Atlanta releases 3.5 star Chipper Jones

[dougaiton]

Originally posted by kingcharlesxii
It seems like it is the current year stats that are causing the problem. For instance in the Lowell case his poor April was 30% of the evaluation. I wonder if something like 30-10-35-25 might work better, (assuming you still want ratings weighted heavily otherwise MD's way would work)?

That's an interesting point. Why would you want the AI to evaluate players more on 2004 performance than 2003, come to think of it? Shouldn't the default be 2003>2004>2002, in that they judge players on how they performed the season before.

Moreover, when do these stats update? In FA of 2004/2005, do they use 2004/2003/2002 stats or 2005/2004/2003 stats (with 2005 being 0 so far) because you're technically in 2005? That may not sound like a big deal, but the most sensible way to have stats ranked for FA (04>03>02) is not the most sensible for the season (03>04>02), if you see what I mean,


I don't mean to harp on about this, but there are too many things that we aren't being told from one version to the next. When we weren't told what MM did for ages, or clutch, or consistency, because of the 'fog of war,' that was one thing. But things like what homefield advantage effects, what mov/stuff/cont actually mean, and especially what this means, is vital if we actually want to use these features.

[dougaiton]

Originally posted by kingcharlesxii
I'm pretty sure it uses minor league stats when they have no big league experience.

I'm not entirely sure it does. The bit underneath says if no stats are available (I presume big league stats) then the weighting will be changed automatically. To me, that implies no 04/03/02 stats, and it switches to ratings.

[CoachL]

I wonder if creating a sliding scale would not be the way to go here

Ratings/04/03/02
April 25/0/50/25
May 20/5/50/25
June 20/10/45/25
July 20/15/40/25

or something like that, so that as you have a larger sample, the computer takes it into account more.

[sixfour210]

10,15,40,35 didn't work either.

Mike Lowell still gets released and so did 2.5-star, $6.7M Tim Salmon. And to top the AI off, the Marlins re-signed Lowell 4 days after releasing him.

I'm going to try what doug suggested. Say. . . 25% ratings, 0 (2004), 50 (2003), 25 (2002)?

[Henry]

Originally posted by dougaiton
I don't mean to harp on about this, but there are too many things that we aren't being told from one version to the next. When we weren't told what MM did for ages, or clutch, or consistency, because of the 'fog of war,' that was one thing. But things like what homefield advantage effects, what mov/stuff/cont actually mean, and especially what this means, is vital if we actually want to use these features.

I'm not sure that's quite fair D, these have been discussed in various threads....

mov/stuff/cont

Good Movement (Primary Rating) & Good Control = less Homeruns
Good Stuff (Primary Rating) & Good Velocity & Good Control = more Strikeouts
Good Control (Primary Rating) = less walks

home field advantage

Based on actual hard Home/Away statistics, implemented by home team having a slight advantage and away team having a sligh disadvantage.... basically enough to swing about 5 games a year from V5.


And this issue is new optional variation. We have to figure out what works best.

Henry

[dougaiton]

Good luck, and thanks for being our guinea pig!

[tward13]

Coach, I was just about to suggest a sliding scale.

The problem is that while more recent stats are more relevent, April stats are pretty much irrelevent, because it's such a small sample.

The current season should become more important as the year progresses. Also prior years should only be relevent if there was some significant number of ABs.

It gets more complicated and I'm not sure of course exactly how the programmed AI works but it would seem to me that the weighting should have overrides for insignificant statistical samples. At worst you'd be back to ratings being the deciding factor for the AI.

Perhaps the relevency of certain stats should be discounted for years with less than a certain number of ABs or IP.

[sixfour210]

Through May and so far so good. . .

Instead of getting released, Tim Salmon was traded to the Giants for Kirk Rueter and a minor league reliever.

[jbmagic]

sixfour210 how do you have your trade tendencies set too in league option?

[dougaiton]

That's one thing we haven't mentioned - the 'Triple Play' effect (thanks EA!). I don't have time to do a proper check, but if you change the way players are judged, it will affect the players that are offered in return. Seeing how little the AI will give up for a superstar having a bad 2 months at different settings would be a quick way of getting a rough idea of how powerful the evaluation modifier is.

[sixfour210]

Originally posted by jbmagic
sixfour210 how do you have your trade tendencies set too in league option?

All "Normal" and "favor heavily for prospects".

[sixfour210]

Weird, but I had ratings and 2004 stats set at 0% and 60% for 2003 and 40% for 2002 but Geoff Jenkins still got released for having a bad start.

As far as I've seen though, weighing ratings and the current year very low is the way to go.

ok, nap time. . .

[kingnebwsu]

I'm using 50% for ratings, then 25/15/10 for the rest. That way if a guy is s'posed to be good, he won't get released til he sucks for an extended period

[TRO]

In the same vein as kingnebwsu, mine are set at: 45/15/25/15.

Personally, I like the "feel" that the has the AI going off of stats but I know that I'm playing a game and the ratings matter, the AI should too.

[jbmagic]

TRO and kingnebwsu

you guys get good result with those setting..trades seem realistic, releases, fa signing, etc..?


i guess favor prospects the way to go too for trade tendencies to be realistic too...


thanks

[satchel]

File this thread under: The computer GM didn’t do what I would have done, so it’s a bug.

[BleedDodgerBlue]

I've got mine set to something like 40% ratings, 25% current year, 20% previous year, 15% third year, although now that I think about it, it might be a better idea to reverse the current and 2 years' back percentages.

The only cuts I've seen so far, though, have been career-ending injuries, and I'm at the All-Star Break...

[SoxWin]

Originally posted by Giants44
name the last 15M player to be released in reali life - it doesn't happen in the game, so it shouldn't happen in the game.

Kevin Appier of the Angels.

Owed $16 mil over 2 years.

[Henry]

I think a 50% - 10% - 25% - 15% should work out well.

Henry

[Hank Greenberg]

Originally posted by Henry
I think a 50% - 10% - 25% - 15% should work out well.

Henry

I just applied those same attributes to my league.....