Trade logic

Temporary forum for the discussion of the Tuesday Morning project, to enhance the Madden 08 off-the-field experience.
User avatar
torontogrudlies
MVP
Posts: 837
Joined: Fri Jul 15, 2011 4:56 pm

Trade logic

Postby torontogrudlies » Wed Apr 25, 2012 12:10 am

We have another thread dealing with player valuation, which is a work in progress. The idea, as I've seen it so far, is to determine a "point" value which represents what a particular player is worth to a particular team/GM. How to actually arrive at this value is best discussed in the other thread.

Basically, a team runs this formula, and all the players beneath a certain point value, they put on the trade block. Other teams then look at that trade block, and they run the same formula on the same player, but based on different team/GM ideals and traits, plus the (likely very different) needs and goals of their own organization. Then if a player ranks highly, they're interested.

The dilemma is how to program the logic for something like this, as a form of AI. How do we know when we've found the best deal for both teams in the transaction?

As I see it, each player will have three values in a proposed transaction:

1. What he's worth to the team who wants to trade him... most likely this is going to be a very low number, otherwise they wouldn't be considering trading him.

2. What he's worth to the team who is interested. Most likely this will be higher...again, otherwise they wouldn't be expressing interest.

3. His general commonly accepted value. My thought on this was to have all 32 GMs evaluate the player, and just average the value. This number is likely to fall in between item 1 and item 2.

I have code which establishes a trade block list for each of the 32 teams. When a team adds a player to their trade block, the player is also added to a more general list which contains every player in the league who is being shopped. When teams analyze for prospective players they might want to add, they start with this general list, and can get a sorted version of the list, starting with the strongest candidate for their team. (Could also have it get all the QBs, all the linemen, all the defensive players, etc. if we needed it to.)

So, a scenario: each player has completed its list of unwanted players. In other words, "the word is out" and now the GMs begin analyzing. Team 1 sees that Team 24 has QB Smith, whom is a great fit. Team 1 ranks him a 480. Because Team 1 is expressing interest, Team 24 now looks at Team 1's unwanted list. Best player on that list (to Team 24) is RB Jones, whom they rank a 300. Next down on the list is DT Washington, whom they rank a 170.

One school of thought would be "trade Smith for Jones+Washington, Team 1 is getting 480 of value, Team 24 is getting 470, pretty even.

But: in the valuation by all 32 teams, the common consensus is that QB Smith is only worth 300. They also rank RB Jones even higher, a 400. (Jones is a 91 OVR, but got ranked so low by his team because Team 1 already has a RB with 97 OVR, and they primarily air the ball out.) DT Washington, for the sake of our argument, is by consensus a 175, pretty close to how Team 24 rated him.

So, what looked like a win/win for both sides a moment ago, now looks like Team 1 is getting ripped off. They're getting 300 of value, while Team 24 is getting 575. Lopsided, if you look at it that way.

Maybe QB Smith's value to Team 1 (the 480) should be used only to decide how strongly they want that player, and then the common value takes over? Like, they're saying "ok, Team 24, find one or more players on our list whose common value adds up to the 300 which Smith is really worth?"

(A variation of an analogy I posted awhile back: I have a new Corvette. My neighbor has a 5-year old minivan, which has 56000 miles on it but runs well. I have a family of six and the minivan would be a much better fit. The neighbor is single with a 22 year old girlfriend. Clearly each of us would be better off with the vehicle the other has, but anyone who knows anything about car values would know not to make such a trade.)

So what are the basic guidelines to set up logic for something like this? For the moment, assuming we're starting with one-for-one player trades, no draft picks being packaged in (although we will certainly want this by the time we are finished.)

Trade logic

Sponsor

Sponsor
 

Sleeper66
All-Pro
Posts: 316
Joined: Fri Jan 06, 2012 10:46 pm

Re: Trade logic

Postby Sleeper66 » Wed Apr 25, 2012 12:51 am

As I see it, each player will have three values in a proposed transaction:

1. What he's worth to the team who wants to trade him... most likely this is going to be a very low number, otherwise they wouldn't be considering trading him.


Maybe I'm misreading here, but this part of the premise seems off, specifically the assumption that, because a team would consider trading a player, his worth to that team would be low. That may be true in terms of playing/contributing, but their valuation of the player would also include his trade value.

The Kevin Kolb situation is one example. He was deemed expendable, with the assumption that Vick would be "the man," but they clearly didn't de-value Kolb, and received a solid return on that deal.

Another (hypothetical) example: I've got Stafford and Hill, but at the top of the second round a QB with a first-round grade has slipped, so I draft him. He has an 82 rating, but because of my QB situation he has very little playing value to me. However, he has enormous trade value, which I would be quick to take advantage of. I'd put him on the block ASAP, with the goal being a future first, or an excellent player at another position.

Again, his value to me on the roster and in the game is minimal except for his trade value.

In fact, if you're looking at realistic scenarios, teams trading players would likely begin the trade process by overvaluing players, while teams looking to acquire that player would try to drive down the price. If the trade partner values the player more than his current team does, wouldn't that lead to some "highway robbery" situations?

Not sure how this logic (if it is logic) would be incorporated, but that's my perception of it.

dawg317
MVP
Posts: 652
Joined: Fri Jul 15, 2011 5:39 pm
Location: Concord Twp, OH

Re: Trade logic

Postby dawg317 » Wed Apr 25, 2012 5:55 am

One thing to consider is how often teams trade players. If you were to ask my opinion, I wouldn't spend too much time on it because trades happen infrequently and most times are because team A is targeting a player or team B wants to get rid of them for whatever reason.
"Self-praise is for losers. Be a winner. Stand for something. Always have class, and be humble." - John Madden

Drizzt_13
All-Pro
Posts: 202
Joined: Wed Apr 18, 2012 7:32 pm

Re: Trade logic

Postby Drizzt_13 » Wed Apr 25, 2012 11:39 pm

Instead of averaging the other teams views just have number 3 in the list be the value to the team before need, salary, and GM preferences for age/whatever come into it. He may want to get rid of the player for those reasons but he isn't going to negotiate that way. If you have a Porsche but can't afford to pay insurance and need a minivan you aren't going to negotiate for it like it's a minivan, you're going to negotiate for it like it's a Porsche you're willing to take a little less for.
Here's a random suggestion for a formula

The value a team is willing to accept for a player is the average of the value before need and the value after need weighted strongly towards the value before need, but that weighting shifts more and more towards need as time progresses towards the end of the off season, or the draft, or the trade deadline. That way if I really need to move a player cause I cant afford to keep him the price will drop dramatically, but if I'm just willing to shop him cause I'm open to an upgrade the price won't shift much. This might also solve some problems with deciding if two teams have different, valid deals, which one will actually go through.

I was also thinking we could include a GM ratting for how good they are at negotiation, and have the difference between different GM's negotiation skill modify what a deal goes for.

We're also going to have to come up with a way to determine what players teams offer if they have multiple options.

User avatar
torontogrudlies
MVP
Posts: 837
Joined: Fri Jul 15, 2011 4:56 pm

Re: Trade logic

Postby torontogrudlies » Thu Apr 26, 2012 12:04 am

This topic has been complex to get a grasp on.

I think maybe you're saying, a team may have two valuations on a player.... one, what he's worth to the team with all factors considered (and this is probably lower, hence the consideration of moving him) and two, what the team feels he's worth to all other teams? Then maybe what the receiving team feels the player is worth?

When you talk about "before need" do you also mean without factoring in the team's run/pass philosophy, etc?

We should try to put together a few spreadsheets which we might be able to plug data into, and look at them that way?

Drizzt_13
All-Pro
Posts: 202
Joined: Wed Apr 18, 2012 7:32 pm

Re: Trade logic

Postby Drizzt_13 » Thu Apr 26, 2012 12:36 am

Yeah "before need" being just the x^(ovr-70) part before we factor in run pass philosophy and other factors, this should still factor in any uncertainty about the player based on our scouting formulas. Wes Welker was a return man before he became a star wide receiver, and I doubt the improvement was due entirely to progression so the chiefs must have not known what they had in Welker and New England somehow figured it out.

User avatar
torontogrudlies
MVP
Posts: 837
Joined: Fri Jul 15, 2011 4:56 pm

Re: Trade logic

Postby torontogrudlies » Thu Apr 26, 2012 12:57 am

Drizzt_13 wrote:Yeah "before need" being just the x^(ovr-70) part before we factor in run pass philosophy and other factors, this should still factor in any uncertainty about the player based on our scouting formulas. Wes Welker was a return man before he became a star wide receiver, and I doubt the improvement was due entirely to progression so the chiefs must have not known what they had in Welker and New England somehow figured it out.


So I think we're saying that a QB with 93 OVR will hold the same value as any other QB with a 93?

Drizzt_13
All-Pro
Posts: 202
Joined: Wed Apr 18, 2012 7:32 pm

Re: Trade logic

Postby Drizzt_13 » Thu Apr 26, 2012 1:29 am

Yes, but this is perceived OVR rather than actual OVR, so it would be effected by how much the GM in question knows about the player and wouldn't just be the OVR in the players statline. That is assuming we still want some sort of system for determining how much a GM knows about a player.

Drizzt_13
All-Pro
Posts: 202
Joined: Wed Apr 18, 2012 7:32 pm

Re: Trade logic

Postby Drizzt_13 » Thu Mar 28, 2013 4:10 pm

Just something to realize here. From 1990-2009 only 6% of trades are player for player trades, the vast majority of trades are player for Draft picks or Draft pick swaps.


Return to “Tuesday Morning Manager”

Who is online

Users browsing this forum: No registered users and 3 guests