For the fun of it I am pasting this message from the beta forum that I posted today. The genesis was some amazing analysis by Amaroq, which he said was brought on by comments from the folks I mentioned above.
========== Message from beta forum starts here ==============
I love it when a plan starts to come together. Amaroq, you are the man!
Seriously, this was just what I needed -- that career arc stuff has needed to be rebuilt ever since I decoupled the real player aging algorithms from fictional (real player works great as is these days, since it is based more on real-life performance from Lahman). This code is also very interesting to write and has been the antidote for "almanac and A-Bug burn out."
So, using a "Modified Amaroq" approach (by modified I don't mean better either, just that it fit in the framework of the engine without too much code churn) I think I am really on to something!
The key is each player now has internal variables for what I call "Peak Age", "Decline Age", And "Cliff Age". These are distributed to mirror real life distributions of player aging analysis I have done. Of course in true PureSim fashion we can also modify the spreads via the XML too! So by nature of these variables we'll see much better career arc trajectory variability. An idea pretty much ripped straight from our buddy Amaroq thank you very much
Even cooler, the little green/red/yellow indicator next to potential is a great way to see that your scout "thinks" a player has still not reached his peak age (green), is sort of holding steady (yellow) and appears to be on the decline (red) . Decoupling that from the potential rating was the key! The potential still influences how *much* a player can improve as he approaches his peak, but it is not the sole indicator of "if" he will improve. So, a low potential guy can sort of creep up slow and steady in some cases if he has a late Peak Age... Of course the best guys to find are the high potential ones, that said what if they peak at 26! See how cool this is
I actually built a very similar engine about 2 years ago during what was going to be a re-write which was scrapped, so it came together pretty quickly.
Additionally "inaate" ratings are much less volatile from year to year for a given player (Like Power and Velocity). Like Amaroq said, while power and velocity can certainly improve, they can't really be "taught" hence you won't be seeing guys with 20 power at 19 years old turn into home run machines. We may want to do some analysis here as there are some implications on initial player generation. So far it looks great though.
I also apply an algorithm that will determine what the optimal level a player should be at for the best possible player development before he reaches his peak, so be careful when you rush that prospect or try to hoarde 27 year old talent in AAA
This algorithm's boundaries can be tweaked as well. Last but not least I made some major modifications to the AI's player evaluation algorithms to improve valuation of youth and potential. I think it will be the most challenging AI in PureSim history. I'm not letting the AI "peek" at the age variables but that might be something I look into to ratchet up the challenge further.
The results so far:
* Career arcs vary MUCH more, way less predictible. Some players peak early, stay at the same level for a number of years then slowly decline, others bloom late but have shorter spans of consistent performace, e.g. they start to declient young, others never really blossom, still others can tail off rapidly, again this is centered around a mean, so the edge cases are infrequent. When you think about the combinations of the three age variables it can get quite interesting.
* I'm seeing a much better average age in the majors 29.xx which is much better, and since players are playing into their 30's much more consistently I'm seeign career stats that look great.
* The homerun explosion problem seems to be gone for good.
In case you can't tell, I'm excited. This was bugging the crap out of me that a former strength of PureSim had shown so much atrophy in the PS 2006 code base. Its not perfect, but I can't wait to get it in the testers' hands
Amaroq, I'm adding you to the credits as a design consultant if that's ok with you?
