11.10 AI not signing some free agents

[Keith L]

I started a brand new franchise in 10.33 (my first time trying out 10.33) and was having a great time getting it organized initially. I traded for some players, set my rosters, renegotiated some contracts, and started the season.

I played four games and decided I had better extend my closer's contract since he was off to a fantastic start and I remember reading that Clay adjusted the game to take stats into account when setting line-ups AND contract valuations.

At the beginning of the season he was asking for an extension of about 2 years @ 12.5M, so I decided to hold off at that time. However, after the great start in the 1st four games I checked again and behold -- he wants $450K/yr. But if I haggle with him gently I can get him to agree to 7 years at $400K.

Even if I intentionally piss him off repeatedly, (by leaving negotiations without finalizing after he's agreed to the contract amount), he doesn't really raise his requested amount. The best I got him to ask for was $500K.

I quit without saving and loaded up the same .mog in 10.31 and funnily enough he wants 2 years @ $9 million in that version. SO there is obviously something amiss in the new salary algorithms in 10.33.

That was a very deflating experience after spending basically all evening setting up the new dynasty and getting psyched up to try out 10.33.:(

[ohms_law]

oops...

I wonder what triggers that? I played a game out to 2012 with Tampa since 10.33 has been released, and never had that happen. Although, I don't often check my players contracts mid season.

[jkwici]

oops...

I wonder what triggers that? I played a game out to 2012 with Tampa since 10.33 has been released, and never had that happen. Although, I don't often check my players contracts mid season.

I'm out in the 2020s in 10.33, and I have certainly noticed that star players with monster stats ask for so much that it's as if they've done a full lap around the salary algorithim, and I end up signing them for just a few million a year versus the 12 million they had been making when their contract expired.

[ohms_law]

A roll over is certainly a possibility.

[Technetium]

Playing in 1906 currently. Honus Wagner and Danny Greene (both rated 95/95 Overall) went on the free agent market after the 1904 season. They demanded too much money, and as a result, did not play at all in 1905. Their skills deteriorated slightly because of this (Wagner went to 94/94, and Greene went to 90/91). At the beginning of the 1906, I noticed they were still on the market, and demanded less than $1,000 per year for two year contracts. I was surprised they hadn't been picked up. I already have 90+ players in both their positions, so I did not particularly need them, but at that price, and with 40k under budget, I could not really refuse. So I picked them up, and decided to try and trade them for pitchers (which seem to be harder to get). I could not get anyone to agree to trade a single 85+ pitcher for both these players combined. Okay, pitchers are harder to get high ratings on, so I can see that. I then decided to see what I could get for these players, and put each of them on the trading block to see what offers I got. I could not get a single offer for either player, either singly or as a pair, even though they both have less than half the contract cost of many players in the league who are rated only in the low 80s. This doesn't make any sense at all. Why are these players unwanted?

[Keith L]

Well the 2nd part I can answer. Mogul somehow artifically "marks" players you've signed as free agents for some period of time (I don't know if it's a full season or what) and doesn't really offer you anything if you try to trade them during that period. It's an attempt to keep the human player from signing all the FAs solely for the purpose of trading them.

Now, I don't know why they were on the FA list to begin with.

[Technetium]

Well the 2nd part I can answer. Mogul somehow artifically "marks" players you've signed as free agents for some period of time (I don't know if it's a full season or what) and doesn't really offer you anything if you try to trade them during that period. It's an attempt to keep the human player from signing all the FAs solely for the purpose of trading them.

Now, I don't know why they were on the FA list to begin with.

Yeah, I couldn't believe that no teams wanted to pick up either 90+ rated player when they were only asking for 700 and 950 per year for two years. Something's not right there.

[ohms_law]

If anyone has a save game where this is going on, please post it. Thanks.

[Technetium]

If anyone has a save game where this is going on, please post it. Thanks.

If I see it happen again, I will do so. Unfortunately, since I went and signed the players, I can't really demonstrate it anymore with my save file, except for the fact that the players are presently unmarketable in trade (which is apparently not a bug according to the first reply).

I'm also thinking the reason they might not have been signed by anyone is that all the other teams in the league (or close to all of them) have been having money problems, and they just might not have been able to afford the prices usually demanded by 90+ rated veteran players. Still, to go the whole year without playing...

[Technetium]

Okay, here's another guy... rated 93/93, only asking for $950 for one year, but no team has picked him up. There are some other 90s here too, and though they are asking for more, it's still a very good price for their talent (there are plenty of players getting over $15,000 per year).

http://technetium.freeyellow.com/unwanted90s.jpg

Here's the save file, too. You can also see that I am unable to get any offers for Honus Wagner and Danny Green.

[ohms_law]

:cool:
Thanks.

[rogue9]

I have been seeing this same thing playing in the 50s and 60s. Players like Lou Brock, Vada Pinson (sp?), Frank Robinson, all in their primes.

What I noticed was if you sign them, usually really cheap 10k for 2 years, play them a little you can trade them later. Other wise my findings were the same as his.

[ohms_law]

FYI: I think I know what the problem with this is, I just don't quite understand why it's a problem yet...
That save game really helps.

[Keith L]

Ohms,

I still have the .mog which contains the salary rollover (or whatever it is) illustration that I posted about at the beginning of the thread.

I've attached it in case it sheds any further light on the issue. Recall that the player in question (Mario Barron) is asking for 3yr@450K when this game is loaded in 10.33, but if the same file is loaded in 10.31, the same player requests a more reasonable $9.5M.

[ohms_law]

I've got the one from above, and (I think) I see what's happening that causes this (what's happening internally, in the AI, that is). Like I said above, I simply don't understand enough about the AI at this point in order to understand why the AI is behaving as it is. I have a suspicion that it's related to the inflation system, but I have to figure it out.

Thanks for the save though. I'm sure that it'll probably come in handy.

[Keith L]

Like I said above, I simply don't understand enough about the AI at this point in order to understand why the AI is behaving as it is. I have a suspicion that it's related to the inflation system, but I have to figure it out.


Well, I'm pretty certain we're NOT talking about the same problem. The one I've described is definitely related to the revised way contract valuations are calculated in 10.33 (taking stats AND ratings into account).

In fact, I play with inflation OFF.

I did some futher testing and observed some interesting behavior. I loaded up my game save in 10.33 and changed the new slider that affects how much the game weighs stats versus ratings to -100% and simmed forward day by day to see what happened to Mario Barron's contract demands.

Initially his demands stayed at $450K. After just a few days they started to rise. His requested salary progressed from 450K, to 2.5M, 6M, 10M -- after about a week or so he was asking for 18.5M.

At that point I reloaded the game and did the same test with the slider at 0% (default). His salary demands stayed at 450K for about a week and then jumped to $16M, rising after that and eventually settling at about 18.5M.

So with either slider setting he eventually ended up asking for the "correct" salary, but with the slider at -100% he got there more gradually.

For completeness sake I performed the test again on the same file using 10.31. The overall salary demand was lower (around 9.5M) but it did not experience very wild flucuations -- staying in the range of 8.9-9.5M for about a month that I simmed.

Conclusions/hypotheses:
(1) The issue is definitely related to the new stats versus ratings slider. There is obviously an effect as observed when the slider is changed to -100%
(2) Even with the slider at -100%, the new algorithm is still taking stats into account, since the salary demands eventually stabilize at the same level-- and that level is much higher for the player in question when the game is played in 10.33 versus 10.31 (for a player with consistently fantastic stats)
(3) The new alogithm has issues during the first few weeks of a new season with certain players. This may also be possibly related to the fact that the player in question is a closer, thus it takes quite a while before he gets many innings pitched in a new season.

I hope this clarifies the issue. I just want to make sure it's not lumped in with the other issue described in this thread because I pretty certain that they are not the same issue.

[ohms_law]

ahhhh... That should help. Good thinking! The stats setting didn't even cross my mind.
Now the real question is if I can fix it without using a kludge technique...
;)

[ohms_law]

Edit: ah hah! ...OK, I found the problem, now I just need to figure out what to do about it.
*scratches head*

The trouble comes from the fact that free agents end up receiving a 0 for the games that their "team" has played, and that value is used later on as a multiplier... anything times 0 is 0, you know?
So, the question is, what value should free agent's receive for their "team's" games played? The way I see it, there are two possibilities: the easy way would be to use a set value such as 3/4 of the season played; the other possibility is to find how many games the player has played before they became a free agent. Either method comes with it's own problems though (beyond just complexity)... I'm not even sure if I should try to fix this myself at this point, or if I should lay it on Clay's lap.

[Keith L]

I see... Yeah 0's are a royal pain in the butt. At least it wasn't a divide by 0, that could have caused a crash. Of course, those issues become obvious more quickly and aren't nearly as intermittent.

Anyway, glad you found something. I can see where you're leading with the idea of "seeding" the value to ensure it is never a zero. That sounds pretty reasonable as it's always good to check for those type of boundary conditions and clamp the value to within a reasonable range. But without understanding what the entire algorithm is, it's hard to assess the impact that would have on Clay's design intent.

Nevertheless, something needs to be done and your groundwork will obviously make that much easier now that the problem has been identified.

Thanks for the effort!

P.S. Thinking about it more... the algorithm in 10.31 was quite reasonable in most cases with determining a salary demand. Is that algorithm still present in the code (possibly commented out) or is it wiped out completely in 10.33? One possible suggestion is using the old algorithm for cases where one or more of the variables in the new one are unreasonable (such as the 0 games you identified) using a Select/Case type construct. The only issue doing that would be trying to make the transition between algorithms somewhat smooth so that there wasn't a extremely step jump when transitioning between the two. Anyway, just a thought to add to the brainstorming process in case you hadn't already considered that possibility.

[ohms_law]

P.S. Thinking about it more... the algorithm in 10.31 was quite reasonable in most cases with determining a salary demand. Is that algorithm still present in the code (possibly commented out) or is it wiped out completely in 10.33? One possible suggestion is using the old algorithm for cases where one or more of the variables in the new one are unreasonable (such as the 0 games you identified) using a Select/Case type construct. The only issue doing that would be trying to make the transition between algorithms somewhat smooth so that there wasn't a extremely step jump when transitioning between the two. Anyway, just a thought to add to the brainstorming process in case you hadn't already considered that possibility.
Oddly enough, I had almost the exact same thought process immediately after posting the above...
:)

So, I'm doing some digging through the change logs to try to find exactly what has been changed. VCS (or CVS... potato, potato; tomato, tomato [you know, that expression really doesn't work when it's written out. LOL]) sure is a great thing.
;)

[Technetium]

I'm just bumping this thread to report that the same problems I reported are still happening in 11.10. Every year there are superstars in the free agent pool who don't get picked up the entire year. Then the next year they only ask for a tiny cost for one year (like $1,000 in 1931 for age 23 player who is rated 88), and they still don't get picked up. It looks like, in this thread, it wasn't corrected yet, and I was wondering if anyone was still working on fixing it.

[HoustonGM]

The strange thing is that this doesn't happen in modern times...