Getting a better handle on AI contract evaluation

[QuikSand]

DECEMBER 2013 EDIT - THIS IS AN OLD THREAD, REVIVED SINCE THE TOPIC SEEMS RELEVANT FOR FOF 7



One element of sports text sims is that the game itself, even in a player-v-player environment, has to make some fairly complicated decisions itself.

I thought it might be timely for this community to discuss, a bit, how we might help define a better system for contract offer evaluation.

How does the game determine which, among several contract offers, the players should prefer/select?


I'll help frame the discussion to suit my interests:

-Let's use professional American football as the test case, with with many different positions of varying overall importance

-Let's use some combination of guaranteed money (like a bonus) and non-guaranteed (like annual salaries)

(this space reserved for future guidance)



As an obvious note, anything you might post here should be considered to be an free and open use contribution. If we somehow buck the odds and do good work here and advance the thinking on this front, I'd want it to pay off as a dividend to the user community as a better game. Nobody here owns anything said here.

[QuikSand]

Quick thoughts:

-I think a pre-defined value for "peak value age" (probably a variable based on position) may be important here

-I think a key concept is whether the player considers the contract to be a good one or a bad one, relative to his expectations (in particular, good contract = want longer duration)

[MIJB#19]

Do you want any mentioning that players may be willing to sign for less in specific situations? Like:
* desire to play with a stronger team (better roster, better results last season)
* likeliness to get playing time on the target team
* preferring specific teams based on other personnel already there, or it being their favorite team (closest to their hometown)

[Julio Riddols]

I think it would be simple to just limit the A.I. to spend a certain amount on each position and have a specific level of talent at any given position that causes it to stop trying to fill that position. It could be different for different teams, maybe randomized at the beginning of the game.. But the CPU should know when it has a good player at a major position, such as QB.

Just look at what players in the NFL are getting paid, find the high end and low end of that, then have the A.I. reproduce that. A player of a certain caliber should be paid a premium, which is then modified higher or lower by his production or lack thereof. Production should be measured on an average per play and weighted so that players with low sample sizes will still be seen as guys who have essentially not played. The more a player has played, the more proven he becomes, which is reflected by him seeking a contract commensurate with his skills and production level.

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

Things that should raise a players value:



Performing in the top 10 (top 20 for positions with more than one starter such as WR/G/T/DE, etc) at your position on a weighted per play basis in a given season. For each season this occurs, a player could be given a slightly larger boost than the last time they earned the boost. It should be noted on a players card when they scored in the top 10 at their position in a season. For return specialists and special teams stars, this could be noted additionally. A guy with long snapping skill or return skills or kick coverage skills should be valued a little more than a guy with the same level of talent who cannot do those things.

Youth: depending on which side of his prime a player is on, his value can be higher or lower. Being on the youthful side of things should give a boost, being in their prime should give an even bigger boost. This would certainly vary by position.

Durability: Lack of injuries should be involved. If a player has not been hurt or hasn't been hurt much, this can be added to their contract demand weighted again by the number of downs they have played.

Leadership: Players who are good leaders or mentors should get a small boost to their contract demands over equally skilled guys without those extra skills.

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

Things that should leave a players value unchanged

Average performance. If a player finishes between 11-22 at his position, (21-44 at positions with 2 starters)

Average injury rate

No positive or negative added skills.

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

Things that should decrease a players value:

Poor performance (players finishing below the top 22/44 at their position out of the top 32/64 qualifiers) This would also be weighted so that players who didn't play enough to really be evaluated wouldn't see their contract value compromised because of it.

High injury rate. Players that are constantly injured should see a pretty significant dent in their value. Players coming off a season ending injury should get an even bigger dent.

Red flag/bad traits.

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

After that, you just have to figure out how to get the A.I. to manage their money properly while still trying to obtain/release the right guys. They should know to dump a guy with bad traits and/or poor performance before they let go of a guy with talent and potential. They should know when they have a guy with good talent and potential at a position and consider that position "filled". In those cases, the A.I. shouldn't sign a guy within 25% in OVR or draft a guy in the first 2 or 3 rounds of a draft at that position.

They should know when they have an average/adequate guy, and react accordingly to that as well. The position could be considered "competitive" and the team could be inclined to sign a guy that is significantly better if available, or draft a guy in the 2nd-4th round to compete at that spot, or bring in a guy with similar talent in a platoon situation. This would be weighted by position, with the most important positions receiving precedence over the others. A team with an average P and an average QB should be more inclined to address the QB situation than the P situation, but should see neither as a major need.

They should know when they are in bad shape at a position too. The position can be considered "needed" and would be one of the first positions in line to be filled by a first round draft pick or an upgrade in FA.

Now lets say you weigh that all by position.

A team with "filled" positions should only look to add to those positions when they have no "needed" positions to fill. This would allow teams with 2 good WR to still draft/sign a 3rd if they have the luxury and the money to do so.

Other than the filled positions, the competitive and needed positions should be weighted as well.

A team with a competitive DE situation but a needed P should still try and find a DE before they worry about a P.

For example, an idea for how the positions could be weighed:

Needed QB (10)
Needed CB (9)
Needed WR (8.8)
Needed DL (8.5)
Needed OL (8)
Needed LB (7.5)
Needed S (7.2)
Needed RB (7)
Needed TE (6.5)
Competitive QB (5.5)
Competitive CB (5.2)
Competitive WR (5)
Competitive DL (4.5)
Competitive OL (4.2)
Competitive LB (4)
Competitive S (3.8)
Competitive RB (3.5)
Competitive TE (3.2)
Needed FB (2.5)
Needed K (2)
Needed P (1.5)
Competitive FB (1)
Competitive K (.5)
Competitive P (0)

Then after the above were taken care of, then maybe the team would look to grab luxury talent, like a 3rd good WR, DB, etc.

Through all of this, the price teams would be willing to pay for talent at a given position would be considered. A team who doesn't like to pay a whole lot for CB's won't go out trying to acquire the most expensive ones. They'll try and acquire them through the draft or they'll sign lower tier guys. Conversely, a team that favors CB's highly will go after the big FA if they are available, and will pay out the nose to keep one around.


Now for an example of a player or two. For these examples, we are assuming that 3 million is the base salary for a 50 OVR MLB.

Joe Schmo is a 50 OVR MLB, a position with decent longevity. He is a free agent with 5 years of experience, in his prime.

He was in the top 10 once as an MLB and once as a ST gunner, has average injury concerns, and has the leadership trait.

Joe Schmo would receive contract boosts from his top 10 as an LB (lets say that adds 5% to value), his top 10 ST (adds 1%), his leadership (adds 3%) and being in his prime (adds another 10%) so lets say he is worth 19% more than 3 million. His contract demand would then be 3,570,000.
-----------------------
Bob Flake is another 50 OVR MLB with 5 years of experience.

He has finished in the bottom 10 3 times in his career as an MLB, and is coming off a season ending injury suffered in week 4 of last season.

He also has a couple red flags.

Bob Flake would see his contract demands helped initially by him being in his prime (+10%) but loses 5%+6%+7% (-18%) for finishing in the bottom of his position 3 times in his career. He loses an additional 20% for his season ending injury (any injury with projected recovery time of more than 16 weeks) last season, and 5% for each red flag. Now he finds himself at -38% of his normal value. He only asks for 1,860,000 a year. Chances are a team that goes cheap at that position would sign him, hoping for his talent to finally shine through. Chances are it would also be his last contract if he again fails to perform or finds himself severely injured again.

[Ben E Lou]

Julio, I think Quik meant how the game's AI evaluates the value of different offers, not how the SP AI evaluates how to make an offer.

[gstelmack]

One thing I ran some experiments on was calculating the Present Value (is that the right term?) of a contract. Bonus + first year + regress based on some inflationary value future years. The trick was to make the regression high enough that future years count some but less. I wanted to come up with something where a backloaded contract was worth less, simply because you might never reach it.

I believe I also came up with a routine to figure out the "actual" terms of a contract, based on year-to-year jump. So, a contract like:

B 20mil
1 5 mil
2 6 mil
3 7 mil
4 8 mil

is a 4 year / $46 mil contract, while:

B 20 mil
1 5 mil
2 5 mil
3 16 mil

is a 2 year / 30 mil contract.

I'd have to go dig up my old notes / routines, I had a bunch of test contracts I could run through it to reach an expected value for the contract. You could then adjust the present value for other factors like hometown, play for winner, current employer discount, etc, then just pick the one with the highest present value.

[QuikSand]

Julio, I think Quik meant how the game's AI evaluates the value of different offers, not how the SP AI evaluates how to make an offer.

Yes, sorry, that's what interests me here.

[QuikSand]

And in general, I think it's wisest to set aside the various add-ons like "play for winner" or "loyalty" or whatnot, and just start with raw numbers. We understand that this foundation could be built upon with other ornaments (and I hope that in any given sim that it would) but I feel the foundation itself is in need of attention.

[QuikSand]

Okay, here's one more concrete example of where I think we need some help.

Two competing contract offers:
S: 1yr deal, $2m salary and $3m bonus
L: 5yr deal, in each year $2m salary and $3m bonus

Right now, one simple way to measure such things is on a purely per-year basis. I have deliberately made these two offers completely identical in that regard (possibly excepting inflation, which I'll shelve for now in the interests of simplicity).

But which is the better offer? L has $15m in guaranteed money, and that means something - the player is completely protected against an immediate injury or a decline in performance to that level. That's good.

But what if this is an emerging star player - one who might only command a $5m contract now, but in a couple more years of development might be worth a good deal more than that? If the sense is that strong that this is an under-market contract, then maybe he'd prefer the short-term deal... right?


This is the sort of pivot that I think makes this interesting and not yet fully explored. It would be nice to try to pin some of this down into a formula.

[QuikSand]

One thing this would do is put a lot of importance on the game's ability to determine "expected contract" for a given player in free agency. That might be difficult - and it gets into some of the nuances Julio started on above.

But I'd like to either (a) make that calculation part of this dicsussion, or alternatively (b) just take that as a given, and work from that as a starting point.

[flounder]

You could develop some sort of concept of the typical career arc for a player in terms of how they will perform relative to other players of that position. Then you could consider the typical salary for a player of a given value above a replacement player. That gives you a measure of how valuable a player of a certain ability will be to a team over the course of a few years.

Then a rookie WR could say something like, "I'll be the 60th best WR years 1 and 2 and that's worth X/yr to a team, but in years 3 and 4 I'll be the 10th best WR and that's worth 5*X/yr. So any 4 year contract that pays me more than 12X is a good contract."

Conversely, a veteran player could say, "I will still be the 5th best MLB for the next 2 years and that's worth 4*X/yr, but I'll be the 10th best the following year and that's worth 2*X/yr and I'll be the 30th best the year after that and that's worth X/yr. So any 4 year contract worth more than 11X is a good contract."

Then you could start adding in complications like differences in how a player perceives their ability compared to different teams, how risk averse or tolerant a player and team are, if a team wants to mortgage the future to win now etc.

Dunno. Just brainstorming here.

[TRO]

Then a rookie WR could say something like, "I'll be the 60th best WR years 1 and 2 and that's worth X/yr to a team, but in years 3 and 4 I'll be the 10th best WR and that's worth 5*X/yr. So any 4 year contract that pays me more than 12X is a good contract."

Conversely, a veteran player could say, "I will still be the 5th best MLB for the next 2 years and that's worth 4*X/yr, but I'll be the 10th best the following year and that's worth 2*X/yr and I'll be the 30th best the year after that and that's worth X/yr. So any 4 year contract worth more than 11X is a good contract."



You would need to be careful in how the player is "determining" what their worth will be in future years. It could turn into a tell to the human player if they see a masked player looking for a big contract.

[OldGiants]

Actually, in "American Professional Football" (do you mean the NFL, or another league of which I am not aware?), your examples of S and L both have the same guaranteed value of $5 million, assuming the first year's salary is paid.

NFL contracts are not guaranteed unless specified in the contract. Players who are released do not ordinarily receive the remaining money on their contract.

[Julio Riddols]

My bad on the misinterpretation. This is why "recklessly enthused" is part of my sig.

As for a player determining the value of a deal, I think they ought to go with the most guaranteed money, then the best situation given their individual preferences. Guaranteed money should be like 95% of what a player looks for, and loyalty/situation/winning/etc should make the player see an offer from that team as more valuable than it actually is, but not enough that it would trump an offer several percentage points higher in guaranteed money.

Ultimately, all that should really matter to a player in free agency is guaranteed cash. The rest is valueless if something happens that results in your release.

Use the average pay scale for players of his rating and then use their history to add or subtract from the expected value for players who are renegotiating prior to entering free agency.

So maybe a WR who you drafted in the second round 2 years ago is entering the final year of his deal and you want to look at re-signing him now before he plays his 3rd season and potentially adds to his value.

The player can use his development, youth, production and health as ways to boost his own expected value. I'd think he would ask for a contract commensurate with that and maybe a little more. He should probably accept a reneg deal if it pays him within a few % what his development, youth, production and health says it should.

As for bonus amount, the player renegotiating a deal with his current team before going FA should expect the percentage of bonus to be at or near (within 5% or so?) to what the average player around his skill level gets in free agency and renegotiations historically.

Elite players should always expect a huge amount of bonus money, maybe something between 40 and 50% of the contract?

I believe the current CBA prevents drafted players from renegotiating their deals until they are in the 3rd year of their contract too, so that will help make player contract valuation a little easier I think for renegotiation purposes.

[Dutch]

My bad on the misinterpretation. This is why "recklessly enthused" is part of my sig.

To be fair, it was a pretty good post.

[Julio Riddols]

I'm one of those people whose head is exploding with ideas and dreams but lacks the motivation somehow to see it through, which results in a lot of vicarious living.. So I tend to get really passionate about things I really want to see happen. If I had the patience to program I'd be spending every waking hour crunching numbers and tweaking code and working on the amalgamation of all the good ideas I have ever seen in a football game.

[QuikSand]

Actually, in "American Professional Football" (do you mean the NFL, or another league of which I am not aware?), your examples of S and L both have the same guaranteed value of $5 million, assuming the first year's salary is paid.

NFL contracts are not guaranteed unless specified in the contract. Players who are released do not ordinarily receive the remaining money on their contract.

Yes, the NFL, I'm being obtuse on purpose, I'd hope you might imagine why.

Of course the first year's salary is not guaranteed, per se. And I'm not including it as such in my calculations.

The $15m guaranteed I reference is the bonus = 5y x $3m/yr, all upfront. Again, trying to keep this simple.

Now, as we try to noodle this, I'll offer one more wrinkle... Signing that L contract does virtually guarantee that first year salary, wouldn't you agree? A team is very unlikely to offer that contract for $15m in bonus and void it before the first season....wouldn't we agree there?

[QuikSand]

Flounder, you're putting flesh on the same bones I am. The "career arc" concept is what I have in mind also, and at some point adorning it with things like risk inclination.

[Ben E Lou]

Guaranteed money is king. For salary amounts, you need to write a table with the cap hits of players at the beginning of each season. Free agents compare salary amounts in contracts to cap hits of players in their range from last year. The further the salary is above the average for similar players, the more likely they aren't going to get it, so the less they value that contract. Then there must be a contract length modifier for where they are on the career arc. I'll put some numbers to this if/when I have time later.

[Ben E Lou]

The CCFL just finished training camp, so this is a fairly good place to look at what I mean. I've got a set of formulas that adjusts the FOF bars to estimate what the players are really worth, based on ratings. In other words, guys with higher BPR get a higher value than their FOF overall rating. That sort of thing. But the scale is roughly the same as FOF's, so I'll use that "value" in place of "rating."

Let's say the player in question is a 50-rated Tight End. Based on the tight ends in the league rated 42.5 to 57.5 (I think a range of 15 is solid here at first pass,) both a sniff test and the actual numbers make me comfortable that such a player should expect to cost around $4.8M against the cap. I think you need to compare that number against the amount of non-guaranteed money in the first year of the contract. That way a 30K,9.99M,9.99M,9.99M contract doesn't beat out a $6M,$2M,$2M,$2M deal, for example. (In FOF right now I'm fairly sure the former wins out.) The formula tells the player that with the former contract, his non-guaranteed money, even in year 1, is *double* what a player of his caliber should expect to receive, so there's a fair chance that with the former deal, he's going to get cut after TC and end up with a one-year deal for significantly less than the $6M bonus he would have pocketed had he taken the $6M bonus.

Putting aside *how* we calculate "expected contract value," I'm thinking this component of the formula would be something along the lines of...

$bonusmoney + ($expectedmoney-$salary1)*k

There's more to it than that, but I think something along these lines gets us started...

[MIJB#19]

Okay, here's one more concrete example of where I think we need some help.

Two competing contract offers:
S: 1yr deal, $2m salary and $3m bonus
L: 5yr deal, in each year $2m salary and $3m bonus

Right now, one simple way to measure such things is on a purely per-year basis. I have deliberately made these two offers completely identical in that regard (possibly excepting inflation, which I'll shelve for now in the interests of simplicity).

But which is the better offer? L has $15m in guaranteed money, and that means something - the player is completely protected against an immediate injury or a decline in performance to that level. That's good.If there are no other factors known, every single player should pick the 5yr deal. All things equal, I think the $15M should always beat the $3M bonus.

But what if this is an emerging star player - one who might only command a $5m contract now, but in a couple more years of development might be worth a good deal more than that? If the sense is that strong that this is an under-market contract, then maybe he'd prefer the short-term deal... right?This is what seems to make it complicated and what I tried to bring up, there are variables that could drastically change which set of numbers is preferable. Even so, the gap between the offers is so big, I think if there are no other options or variables, even the emerging star player should always take the 15M bonus and later on go into holdout mode if he thinks the 5M/year income is no longer fitting with his talent.

[QuikSand]

I think if there are no other options or variables, even the emerging star player should always take the 15M bonus and later on go into holdout mode if he thinks the 5M/year income is no longer fitting with his talent.

Okay... this might be a place we start from. If it's pretty hard to conceive of a situation where the short term deal is better than the long term deal, then that ought to be the starting point.

Total dollars guaranteed, not just per-year, should be a major factor.

Are there situations where this shouldn't be the case?

[QuikSand]

Okay, let's look at two three-year offers to a quality player (let's assume that minsal is a million or less here):

B: each year includes B $5M and S $2m
S: each year includes B $2m and S $12m

"Bonus is king" theory says that B is better. But if we work with the assumption that the meaningful (but not huge) bonus at least guarantees that first year of salary... then doesn't offer S really look better? Worst case absent injury, this guy gets cut after one year, and he pockets $18m for the year -- nearly as much the entire three year offer in B, and then he gets to go out and strike a new deal.

So, I don't think we can simply say that it's bonus, bonus, bonus. I myself have played a multiplayer text sports sim in which I offered quite a lot of contracts like the one labeled S above, FWIW - this isn't totally absurd conjecture.

[Ben E Lou]

I think the bottom line for FOF is that players simply don't hold out frequently enough.

1. Take the most guaranteed money (with small modifiers for play for winner and that sort of stuff)
2. If not being paid "enough"...hold outs should be frequent.

In real life, it probably just doesn't *come* to a holdout. "Joe, you're a starter now. Here's your money." But that's not going to happen in a computer game setting where we're perfectly fine with screwing every one of our players as much as possible. We need to be forced to do something. In the absence of some serious "everyone hates that owner so we won't play for him" logic in the game, holdouts must be frequent.

[Julio Riddols]

I think the reason bonus should trump any competing offer with less bonus is because that high amount of bonus money makes the player less likely to be cut, IMO. Owners are going to dump the 12/2 guy probably after one season, because an offer like that screams low risk for the GM running the team. The likelihood of that good player not receiving an offer better than 12/2 would be slim to none, IMO. If those were the only 2 offers the guy got, I still think he should choose the 2/5 deal because there is still more guaranteed money there. If he were to crash in camp, he would likely be released with only 6 million in his pocket and no future.

I look at contracts as an insurance policy for the player. If I were a player I would go with the best bonus every time. It is the only safeguard you have against a career ending injury or complete drop off in skills, and in a sport where here today, gone tomorrow is a reality for about half the league I feel like that has to be king.

[Julio Riddols]

I think the bottom line for FOF is that players simply don't hold out frequently enough.

1. Take the most guaranteed money (with small modifiers for play for winner and that sort of stuff)
2. If not being paid "enough"...hold outs should be frequent.

In real life, it probably just doesn't *come* to a holdout. "Joe, you're a starter now. Here's your money." But that's not going to happen in a computer game setting where we're perfectly fine with screwing every one of our players as much as possible. We need to be forced to do something. In the absence of some serious "everyone hates that owner so we won't play for him" logic in the game, holdouts must be frequent.

Instead of holdouts, maybe a player can push for a renegotiation without holding out? If you ignore his requests then maybe eventually he does hold out, or it hurts his production on the field, becomes a distraction, etc.

I'd love to get emails from players that ask for raises based on production rather than having surprise holdouts out of nowhere. You should be able to see a holdout coming and have a chance to change that. Maybe if a player doesn't hold out but doesn't get a new contract from you, he will simply refuse to re-sign with you when he becomes an FA unless you overpay drastically.

It seems like this would mostly come into play for players toward the end of their rookie contracts, since that would be the time when the player is generally at his peak value.

[QuikSand]

Instead of holdouts, maybe a player can push for a renegotiation without holding out? If you ignore his requests then maybe eventually he does hold out, or it hurts his production on the field, becomes a distraction, etc.

I'd love to get emails from players that ask for raises based on production rather than having surprise holdouts out of nowhere. You should be able to see a holdout coming and have a chance to change that. Maybe if a player doesn't hold out but doesn't get a new contract from you, he will simply refuse to re-sign with you when he becomes an FA unless you overpay drastically.

It seems like this would mostly come into play for players toward the end of their rookie contracts, since that would be the time when the player is generally at his peak value.

I like the idea. Anything we come up with here in clinical terms needs to be nested inside a system that creates its own fixes. Underpaid guys ought to gripe before holding out, I'd think.

[QuikSand]

If those were the only 2 offers the guy got, I still think he should choose the 2/5 deal because there is still more guaranteed money there.

Heh, I honestly built up the B/S setup to come up with a contrast so obvious that nobody could possibly disagree that the salary can/does count at some level. Guess I failed.

But surely, we could so so, right?

BB = 3yr deal, B=3.0/yr, S=2.0/yr
SS = 3yr deal, B=2.9/yr, S=22.1/yr

Even you would go for the SS deal, right?

[Dutch]

Okay, let's look at two three-year offers to a quality player (let's assume that minsal is a million or less here):

B: each year includes B $5M and S $2m
S: each year includes B $2m and S $12m

"Bonus is king" theory says that B is better. But if we work with the assumption that the meaningful (but not huge) bonus at least guarantees that first year of salary... then doesn't offer S really look better? Worst case absent injury, this guy gets cut after one year, and he pockets $18m for the year -- nearly as much the entire three year offer in B, and then he gets to go out and strike a new deal.

So, I don't think we can simply say that it's bonus, bonus, bonus. I myself have played a multiplayer text sports sim in which I offered quite a lot of contracts like the one labeled S above, FWIW - this isn't totally absurd conjecture.

While I agree that the S offer is better, but for game functionality how should we handle a more likely scenario where owners are "carpet bombing" the top ranks to land their players?

Scenario:

Free Agency Stage 2: Top QB's
Player A (50/50)
Player B (49/49)
Player C (48/48)
Player D (17/17)
Player E (16/16)

Team A makes this offer
Player A--> B: each year includes B $5M and S $2m

Team B makes *these* offers
Player A --> S: each year includes B $2m and S $12m
Player B --> S: each year includes B $2m and S $12m
Player C --> S: each year includes B $2m and S $12m

Team A commits $7M in bonus to grab Player A. Team B commits $6M in bonus to grab Player A, Player B *OR* Player C (with no intention of holding on to all three).

If Bonus is not king, it's possible that Team B wins all three contracts, ultimately decides on Player A...and leaves Team A hoping for the released scraps of Team B.

[Julio Riddols]

Heh, I honestly built up the B/S setup to come up with a contrast so obvious that nobody could possibly disagree that the salary can/does count at some level. Guess I failed.

But surely, we could so so, right?

BB = 3yr deal, B=3.0/yr, S=2.0/yr
SS = 3yr deal, B=2.9/yr, S=22.1/yr

Even you would go for the SS deal, right?

:lol:

100% SS for me. That's a sweet deal compared to the top one. 300 k less guaranteed money that you would likely make up by simply signing the SS contract.

If Jim adds roster bonuses, that will greatly vary the way contracts have to be evaluated too.

I think guarantees should be paramount, but salary could totally sway a guy if the guaranteed money is similar enough.

I think if the bonus is 5% lower per year, a salary 10% higher could be considered an equalizer.

If the bonus is 10% lower per year, a salary 25% higher..

If the bonus is 20% lower per year, a salary 50% higher.

I feel like if the bonus isn't within 20% of the highest bonus offered, the player shouldn't even consider it.

a: 5 million total bonus, 10 million total salary would be equal to
b: 4.75 million total bonus, 11 million total salary, which would be equal to
c: 4.5 million total bonus, 12.5 million total salary, which in turn would equal
d: 4 million total bonus, 15 million total salary.

on a larger scale

50 million total bonus and 100 million in salary would be equal to
40 million total bonus and 150 million in salary.

Of course it couldn't be that plain and simple. The curve would have to change based on the total amount of money being offered or the amount of years in the deal.

If deal one was

2 years
10 million total bonus
1: 3000000
2: 6000000

and deal 2 was

2 years
10 million total bonus
1: 4500000
2: 4500000

Then the player should take deal 2.

I think ultimately it boils down to initially comparing bonuses. The player then takes the deals with the bonus within 20% of the highest deal and compares those. After that, I think highest first year salary should take precedence. If any contract within 20% of the same level of bonus money has a salary that pays the player more in the first year of the deal than the difference in bonus, then maybe he takes that deal.

This way, the player addresses his need for security by eliminating non qualifying bonuses, then addresses the overall deal in a second evaluation. I feel like future salary increases should be ignored unless the salary in a future season is less than the average bonus per year.

So, if a guy got 5 offers:

1: 3 years, 15 million bonus, 3 million, 4 million, 7 million salary
2: 3 years, 12 million bonus, 4 million, 4 million, 6 million salary
3: 2 years, 8 million bonus, 5 million, 8 million salary
4: 4 years, 18 million bonus, 4 million, 4 million, 4 million, 4 million
5: 1 year, 2 million bonus, 5 million salary

He should eliminate deal 5, deal 2 and deal 3 immediately, as they are not within 20% of the bonus value of deal 4.

Then he should reset, and compare deals 1 and 4.

Deal 1's 3rd year should be excluded, essentially making it a 2 year deal offer.
Deal 4's entire value can be considered, because the chances of a team releasing a player who is making more bonus money than salary are slim.

So now you have

1: 2 years, 15 million bonus, 3 million, 4 million salary. (22 mil/2 years)
2: 4 years, 18 million bonus, 4 million, 4 million, 4 million, 4 million salary. (34 million/4 years)

The player should take deal 1. It offers him higher bonus per year, and affords him an opportunity to seek another deal much faster than deal 2.

So, the player signs a 3 year deal worth 29 million, in which the 3rd year is likely to involve a renegotiation initiated by the team. He passes on the 4 year 34 million deal because chances are it will lead to him being stuck in a contract he is outperforming if he continues to increase his value.

An aging vet might take deal 2 though, if it is most likely his last contract. He should want to cash in as much as possible in guaranteed money if he plans to retire within the time frame of the deal being offered.

[MIJB#19]

Heh, I honestly built up the B/S setup to come up with a contrast so obvious that nobody could possibly disagree that the salary can/does count at some level. Guess I failed.

But surely, we could so so, right?

BB = 3yr deal, B=3.0/yr, S=2.0/yr
SS = 3yr deal, B=2.9/yr, S=22.1/yr

Even you would go for the SS deal, right?This to me shows how the free market works in the MP environment, 'tricky' owners will offer SS to tease the player. In return, the player needs to make the final 53-men roster to make more money if he signs SS, he still does if he gets cut after week 1 (20M more in salary divided by the number of games will beat the 0.1 bonus).

I think some variables should come into play to determine the winner

Initially I'd think a large portion of players (maybe 50%) should be confident enough about themselves to make the week 1 roster and get paid for at least a part of that season 1 salary. At what point will the 0.1 extra bonus matter? Only if the player knows for sure he won't make the 53-men roster AND won't get a contract elsewhere?

At the same time, contract SS screams that the 22M is a mirage, particularly for seasons 2 and 3 and the player at hand would have to prove himself in camp 1 to be an indisputable asset to the team. I suppose in the real world, only proven veterans would be considered to get a deal like this. Looking solely at the quarterback position, I'd think of recent years' situations of Peyton Manning, Brett Favre, Donovan McNabb and Michael Vick. Or for the past off-season I'd think of players like Matt Cassel, Vince Young and Matt Hasselbeck, stretching it a bit, maybe Kevin Kolb or Derek Anderson.

Some out of the box thinking for things that might also happen:
1. the 'owner' will downright refuse to allow to offer the 22M salary, deeming the player at hand not good enough to get a star player contract
2. the player himself will feel that he's not getting a fair contract offer and downright refuses to consider the mirage offer

[MIJB#19]

So now you have

1: 2 years, 15 million bonus, 3 million, 4 million salary. (22 mil/2 years)
2: 4 years, 18 million bonus, 4 million, 4 million, 4 million, 4 million salary. (34 million/4 years)

The player should take deal 1. It offers him higher bonus per year, and affords him an opportunity to seek another deal much faster than deal 2.

So, the player signs a 3 year deal worth 29 million, in which the 3rd year is likely to involve a renegotiation initiated by the team. He passes on the 4 year 34 million deal because chances are it will lead to him being stuck in a contract he is outperforming if he continues to increase his value.

An aging vet might take deal 2 though, if it is most likely his last contract. He should want to cash in as much as possible in guaranteed money if he plans to retire within the time frame of the deal being offered.I took out some of your post, as this was basically your conclusion anyway.

I'm not sure Deal 1 should win here. Based on the comments earlier on, Deal 2 should win due to the extra 3M. The salary of years 3 and 4 are way too far in the future to be considered guaranteed money. The player at hand would need to be pretty confident of making the team those years. There are no guarantees that the player gets a new job somewhere else after the second year of Deal 1.

You could be right, but at the moment I'm leaning towards thinking that bonus per year should be a near non-factor. The bonus figure itself is the only guaranteed money, the difference in salary will have to be enormous to make up for it.

[Julio Riddols]

With deal one, the thinking is that the player will either play his way out of the contract and be released after year 2 of the deal, in which case he is lucky he got what he did, or if the player performs well under the contract, the team will be forced to pay him that 7 million in year 3 if they don't negotiate a new contract with him. He will have all the power then, and could force the team to overpay to keep him around, thereby securing another contract with another, probably larger bonus while never really being in limbo. That way, say he signs a new deal in year 3 with an additional 15 million in bonus for 2 more years. He now essentially has a 5 year deal with a 30 million bonus instead of a 4 year deal with only 18.

I would hope the player good enough to sign a deal at this salary level would have enough confidence in himself to earn the contract and force the team to pay him again. This is why the veteran would take the second deal, and the younger up and coming player would take the first deal. If the younger guy took the second deal, he could be looking at 4 long years before he gets a shot at new guaranteed money, and by then he may be beyond his prime.

[BowTieSports]

Interesting exchange of ideas here.

One variable I don't think has been viewed as enough of a variable is age/projected remaining career span.

For example, a 50/50 rated 6th year OL should feel more comfortable with his chances of playing out a 3 year contract than a similarly rated 12th year OL. The 6th year pro should put more weight on total value of the contract, while the 12th year pro should put more weight on the guaranteed money.

Modifying an example used earlier -- 3 year deal
B: each year includes B $5M and S $2M ($21M potential; $15M guaranteed, $17M cash paid out in Y1)
S: each year includes B $2M and S $10M ($36M potential; $6M guaranteed, $16M cash paid out in Y1

In my opinion:
The 6th yr guy above should value option S over option B ... he should view himself as having at least 3 more years left in his career
The 12th yr guy above should value option B over option S ... he should view the option of playing out the full term of the deal as somewhat-to-very unlikely and have more interest in the guaranteed cash

[flounder]

Maybe we should come up with a test suite. Give a bunch of possible contracts, decide which should be preferred (maybe by polling) and see if we can write an algorithm that gives the "correct" result.

[QuikSand]

Bumping this thread - there may be stuff to harvest here. I think refinements to this system in FOF would be a plus, and since the game is open to upgrading, it seems like the right time to be thinking about this sort of thing.

[QuikSand]

Jotting down some thinking here... I'm still really interested in coming up with an all-purpose formula that does a passable to good job of evaluating contracts in the ways that we would.

Key elements:

-Defining a "peak" range of exp and/or age -- players should have different expectations based on their future performance... certainty of a long term deal and total bonus means more to older players than to younger players, etc.

-Calculating a sophisticated way to weigh base salary, looking at the accompanying bonus (or perhaps the cap hit resulting from a cut) and applying a sliding scale... in future years that should drift downward, but not to zero)

-Making a player's views on contract duration partially a function of his assessment of whether the contract is a "good" one -- if it's under market (defined by his own request level) then he should prefer short, if it's over market, then maybe he prefers a longer deal... and this becomes a recursive element in the assessment/evaluation

-Combining this stuff into an overall numeric grade assigned to any given contract that should be comparable by that player... basically, I want to grade out every offer and make sure the resulting output corresponds with what we think is wisest for the player

[flounder]

I'd be willing to devote some effort to this, but I'll need some examples of good vs. bad contracts.

[BowTieSports]

I'm willing to devote time to this as well. Quick I think you have a lot of the basics identified.

I would like to see the players do a better job of evaluating the likelihood of a team cutting them before the end of the contract and using that to weigh the value of future years on the deal.

7 Elements I would like to see weighted from the player side
1 - GM History
(if it's not too much additional coding, having a component where your past history of roster moves stays with you -- % of your contracts signed that last for the full term/75% of term/50% of term)

2 - Player age for each year of the contract
(straight forward concept and probably already in the game, but a 5-year offer which covers a player's 25,26,27,28,29 years is more likely to go to term than a 5-year offer which covers the same player's 29,30,31,32,33 years)

3 - Player position
(hopefully already in the game, but the average player at different positions burn out at different rates. RBs, for ex, burn out fast, so offering a RB an $8M base salary in Y5 of an offer should be viewed as highly unlikely to be paid out, while an $8M base salary in Y5 of an offer to an OL has a greater likelihood of staying on your team's books)

4 - Sliding scale combining how loyal a player is and how injury prone the player is
(a more injury prone player should view future years on the contract as less likely to occur, than a player who has never suffered an injury)

4b - Injury Setting For Your League
(if your league is playing with the injury setting at 100, the AI should recognize that the chance of a career-changing injury is lower and therefor future years of the contract are more likely to be paid out; if the injury setting is 400 the AI should recognize a higher-probability of a career-defining injury and reduce the value of unsecured future year dollars)

5 - Cap hit ratio between bonus money and base salary
(pretty straight forward, if 1/2 a guy's cap hit that year is accounted for from bonus money a team is more likely to keep him, as opposed to a year where 1/10 of his cap hit is accounted for in bonus money)

6 - Player Popularity
(with Popularity playing a role in ticket sales and ticket sales playing a role in team position for the Coaching Draft, savvy owners should be less likely to cut High Popularity players in salary cap moves, so greater chance that a team will honor future years of a multi-year deal)

[QuikSand]

BTS, much of what you're raising has plenty of merit, but I think we may be best off trying to keep this as discrete as possible - meaning the numbers we have in hand, more than stuff that's hidden. That's my starting point, at least. If a goal here is to help improve this for this game, then making it programmable is a worthy goal.

[BowTieSports]

Quick, sent you a Private Message with some other thoughts --- I don't want to get too far off topic on this thread if my thoughts in that message are not a direction you think is useful. But if you think that works as a starting point, I've got some thoughts on ways to use age, popularity and some other "listed" rating to create a salary range within that factor.

[QuikSand]

Okay, if a case study is what it takes to keep this moving, here's one that seems worthwhile.

In my current career (2034, sorry, the salary numbers will be a bit inflated) here's a decent free agent in the open market:

6th year LB, rated 43/43 overall, good run stopper and decent special teamer... in my view a decent MLB/WLB starter or very nice reserve LB... career has been as backup, only 3 starts and 74 tackles over 5 seasons (with 3 teams).

Minimum salary for relevant years: 1,230,000 for years 6-7, then 1,450,000 for years 8+

His salary request: B 860,000 Sal 1,320,000 - 1,610,000 - 1,970,000

To me, this is the standard "tricky" guy in FOF 7. I want to sign him for minsal, he thinks he's worth a god deal more than that. And looking at his career, he's a mixed bag - got cut from his rookie contract, settled for one 1yr minsal deal, then actually got the 3/4.7m deal but cot cut after 2 years... and now I suspect he will demand this serious deal and likely won't get it, and will end up settling for life as a roster-filler after training camp.

Okay...next post = contract options for this guy...

[QuikSand]

So, here are a few options for this decent guy, it seems to me:

1/2.5: 1yr, all salary, 2,500,000

1/2.0: 1yr, min sal 1,230,000 plus 770,000 bonus

3/7.5: 3yr, all salary, 2,490,000 each year with 30,000 bonus

3/6.6: 3yr, min sal 1,230,000-1,230,000-1,450,000 plus 2,690,000 bonus

5/12.5: 5yr, all salary, 2,490,000 each year with 50,000 bonus

5/11.0: 5yr, min sal 1,230,000-1,230,000-1,450,000-1,450,000-1,450,000 plus 4,190,000 bonus


I'm sure there are more options out there... but without testing, I believe that right now in FOF 7 we would see the first four offers get into the mix, but the last two would be shut off by the arbitrary three-year limit.

Anyway... marginal player, wants a pretty serious deal. I'm trying to weasel him down below his demands. What option makes the most sense to him?

[Comey]

The third and fourth options make the most sense. Option 3 nets the most functional money (assuming he plays two years of the deal)...if the likelihood he gets cut after a year are high, then the fourth offer makes the most sense (as it also has him making the most in his bonus, and therefore, keeping him at 1.2 seems more logical for management.

So, to me, he's most likely to take offer 4, then 3.

[QuikSand]

(quick follow-up note: this guy did indeed land a 3yr contract to his liking in late free agency: B 750,000 S 1,310,000 1,610,000 1,970,000)

[flounder]

I think our best approach would be to start with something simple and slowly add complexity as needed, making sure at each step that our algorithm still works for the simple cases. So the simplest algorithm would be, which contract has the most total money. It has to be more complex than that, though, since we know that the money is not guaranteed. So the next step would be to only count guaranteed money. That's still not good enough since the probability that the player gets the money in future years is above 0.

So one first simple step would be to try to assign a percentage likelihood that the player will receive the salary in later years, and weight the value of each contract by that percentage. This can get really complicated because you would have to take into account the likelihood that the contract would get renegotiated, an injury would take place, etc. So lets start with something really simple. Each point of current ability above replacement is worth a certain amount in dollars. If the contract exceeds that amount, there is the potential that the team will cut the player rather than honor the contract. We could come up with some formula to discount future contract years based on this. We could start out assuming that the player's current ability will be constant, then add in career arcs later.

Thinking about it some more, maybe that's even too complicated to start with. Maybe we should just come up with a way to take a league and calculate dollar values for each point of current ability over replacement for each position and then try to evaluate contracts based on that, assuming the player won't be cut. I think that's a pretty big task on it's own. If you could get a list of LBs in your league with their current ratings and salary we could try to generate this formula and see how each of those contracts compare.

Again, I'm just throwing ideas out there. I'm open to other directions.

[QuikSand]

In my head, I'm thinking that the way to weight non-guaranteed salary is along these lines:

-the potential range is between X% and 100%, where X is some number that declines over the life of any contract (like maybe from 50% down to 0%)

-the actual weight within that range is a function of the bonus-related cap hit triggered by a cut in that year... maybe just as simple as (caphit/salary, max of 1)

-the only thing that leaves out is the "big jump in salary" that I want to include, but haven't figured out an intuitive way to do so yet

[QuikSand]

(quick follow-up note: this guy did indeed land a 3yr contract to his liking in late free agency: B 750,000 S 1,310,000 1,610,000 1,970,000)

...and since I'm looking at this guy, here's his update. 12 games, 12 starts and 75 tackles for the Saints after they signed him during last offseason. And who then cut him in early FA in the following season, and (while I have not followed their travails closely) still have $14m in unused cap space as we sit in late free agency.

Now he is sitting unclaimed in late free agency asking for an even fatter 4/17.4m contract with 3.5 in bonus up front (and this is after going unclaimed through the 12 early stages already).

[flounder]

Unless I'm wrong, there doesn't seem to be a way to get salary information out of FOF in any kind of automated way. We might have to use cap cost since we can get that out of the league html.

[QuikSand]

I think the "project" needs to be more discrete than the broad-based completely rethinking of player value than you have in mind, flounder. All I'm shooting for is to give some input that might help Soecismic patch this part of the game - or at least persuade him that it's in need of the attention.

That's why I'm including the player's original requested contract as an important value. The game already generates that (somehow). Whether it's perfect or not (it isn't), it's at least already in the game, so I'd like to think we could work with that as the anchor for what the player believes he is worth based on his performance and skills.

Anyway -- really, all I want to do is make this a bit more dynamic. And I think the way to do that is to have a "preferred contract length" variable that is basically a function of whether the offer is above or below the player's expectations of what he thinks he will command in the open market.

To me, the heart of what the game needs is a boost in the ability to compare different contracts and to decide among them. Back to my post above... what should the game have to say about these two contracts to the same marginal-starter player:

1/2.0: 1yr, min sal 1,230,000 plus 770,000 bonus
5/12.5: 5yr, all salary, 2,490,000 each year with 50,000 bonus

...and so forth. If we could offer some thoughts on how this might be situationally adjusted, I think that might be something that could get picked up in a future patch to this game. I'd hope against hope it could, at least.

[flounder]

Ok. I've been thinking about this a bit, and I agree that starting from the player's demand is a much easier way to approach things. I think the problem then becomes how you evaluate contracts that are longer than the player's demand. To take your example, the player is requesting a 3 year contract of

B 860,000 Sal 1,320,000 - 1,610,000 - 1,970,000

So how can we evaluate contracts in year 4 and 5 in the absence of knowing what the player thinks his market value will be, without having some way of determining it? Or are we assuming we're Jim and we know in general terms what his career arc will be and exactly the market value for a player of a given ability? Sorry if I'm being obtuse.

[QuikSand]

Dang. I wish we had put more into this. I got the sense that the game was completely in the can, and didn't feel like investing time here would be productive.

But had I known there would be a substantive patch this far out, I think we could have put this stuff through some more paces and come up with something that might have been easy enough to plug-and-play for a patch.

Alas.

[QuikSand]

So, it looks like the game may be due for another patch. Any interest in re-stocking these fires some?

I don't feel like the current game is a trainwreck in evaluating contracts, but would there be a formula or pattern that would improve things?

[stevew]

Seems a bit too much for a patch?

Really would like to see more "pay as you go" type contract AI with guaranteed bases/bonuses in the next version among other things. Bulk 're-signings for guys like you described(veteran backups) would be great as well. Also rounder numbers(And the ability to increase offers by numbers larger than 10K when hitting +)

[MIJB#19]

Having ran into the incompetence of the player agents in FOF7 a bit lately, I agree it's time for overhaul of this discussion, although I think my line of thinking is pretty straight forward: "bonus money is everything".
So, here are a few options for this decent guy, it seems to me:

1/2.5: 1yr, all salary, 2,500,000

1/2.0: 1yr, min sal 1,230,000 plus 770,000 bonus

3/7.5: 3yr, all salary, 2,490,000 each year with 30,000 bonus

3/6.6: 3yr, min sal 1,230,000-1,230,000-1,450,000 plus 2,690,000 bonus

5/12.5: 5yr, all salary, 2,490,000 each year with 50,000 bonus

5/11.0: 5yr, min sal 1,230,000-1,230,000-1,450,000-1,450,000-1,450,000 plus 4,190,000 bonus


I'm sure there are more options out there... but without testing, I believe that right now in FOF 7 we would see the first four offers get into the mix, but the last two would be shut off by the arbitrary three-year limit.

Anyway... marginal player, wants a pretty serious deal. I'm trying to weasel him down below his demands. What option makes the most sense to him?Reaching back to this old post, I would think any player should prefer 5/11.0, simply because it has the most guaranteed money. Only when other variables come into play, 3/6.6 is the alternative, if he's young and feels strong about getting a better contract in 3 years worth more than the 1.5M signing bonus for just 2 years of service. Option 1/2.0 is only valid if this player wants to retire after that season and would have to pay back bonus money if he retires early (which in this game does not apply).

Given the background of said player, a journeyman in year six, he should take the large bonus money.