Cohesion study - 6.3

[Hammer]

The purpose of this study is to try to get a picture of the importance of cohesion.

For the control I will sim 10 seasons with the AI plan. Injuries are off.

Secondly I will swop Brady to the Colts and Manning to the Patriots. Then I will repeat the 10 seasons and record the results.

Tom Brady (9 formations)
Peyton Manning (17 formations)



Patriots (Brady) Cohesion is 87 (5th)

Colts (Manning) Cohesion is 100 (1st)

Team winning percentage will be listed, along with passers average per attempt and QB rating.


RAW DATA
Year One
Patriots 13-3. 8.4. 113.1

Colts 14-2. 8.5. 110.1

Year 2
Patriots 14-2. 7.6. 106.5

Colts 16-0. 7.6. 97.5

Year 3
Patriots 14-2. 8.5. 106.4

Colts 13-3. 8.5. 108.8

Year 4
Patriots 8-8. 7.3. 84.7

Colts 16-0. 8.3. 115.4

Year 5
Patriots 13-3. 7.6. 96.1

Colts 13-3. 8.8. 116.0

Year 6
Patriots 12-4. 7.3. 93.8

Colts 16-0. 8.5. 114.3

Year 7
Patriots 11-5. 7.8. 109.3

Colts 15-1. 8.5. 99.9

Year 8
Patriots 12-4. 8.4. 107.6

Colts 14-2. 7.9. 99.2

Year 9
Patriots 12-4. 7.5. 100.3

Colts 15-1. 8.1. 104.6

Year 10
Patriots 12-4. 7.7. 93.0

Colts 15-1. 8.7. 115.0


Observations
From the results I can immediately pick out that that Colts are fairly consistently getting a better record than the Pats, although not always. Brady and Manning themselves are fairly up and down in their results. The Patriots posted 1 season that appears to be a freak, at 8-8.

Now I shall move Manning to the Patriots and Brady to the Colts and repeat the procedure.

As expected the cohesion of both teams drop considerably.

Patriots 67 (25th)
Colts 76 (17th)


RAW DATA
Year 1
Patriots 10-6. 8.2. 108.8

Colts 14-2. 7.6. 99.6

Year 2
Patriots 10-6. 7.9. 97.1

Colts 16-0. 8.9. 122.4

Year 3
Patriots 13-3. 8.4. 104.9

Colts 14-2. 7.7. 99.1

Year 4
Patriots 12-4. 8.3. 102.8

Colts 14-2. 7.3. 97.2

Year 5
Patriots 14-2. 7.9. 102.4

Colts 15-1. 8.7. 114.7

Year 6
Patriots 14-2. 8.3. 104.0

Colts 15-1. 8.4. 107.9

Year 7
Patriots 14-2. 8.3. 105.9

Colts 16-0. 8.8. 115.6

Year 8
Patriots 13-3. 8.1. 103.8

Colts 12-4. 7.6. 98.2

Year 9
Patriots 13-3. 8.5. 117.0

Colts 13-3. 7.7. 110.5

Year 10
Patriots 13-3. 8.4. 108.4

Colts 14-2. 7.3. 97.5


Observations.
Manning does appear to be the better QB on face value, and may have lifted the Patriots. The Colts don't seem to have suffered much with Brady. As with the previous study, I fear cohesion has had no, or almost no impact. Time to get the calculator out.


Control Group (High Cohesion)
Patriots
AVG Wins 12.1
AVG Att. 7.8
AVG Rating 101.1

Colts
AVG Wins 14.3
AVG Att. 8.3
AVG Rating 108.1


Experimental Group (Fairly Low Cohesion)
Patriots
AVG Wins 12.6
AVG Att. 8.2
AVG Rating 105.5

Colts
AVG Wins 14.3
AVG Att. 8.0
AVG Rating 106.3


Results and Conclusion.
The Patriots did a little better with Manning as their QB. The Colts did a little worse with Brady as their QB, although their wins were exactly the same. Put simply, Manning appears to be a better QB than Brady.

In terms of cohesion though, it appears the variable had no measureable impact on the results. Its as if it doesn't exist.

My final conclusion is that cohesion isn't working as it should in the game.

[Stokes36]

Thanks for the info and the time spent!!! :)

[Yoda]

I donno.... Maybe when I get time I will do something similar, but I don't see how swapping 2 of the top QB's in the league would affect much. For me, I would want to see swapping Peyton/Brady with Warner or a similar lower rated QB where the team does well with them.

It's a good study, don't take me wrong, I just don't agree with the sample set. Actually, let me finish up fighting with my wife's webpage and I'll run it.

[Hammer]

As with any study, its great for people to replicate as they see fit. It doesn't prove anything, but its a solid starting point. Look forward to your results. This took me about an hour.

[Yoda]

Ok, for the study I will do a SP league, named Cohesion, and I will be controlling the Colts (the AI will not trade Warner for Manning the other way around). All the AI options will be turned on and injuries will be off.

[Yoda]

Almost done with the control.

[Yoda]

Awesome...the Cards won't trade Warner to the Colts because of his cap hit... bastards.

[Yoda]

ok, 75% done

[Yoda]

http://www.goldenerafootball.com/images/cohesiontest1.jpg

[Yoda]

Colts w/ Manning 14.4 wins per season
Colts w/ Warner 11.5 wins per season

Cards w/ Warner 13.4 wins per season
Card w/ Manning 11.6 winx per season

Manning is clearer a much better QB than Warner.

Manning's bars are as follows:
Screen 85
SP 96
MP 85
LP 96
DP 91
3rd 81
Acc 75
Timing 82
SR 92
RD 96
2m 87
SF 40
KH 84

Warner's
Screen 39
SP 80
MP 26
LP 37
DP 89
3rd 83
Acc 65
Timing 50
SR 44
RD 10
2m 47
SF 17
KH 8

Also, for the record, they are #5 (Manning) and #8 (Warner) on Greta's Most Careful QB's.

[Yoda]

Also, it appears that if you trade the player back, the cohesion hit stays. That might be the next direction to take. To verify the hit stays, then trade the players back. Could actually do that for the whole league actually, to get a wide range of QB's and how they are affected.... but it is WAY to late and I need to get to bed.

[Hammer]

Interesting. I will do a similar study to that within the next couple of days, see what results come up.

[Dutch]

Also, it appears that if you trade the player back, the cohesion hit stays. That might be the next direction to take. To verify the hit stays, then trade the players back. Could actually do that for the whole league actually, to get a wide range of QB's and how they are affected.... but it is WAY to late and I need to get to bed.

In MP I resigned two aging players with 10+ years each on my team after they sat in the free agency pool for a year or two (I think they were on other teams first)...but neither ever provided a cohesion boost according to the Cohesion screen.

[Ben E Lou]

It's pretty clear that cohesion is calculated based on the value in the "Joined Team" field. If someone is cut/traded/etc. and then comes back, that number resets to the latest time he joined the team.

[Hammer]

I was thinking of recording a control group of 6 QBs of varying abilities. Trading them in the pre season, and then back to their original teams. Running the same season 10 times for each scenario. Seems like a good way to control the variables to me.

[Hammer]

I donno.... Maybe when I get time I will do something similar, but I don't see how swapping 2 of the top QB's in the league would affect much. For me, I would want to see swapping Peyton/Brady with Warner or a similar lower rated QB where the team does well with them.

It's a good study, don't take me wrong, I just don't agree with the sample set.


Give that man a cigar. 80% through. The Colts (best team used) are seemingly unaffected by the cohesion hit, or only slightly affected. The Texans (the worst team) have really been squashed by the cohesion hit so far.

My guess is that cohesion doesn't impact elite teams anywhere near as much as weaker teams. This ties in with what Ben said, saying that he felt weak teams lived off of cohesion.

Full results to follow.

[Hammer]

Okay, not as conclusive as I would of hoped. While Houston's wins dropped, the passing game didn't. It may have just been noise.




Subjects

Colts - Peyton Manning (82/82. SR 93) F 18. AV 5th. Cohesion 100.
Steelers - Ben Roethlisberger (75/75. SR 93) F 16. AV 8th. Cohesion 98.
Bengals - Carson Palmer (54/54. SR 39) F 18. AV 2nd. Cohesion 86.
Cardinals - Kurt Warner (46/46. SR 40) F 14. AV N/A. Cohesion 87.
Texans - Matt Schaub (35/48. SR 66) F 11. AV N/A. Cohesion 76.
Eagles - Donovan McNabb (54/54. SR 39) F 11. AV 9th. Cohesion 94.

I am recording sense rush (SR) and Avoid Interceptions (AV) from Greta’s list as they are large variable of a QBs ability. Formations (F) are also listed.

I will now run 5 control season recording Wins, Average per passing attempt and QB rating.

I am looking at average-good teams here. From a selfish point of view, I am more concerned with seeing how cohesion affects teams that are already strong. I also wanted teams that were fairly high in cohesion currently.

Year 1
Colts W12 7.0. 91.4
Steelers W10 7.0. 88.2
Bengals W10. 8.1. 96.2
Cardinals W12 7.4. 88.0
Texans W12 8.0. 91.3
Eagles W 10 6.7. 87.5


Year 2
Colts W15. 8.2. 111.7
Steelers W11. 6.9. 85.9
Bengals W11. 7.7. 100.2
Cardinals W8. 7.6. 92.9
Texans W8. 7.1. 81.2
Eagles W11. 6.6. 87.1


Year 3
Colts W14. 8.1. 96.7
Steelers W11. 7.3. 93.0
Bengals W12. 6.4. 84.4
Cardinals W10. 7.7. 90.4
Texans W8. 6.7. 79.6
Eagles W10. 6.7. 86.1


Year 4
Colts W13. 7.4. 94.6
Steelers W10. 7.1. 96.2
Bengals W8. 7.3. 92.1
Cardinals W12. 8.3. 94.2
Texans W7. 7.4. 80.3
Eagles W10. 6.4. 84.2


Year 5
Colts W14. 7.7. 93.2
Steelers W12. 8.8. 99.6
Bengals W11. 7.9. 92.7
Cardinals W12. 7.5. 90.7
Texans W9. 8.0. 81.7
Eagles W12. 6.7. 88.2.




Experimentation group


Here are the cohesion changes when I have traded the teams their Qbs back.
Colts 100 (1st)-->75 (19th)
Steelers 98 (2nd)-->89 (6th)--> 75 (18th)
Bengals 86 (9th)--> 75 (18th)--> 64 (31st)
Cardinals 87 (8th)--> 81 (12th)--> 66 (25th)
Texans 76 (17th)--> 75 (21st)--> 64 (32nd)
Eagles 94 (4th)--> 72 (24th)


Well that has thrown me. Only the Eagles and the Colts dropped like I expected. Thinking on my feet now, I am going to exchange veterans Wrs between the 4 teams, and then bring them back. Just
to force a further cohesion drop in the passing game. Results posted in the table above.

That’s better, nice big pronounced changes. Here we go.


Year 1
Colts W14. 7.7. 99.5
Steelers W11. 8.0. 94.5
Bengals W9. 7.8. 96.4
Cardinals W9. 6.7. 78.9
Texans W 8. 7.4. 81.6
Eagles W 8. 6.2. 84.7


Year 2
Colts W13. 7.6. 95.7
Steelers W9 7.0. 84.1
Bengals W13 7.0. 93.1
Cardinals W13 7.6. 96.4
Texans W6 7.0. 81.5
Eagles W9 7.1. 92.1


Year 3
Colts W12. 7.2. 94.0
Steelers W10 7.3. 89.5
Bengals W11. 7.4. 94.6
Cardinals W9. 7.6. 92.2.
Texans W6. 7.0. 79.8
Eagles W11. 6.6. 93.1


Year 4
Colts W14. 7.3. 93.8
Steelers W13. 7.6. 99.1
Bengals W7. 6.8. 85.5
Cardinals W10. 8.0. 92.1
Texans W6. 7.1. 82.1
Eagles W9. 6.2. 85.2


Year 5
Colts W16. 8.3. 106.5
Steelers W10. 7.2. 90.3
Bengals W 8. 6.4. 82.2
Cardinals W8. 7.1. 86.3
Texans W5. 7.4. 87.1
Eagles W10. 6.7. 88.1


Observations
The Colts seem totally unaffected by the cohesion drop. The Texans have been floored by it, in terms of wins at least. Perhaps not coincidence these are the strongest and weakest teams respectively.


I will now calculate the averages of the 5 seasons for both groups.

IND High cohesion - Wins 13.6. Avg A 7.7. QB Rating 97.5.
IND Low cohesion - Wins 13.8. Avg A 7.6. QB Rating 97.9

This result provides support to my previous study that cohesion isn’t working at it should. It appears now, that talent makes cohesion a non factor. This is a great team though, it will be interesting to see what happens with your 10/11 win type teams.


HOU High cohesion - Wins 8.8. Avg A 7.4. QB Rating 82.8.
HOU Low Cohesion - Wins 6.2. Avg A 7.2. QB Rating 82.4.

Huge differences in the wins here. Not so much in the passing stats. While this gives up cause for further research it doesn’t provide the cut and dried results I’d hoped for. If the passing game isn’t any worse, you might argue the wins are just noise. But that’s a big difference right there. The Texans are not strictly a bad team though, perhaps things would magnify if we dealt with a very poor team. Perhaps that should be the next study.


PIT High Cohesion - Wins 10.8. Avg A 7.4. QB Rating 92.6.
PIT Low Cohesion - Wins 10.6. Avg A 7.4. QB Rating 91.5.

Very little difference here.

CIN High cohesion - Wins 10.4. Avg A 7.5. QB Rating 93.1
CIN Low cohesion - Wins 9.6. Avg A 7.1. QB Rating 90.4

Seems to be something with the Bengals perhaps.

ARI High cohesion - Wins 10.8. Avg A 7.7. QB rating 91.2
ARI Low Cohesion - Wins 9.8. Avg A 7.4. QB Rating 89.2

Again, it appears we have something with the Cards.

PHI High conhesion - Wins 10.6. Avg A 6.6. QB Rating 86.6
PHI Low cohesion - Wins 9.4. Avg A 6.6 QB Rating 88.6

Odd set of results. Can we blame the passing game for the drop in wins? Is it just “noise” or is the cohesion at work here? I can’t see how we can blame it, as much as I would like to. Open to suggestions.


Results and Conclusion

We are seeing a drop in wins in some cases, but very little change in the vital passing statistics. At this point my feeling is that the study is inconclusive.

If cohesion is at work here, I can’t get my head round why it would be shown in wins, but not in the passing stats. We may just be dealing with noise here in the win column, just not sure at this point. Hopefully I’m not thinking of something really obvious!?!?

Lets remember that these were very large cohesion drops we were dealing with here.

Next study will revolve around weak teams, in the hope of seeing something more obvious in the results.

[TheMeat]

My initial observations on the testing. Need more seasons, 5 is not enough to measure a change in wins (especially) and other stats. When I test I use a macro and do 20+ minimum.

This is very interesting as I had witness a SP career where a drop in cohesion from using manning one year and switching to a 97/97 3rd year QB (who had all the formations I use) made a huge difference. And my WR are 76 and 89 OVR. At least 2 wins a year DROP in results (100seasons per test) with the new QB, who is better than Manning was. BUT the switch occurred over an off-season. I plan to do some of my own testing on this very soon.

The plan:

1. Take a very talented passing team, and run 100 seasons on one of my MP gameplans.

2. Trade QB away and back, run 100 seasons.

3. Trade the rest of the WR/TE away and back, run 100 seasons.

I'll be posting results.

[Hammer]

Totally agree with you on the sample size. Simply can't be arsed. Good luck!

[Hammer]

Testing weaker teams


Houston - Matt Schaub (Cohesion 73 - 16th) TEST Cohesion 60 - 31st
Buffalo - Ryan Fitzpatrick (Cohesion 71 - 17th) TEST Cohesion 62 - 26th
Tampa Bay - Josh Freeman (Cohesion 71 - 18th) TEST Cohesion 62 - 28th
Washington - Jason Campbell (Cohesion 79 - 12th) TEST Cohesion 64 - 24th
St. Louis - Marc Bulger (Cohesion 81 - 9th) TEST Cohesion 64 - 25th
Seattle - Seneca Wallace(Cohesion 77 - 15th) TEST Cohesion 59 - 32nd



Control group raw data
Houston W10. 6.8. 84.5
Buffalo W7. 5.8. 71.1
Tampa Bay W4. 6.2. 76.5
Washington W7. 7.0. 81.3
St. Louis W6. 6.6. 65.2
Seattle W9. 5.9. 75.4



Houston W9. 6.5. 75.8
Buffalo W9. 5.9. 80.1
Tampa Bay W4. 6.2. 76.7
Washington W8. 6.8. 79.4
St. Louis W8. 7.4. 80.8
Seattle W2. 5.8. 66.0



Houston W8. 6.8. 84.2
Buffalo W4. 5.8. 71.9
Tampa Bay W4. 6.9. 80.5
Washington W7. 6.8. 80.2
St. Louis W7. 7.9. 87.3
Seattle W4. 5.9. 75.1



Houston W9. 6.6. 75.4
Buffalo W2. 5.2. 62.4
Tampa Bay W5. 7.6. 83.2
Washington W11. 7.0. 86.6
St. Louis W9. 7.4. 85.1
Seattle W6. 5.5. 70.0



Houston W8. 6.7. 84.1
Buffalo W5. 6.0. 66.9
Tampa Bay W7. 6.5. 79.6
Washington W6. 7.2. 74.3
St. Louis W7. 7.5. 80.4
Seattle W10. 6.1. 85.0


Some of these teams were not all that high in cohesion. I am going to make sure to knock them right down by taking receivers with the QB as necessary. New cohesion rating listed in the top paragraph.
In addition I decided to do something else with Houston and Tampa. I took out 12 or so starters, and traded them back a week later to give us a huge cohesion drop across the board. These 2 teams had massive cohesion drops in each of the other 3 categories, in addition to the passing one. If this doesn’t produce a drop in wins due to cohesion, I don’t think anything will.


Test group raw data
Houston W3. 6.6. 72.3
Buffalo W6. 5.0. 56.6
Tampa Bay W5. 6.4. 78.2
Washington W7. 6.9. 81.6
St. Louis W7. 6.7 73.0
Seattle W5. 5.9. 78.2


Houston W10. 6.9. 84.4
Buffalo W5. 5.0. 56.2
Tampa Bay W4. 6.4. 80.9
Washington W8. 7.3. 79.8
St. Louis W4. 7.0. 81.9
Seattle W6. 6.4. 75.7


Houston W4. 6.6. 74.1
Buffalo W5. 6.0. 68.3
Tampa Bay W3. 6.1. 60.4
Washington W9. 7.1. 84.5
St. Louis W5. 6.7. 71.8
Seattle W7. 5.8. 72.3


Houston W9. 6.2. 76.5
Buffalo W6. 5.6. 72.2
Tampa Bay W4. 6.3. 73.9
Washington W8. 6.5. 81.9
St. Louis W4. 6.3. 76.7
Seattle W10. 6.3. 77.0


Houston W6. 6.8. 79.4
Buffalo W5. 5.7. 67.7
Tampa Bay W7. 6.8. 79.4
Washington W7. 6.4. 74.7.
St. Louis W6. 6.3. 71.5
Seattle W5. 6.1. 72.4



Lets start with the 2 teams who suffered big cohesion drops in all 4 areas of the team. Certainly hoping to see something here.

Houston - High cohesion W 8.8. 6.7. 80.8
Houston - Low cohesion W 6.4. 6.6. 77.3

Big difference in wins, it looks quite significant. Hopefully it will be the same with Tampa.


Tampa - High cohesion W4.8. 6.7. 79.3
Tampa - Low cohesion W4.6. 6.4. 74.6

Never straight forward is it. While the passing games declines wins, hardly at all.


Buffalo - High Cohesion W5.4. 5.7. 70.5
Buffalo - Low Cohesion W5.4. 5.5. 64.2


Washington - High Cohesion W7.8. 7.0. 80.4
Washington - Low Cohesion W7.8. 6.8. 80.5

St.Louis - High conhesion W7.4. 7.4. 79.8
St.Louis - Low cohesion W5.2. 6.6. 75.0

Big difference again here.

Seattle - High cohesion W6.2. 5.8. 74.3
Seattle - Low Cohesion W6.6. 6.1. 75.1



Conclusion

We are dealing with a small sample size here. With this in mind I am trying to look at the stats as a whole. For example the 2 teams I totally broke up had 2.4 and 0.2 wins more in the high cohesion group, so I will add them together and divide them by 2, giving me 1.3 wins.

We saw little movement overall in the other 4 teams. 0.45 of a win between the 2 groups overall. There was also a small tendency for the high cohesions teams passing game to be better overall.

My personal opinion and final conclusion is that cohesion IS at work here. The variable does seem to raise its head with weaker teams, but I have seen no evidence whatsoever that it does with elite teams.

Having said that the impact cohesion has seem to be little. The difference in cohesion between the 2 groups tested here is large. The results are not. I would suggest cohesion is a relatively minor factor to bear in mind, it shouldn’t stop you doing what you want with your team.

[Koprnkc]

Would the Cohesive hit be bigger with the trade of the QBs best WR? IE, instead of trading Manning / Brady. Trade Wayne / Moss.

The theroy being the QB has not worked with the new WR, and thus has not "developed" the timing, etc...

[Hammer]

I don't think so.

[Hammer]

Well I got some time today afterall, as the other half went to see the in-laws.

I ran two sets of 50 seasons and recorded the wins for 6 teams. I decided to forget all other stats for time reasons. I figured win-loss is the ultimate test. With this good sized sample, I would expect to see some evidence of cohesion at work. As we mainly saw evidence of movement in wins, rather than passing stats in previous studies, I felt this added further validation in using the number of wins as my measuring tool.

For 4 teams I reduced the cohesion hugely by trading away their star veterans, and then trading them back the following week. The other teams I used as controls, I never touched them. I just wanted to do this to see how consistent my results were, and that the sample size was enough. The 2 control teams were very consistent in terms of wins in both groups.


Below I recorded the results for both initial cohesion, and after I had traded, and then traded back the players to their original teams, in each of the 4 cohesion catogories.


High cohesion

Indianapolis 100(1st) 95 (3rd) 88 (3rd) 87 (4th)

St.Louis 68 (21st) 68 (31st) 67 (21st) 65 (25th)

Tennessee 66 (25th) 81 (14th) 75 (12th) 67 (21st)

Carolina 78 (13th) 85 (12th) 77 (11th) 70 (16th)



Low cohesion

Indianapolis 60 (32nd) 67 (29th) 66 (23rd) 66 (23rd)

St.Louis 63 (31st) 61 (32nd) 58 (31st) 56 (30th)

Tennennee 63 (30th) 64 (30th) 62 (26th) 65 (25th)

Carolina 68 (25th) 72 (24th) 61 (29th) 64 (27th)


As you can see I really whacked every team hard in the cohesion department.

　

Number of wins with High cohesion

Indianapolis 14.1

St.Louis 5.2

Tennessee 5.1

Carolina 8.4

San Diego 12.8 (Control)

Miami 5.9 (Control)

32.8 total for the 4 teams tested.





Number of wins with Low cohesion

Indy 13.6

St 4.6

Ten 5.6

Car 8.6

San 12.7 (Control)

Mia 5.9 (Control)

32.4 total for the 4 low cohesion teams.


The difference in wins I saw in the previous studies pretty much disappeared. I have emailed this study to Jim. It might be that when the teams get their players back cohesion within the game returns to the original higher level. However, the in-game ratings certainly do not, so at this point I presume that is the case.

No evidence of cohesion having any real impact in this study.

[Sef0r]

It's pretty clear that cohesion is calculated based on the value in the "Joined Team" field. If someone is cut/traded/etc. and then comes back, that number resets to the latest time he joined the team.

If I forget to resign a player and they hit FA but can successfully sign them again in FA will there be a cohesion hit?

[TheMeat]

No, the Joined Team field stays as the original year they came to your team. If you let them sit long enough to hit preseason, they are officially off your team.