out!" in stores everywhere next month.

[Daimyo]

ez: I have the spreadsheet with all the items i listed above. I could easily calculate pts per 100 yards and see how it correlates with wins. Let me download the spreadsheet from my work computer and if i have space on my webserver I'll make it available for all to look at.

[Daimyo]

the spreadsheet

Its saved as in a comma-delimitted format to both make it smaller and compatible with non Excel users.

Basically my team was the Colts, but I let the scout do EVERYTHING. At the end of each season I pprinted all the stats to a text file and imported them into the spreadsheet. There is data for 524 team-seasons which should give pretty reliable data I hope.

[TheDawgsAreOut]

Well, I have put my money where my mouth was. In a position/winning correlation study, I also looked at other team stats not previously mentioned. What I found for r^2:

Turnover Margin: 0.12
3rd Down Conversion % Margin: 0.36
Penalties: 0.00
Red Zone Margin: 0.03

None of these appears to have a particularly large effect on winning percentage. To explain, to find margins, I subtracted defensive 3rd Down Conversion % or Red Zone Points per Appearance from Offensive totals. I figure the two can be combined because they're largely determined from a common source offensively and defensively - the special skills part of training camp. 3rd Down conversion does seem somewhat important. Turnovers and Red Zone ability showed much less correlation than I expected. Penalties did have a slightly negative correlation, but seemed to have very little effect.

------------------
The best damn announcer the fake FOFL has ever seen! Check out my cd "Who cares who let them out, I know they're out!" in stores everywhere next month.