So, What Good Is Team Chemistry Anyway?

[QuikSand]

So, What Good Is Team Chemistry Anyway?

For the last couple of versions of Front Office Football, there has been a “team chemistry” function that assigns affinities and conflicts among the players.

In the basic sense, each position group has a leader, and each player within that group (plus each QB for the offensive player groups) are tested for compatibility with that group leader. The results of these compatibility tests are the affinities and conflicts that appear on the “Attitude Advisory” tab of your “View Rosters” screen.

For some background on this system, here are some links that may be useful:

Initial Discussion thread on Team Chemistry – Warning, this may be a “spoiler” thread for those who don’t want to understand how this stuff works, as it explains with some detail how team chemistry is generated.

Anything new with affinities/conflicts? – some continued discussion on what it takes to generate an affinity or conflict

There are probably more discussions on this subject that shed some light, their omission here is unintended, I just did a quick search to find the threads that I remembered on the matter.

- - -

With that, the remaining question seems to be: Even if we understand how to build affinities and avoid conflicts…How big a deal is this stuff anyway?

I recently had a single player career that afforded a great opportunity to put this concept to the test. I had played a team that, for various reasons (though not the initial or primary idea behind the career), ended up becoming an extreme case of high-affinity. In fact, at the point of this testing, the team roster included an affinity with every possible player on the team save one (whose personality score was apparently just a shade too low to register). That seemed like an interesting platform for some experimentation.


THE EXPERIMENT

A single player FOF 2004 team has been assembled, with very high affinity levels. As indicated above, the nominal team has position leaders in every group who maintain an affinity with every player on the roster (only one exception, at DE). Each quarterback, as well, has an affinity with at least one offensive position leader.

I saved this high-affinity team, which was also a very cohesive unit with many players who had played together for a substantial time period. To determine a basis for expected results, I ran a full season with that team in a league with injuries set to zero. From fifteen such full season trials, I excerpted some basic outcomes from the team, focusing on its offense in particular. For each season, I gathered the following data:

Regular Season Wins
Rushing Yards
Yards per Carry
Passing Yards
Yards per Attempt
Points per Game


For the experiment, I made modest changes to the team’s roster, only on offense. For each position group, I released or traded the incumbent position leader. In each of the three cases, this player was neither a starter, nor did he play a significant role in the depth chart. In each case, I acquired a new player to become the position leader – and in each case that new position leader had a substantially poor chemistry with the neatly-assembled position group. I do not have details readily accessible, but there were at least one or two conflicts within each group. In short, the affinity summary went from an absolute extreme positive (every single offensive player having an affinity) to a fairly extreme negative (where there were no affinities at all, and several conflicts).

The entire defense was unchanged – the strong positive affinity structure was left in place.

So, the change of these offensive leadership roles, nominally, should have the following effects on the team:

-no meaningful effect on the ratings of on-field personnel
-some (presumably modest) loss of team cohesion resulting in loss of veteran player
-substantial loss of affinity from position leadership change

This, to me, yielded the best possible contrast in player productivity – presumably with enough trials, the difference in outcome between these two teams should be almost exclusively attributable to the affinity differences. (The loss here of three non-starting veterans, replaced with new veterans, simply can’t amount to a huge swing in cohesion in my judgment)


So, I made these adjustments to the FOF team, and recorded the same outcomes from each of 15 seasons using the modified team. Again, no injuries and no gameplan adjustments – so the same players would be present in roughly the same roles on both sides of the experiment – just to reduce the number of variables as best we can.

I am fairly confident that any differences observed between the all-affinity team and the non-affinity team in this controlled environment is largely, or nearly entirely, attributable to the difference in team chemistry. This chemistry difference is about as pronounced as can be orchestrated, so if there’s anything at all to be found, it ought to show up here.


THE RESULTS

		
BROKEN OFFENSE AFFINITY	AVG	StDev
Wins                   	10.5	1.76
Rushing Yards	        1816	124
Yards per Carry	        3.78	0.17
Passing Yards	        3762	198
Yards per Attempt	7.50	0.51
Points per Game	        20.5	1.80

WITH FULL AFFINITY	AVG	StDev
Wins                  	13.2	1.21
Rushing Yards	        2204	180
Yards per Carry      	4.15	0.24
Passing Yards	        3776	246
Yards per Attempt	7.93	0.54
Points per Game	        25.6	1.53

Clinical Observations:

Wow.


Continued Observations:

With the exact same personnel on the field, and largely everything held constant, the high affinity offense team was substantially better than the poor affinity offense team in pretty much every meaningful way. They ran better, they passed better, they scored more points, and they won more games.

Keep in mind that even without the benefits of affinity, this is a pretty good team. The “broken affinity” experimental version still was winning 10 or 11 games most of the time, frequently a top seed in the playoffs. A swing from 10-11 wins to 13 wins might better be viewed as a reduction in losses. The broken affinity team only lost 5 or 6 games a year – but the full affinity team reduced those losses from 5-6 to about 3 per year, essentially eliminating about half their expected losses.


A few more tidbits from the data:

- The broken affinity team managed to win 13 games twice in 15 trials. Again, this was a good team already, with solid players and very high cohesion. With the broken affinity, though, the team did miss the playoffs in several trials – with three seasons of eight wins, and two more with only nine wins. The full affinity team, however, won at least 12 games in every single trial, had two seasons of 15 wins and one unbeaten season.

- The broken affinity offense managed to top 2,000 rushing yards once in 15 trials. The full affinity offense failed to rush for 2,000 yards only once in 15 trials. This is presumably a function both of rushing effectiveness and overall team success, but the difference here was marked.

- Passing efficiency, which seems like it might have a higher variance, still showed a marked split. Only three of the broken affinity trials yielded a yards per attempt that matched the average of the full affinity team.

- - -

CONCLUSIONS

It is still far too early to do any pinpointing. I cannot isolate the value of a single affinity, or the impediment posed by a single conflict. The system we’re dealing with is probably too complex to ever be able to isolate such detail – we’ll probably never know.

But, if your general concern is “does this stuff matter at all?” the answer has to be a resounding yes.

A team with chemistry working against it (on one side of the ball, anyway) honestly didn’t have a great deal of room to improve – this was not a team where a small boost could turn a few tough losses into wins, and make an easy swing from 6-10 to 10-6. But reversing its chemistry problems and building strong affinities – and essentially no other change at all -- made a marked difference in the team. That team became more effective at all its offensive endeavors, and far more consistent in winning games – even with all the same players on the field.

This is, admittedly, an extreme example. The swing in chemistry from something like 20 offensive affinities to six or eight conflicts is pronounced and artificial. It’s on a scale larger than anything you’re likely to consider in any FOF team you manage. That’s deliberate, of course.

But the magnitude of the effect is pretty great, in my judgment. If you are playing FOF, either in single player or multi player, and you haven’t been paying attention to team chemistry – you are missing out on a potentially meaningful element to this game. It takes work to build a team this way, but it’s possible that this is a way to better your team that doesn’t necessarily require a huge investment in top-skilled players (the sort who require high draft picks and/or big money contracts).

Feel free to use this information as you see fit.

[albionmoonlight]

Thanks a lot for sharing. FWIW, I noted in IHOF that Jim's team was conflict free and decided to make mine conflict free as well. People attribute the success of Fairbanks (which, on paper, is a good but not great team) to gameplanning. My ego likes to think that gameplanning is part of it, but I beleive that some of our success also comes from simply paying attention to conflicts/affinities.

FWIW, I also recall Jim mentioning in one context or another that he signed a guy in order to "gain another affinity" or some such thing. The player in question was a backup, which indicates to me that having affinities is important, regardless of whether the players are starting or not.

[Ben E Lou]

I've never done a study. I didn't have to: http://www.younglifenorthdekalb.com/fofc/annarbor.pdf

[Buzzbee]

I also know that North Plainfield makes an active effort to be conflict free, FWIW.

Thanks for the analysis Quik. Many have speculated that chemistry had an effect, but you've put an empirical spin on that.

Any thoughts on how chemistry might work under the hood? Do you think an affinity might be a positive modifier to a player's ratings? Could it possibly be a modifier on the results of a play? My guess would be that the easiest way to implement it would be as a modifier to ratings. Of course if you have a position leader that has multiple affinities, would he get multiple 'modifications' to his ratings? Could a 50 rated center play like a 60 rated center, simply because of multiple affinities?

Again, thanks for taking the time and for sharing your results.

[QuikSand]

Originally Posted by Buzzbee

Any thoughts on how chemistry might work under the hood? Do you think an affinity might be a positive modifier to a player's ratings? Could it possibly be a modifier on the results of a play? My guess would be that the easiest way to implement it would be as a modifier to ratings. Of course if you have a position leader that has multiple affinities, would he get multiple 'modifications' to his ratings? Could a 50 rated center play like a 60 rated center, simply because of multiple affinities?

I have no idea, and don't see how I/we would ever be able to measure this.

All I have done with this, I think, is convinve myself beyond any doubt that this stuff does indeed matter. Whether it's a separate calculation done in the game, or some kind of ratings magnifier, or something else -- I don't have any idea.

[timmynausea]

I had been wondering about this a bit lately. Good work, Quik.

[Taco]

I think the main reason for my early season success (8-1) in the GEFL is due to my good team chemistry. My starting quarterback has affinities with all 3 offensive position leaders, and I have 10 affinities and 11 mild affinities. There is only one mild conflict and he's my 3rd string MLB...hmmmmm....someone might get cut soon...

[QuikSand]

To me, it is a great mystery how much these affinities matter, and also how much conflicts matter. At one point, I started calculating it as a score... but even just separating the affinities (mild, affinity, strong) how much weight do you assign the different levels? 1-2-3? 2-3-4? And then, how much should a conflict count against you in that kind of metric? Is a conflict enough to offset one affinity? Two? Ten?

A total unknown to me, and I really don't see how you could ever reliably measure it.

On teams where I am taking this seriously, I have basically gone with a philosophy of: no conflicts, period, and affinities are more than just a tiebreaker when making my roster decisions.

[Celeval]

I think it's a reasonably accepted belief that the mild/affinity/strong values are mapped back to a particular set of values (trying to be non-spoilerish in this one) - I think it may be reasonable to assume that those labels are only that - labels - and the effects are related to the 'pure' affinity score behind the labels.

[QuikSand]

Originally Posted by Celeval

I think it's a reasonably accepted belief that the mild/affinity/strong values are mapped back to a particular set of values (trying to be non-spoilerish in this one) - I think it may be reasonable to assume that those labels are only that - labels - and the effects are related to the 'pure' affinity score behind the labels.

Yes, I agree that's almost certainly how it works. But how these values are calculated is a mystery to me. It's obviously a function, at least in part, of the players' personality scores -- but if they sum to 100 does that mean that the affinity counts twice as much as another case where they sum to 50? Or is this (liek many things in FOF seem to be) a nonlinear sort of value? Does one player count more than the other? Do conflicts work the same way? What about the labgguage in the traning camp page that references conflicts? How do "team functions" affect all this?

Lots of levels...

[cthomer5000]

Originally Posted by Buzzbee

I also know that North Plainfield makes an active effort to be conflict free, FWIW.

That's true, although I never go out of way to find an affinity. I'm talent first, so i'll take a "neutral" guy over an affinity guy if i like his skills better. Adding conflict guys are simply out of the question for me.

[Buzzbee]

Definitely many layers, and I agree that it would be difficult, if not impossible to try to determine the actual effects of individual conflicts or affinities. Perhaps the affinity/conflict might be compared to a +/- scale and act as a modifier. Perhaps no conflict or affinity would rate a 0 modifier, an affinity would rate a + modifier and a conflict would rate a - modifier. Since the range of the raw numbers can be from 0 to 200, it seems this could convert pretty well to a %.

Another thought is that since Jim likes to use the 355-1000 range (it is 355 or somewhere in the 300's) this might help explain players whose personality is too low to register. The 0-200 range could equate to 0-1000. Personality combinations below the 3xx level simply don't register.


I have no clue if the modifier would modify individual ratings such as breakaway speed, either singularly or en masse, or if it might modify a player's overall rating. My guess is it would modifiy all of a player's individual ratings. To me this seems to be the most logical way to implement this. As I see it + chemistry leads to better results. Those better results have to come from a modification to the player ratings, or the results of individual plays. It just seems easier to modify ratings rather than plays.

Also, traning camp might serve to lessen the negative modifier. Instead of having a value of -.0042 training camp might adjust that down to -.0026. Purely speculatve numbers, used for illustration only.

Those are just some thoughts. Take them or leave them as you see fit.

[Dutch]

Originally Posted by Buzzbee

Definitely many layers, and I agree that it would be difficult, if not impossible to try to determine the actual effects of individual conflicts or affinities. Perhaps the affinity/conflict might be compared to a +/- scale and act as a modifier. Perhaps no conflict or affinity would rate a 0 modifier, an affinity would rate a + modifier and a conflict would rate a - modifier. Since the range of the raw numbers can be from 0 to 200, it seems this could convert pretty well to a %.

Another thought is that since Jim likes to use the 355-1000 range (it is 355 or somewhere in the 300's) this might help explain players whose personality is too low to register. The 0-200 range could equate to 0-1000. Personality combinations below the 3xx level simply don't register.


I have no clue if the modifier would modify individual ratings such as breakaway speed, either singularly or en masse, or if it might modify a player's overall rating. My guess is it would modifiy all of a player's individual ratings. To me this seems to be the most logical way to implement this. As I see it + chemistry leads to better results. Those better results have to come from a modification to the player ratings, or the results of individual plays. It just seems easier to modify ratings rather than plays.

Also, traning camp might serve to lessen the negative modifier. Instead of having a value of -.0042 training camp might adjust that down to -.0026. Purely speculatve numbers, used for illustration only.

Those are just some thoughts. Take them or leave them as you see fit.

Having a hidden value on the player seems logical. A plus or minus modifier for certain attributes.

Maybe it's as simple as a {-3, -2, -1, E, +1, +2, +3} modifier on top of one, some, or all player attributes.

[MalcPow]

I think there is already a kind of "universal modifier" applied to a team's performance. Aren't coaches' motivation ratings some kind of positive or negative modifier on overall performance? Affinity/conflict probably affects the game in the same way, and might even affect the same general modifier. I don't think it is necessarily a ratings thing, but something that affects the "rolls" in game. Obviously I don't really know, just spitballing ideas, but I always kind of intuitively thought cohesion worked in a similar way, not necessarily affecting the players themselves, but positively impacting the randomness being generated behind the scenes.

[Dutch]

I would imagine putting bonus/penalty modifiers on the "player card" would be easier than to add calculations to the game sim during the weekly sim. But I admit I would just be guessing.

[mhass]

Very nice work, QuikSand. Interesting that, in general, the affinity team had a wider range of values. So affinities definitely are better, but, if you take your control group as a "universal" control, more unpredictable or, to confuse terminology, volatile.

You mentioned the win totals, but in a theoretical head-to-head, Team Affinity would (statistically) beat Team Broken a significant majority of the time based on the points/game range (24 minimum for Affinity and 23 maximum for Broken).

[QuikSand]

Originally Posted by mhass

Interesting that, in general, the affinity team had a wider range of values. So affinities definitely are better, but, if you take your control group as a "universal" control, more unpredictable or, to confuse terminology, volatile.

I decided to measure and share my values for standard deviation, but I honestly don't know what, if anything, to take from those values. The variability in the total number of yards passing or rushing, to me, is not necessarily indicative of much - since the team's decision to attempt passes or rushes is a function of many things that themselves are likely to vary a lot. I don't know if a 15-season trial is enough to draw any conclusions about reliability in one verus the other.

But the yards per carry and yards per attaempt, I'd think, ought to be more normalized -- so maybe the greater variance in the high-affinity team's results there is meaningful. I don't even have a theory to explain this... but I guess it is a subsidiary finding here, or at least a temporary/tentative one.

[Cotton]

Linemen seem to be the grumpiest bastards. Is that just my teams?

[Warhammer]

I think it is because there are more linemen to keep happy. That is where I have the most injuries, so I am constantly tweaking the lines. Also, during the off-season, I have the most turnover at my backup OL positions.