Monday, February 28, 2005

MVP 2005 (PC) Player Progression Project

This is probably the only game I'll purchase from EA this year (due to the embargo, I now have a Gamefly subscription and get EA console titles through them as rentals), but as soon as MVP 2005 was released, the mod makers were going crazy again with hi-res uniforms and new bats and whatnot, and I got sucked in.

I'll say this: intead of adding 500 new features, EA focused on fixing what was wrong last year, and I've been very impressed so far. With the exception of Tiger Woods (PC), this is probably the best sports games EA has put out in the last year. I think with this version High Heat can officially be put to rest. It had a long run.

MVP is highly moddable, not due to any openness on EA's part, but due to the persistence of the modders and some very well-developed tools to extract individual data files from the .big files in the game directories. If you've read this column for a while, you know that last year we had a big project to improve player progression, which was totally broken in the shipping version of the game. With the help of some very hard-working volunteers, we modified the player progression system to produce stable stats for 80 years.

So I did an initial test of the player progression system this time to see what EA had changed. Last year, offense collapsed over time with the default player progression values. This year, it's the opposite--home runs went up 35% over 20 years. That's a massive increase.

That sounds like bad news, but it's not. The reason it's not is that the player progression system is entirely new and much more sophisticated. Last year, the progression system was very simplistic. Players were categorized from 1-5 stars, and for each star level, there was a potential min/max for a rating at three ages : 18, 31, and 45. Players had a defined potential (which was inside the min/max for their star rating and age category), and they inexorably marched toward it until they reached age 31. It was very linear, and their decline after age 31 was linear as well. Very basic and very predictable.

This year, much to my surprise, the system is totally different. Instead of a linear system, it is probabilistic, and it's far more dynamic and interesting. Here's how it works: For every age (by year, not in ten-year chunks) and every star rating, there is a table that lists the probability of a rating change. A random number is generated each year, and the number is compared to the table to see whether a rating should go up, stay the same, or go down.

Having those definitions available for each year makes the system much more powerful than last year.

So let's say the rating is supposed to increase. A second table looks at the existing rating and assigns probabilities for three possible categories: 1-5 points, 6-10 points, and 11-15 points. Again, there is a definition for every age from 18 to 45, not just age categories. And there are additional categories for the existing value of the rating. Another number is generated and the table is read.

Ratings declines work the same way, with their own tables.

As far as I can tell, that's how things basically work. Put all this together and it produces a player progression system of tremendous depth. It's much less linear and far more dynamic. It's now possible to model player development far more realistically.

It's also very difficult to balance. I can't imagine trying to balance this complex a system while also facing the pressure of shipping the game.

So I didn't think I was going to get into player progression again this year, but it looks like I am. And I'll need a few testing volunteers. If you'd like to volunteer, you'll need a copy of the PC version of MVP 2005 and the ability to sim a minimum of twenty seasons (which takes 2-3 hours). You can sim as many seasons as you'd like, but 20 is the minimum. Like last year, I'll try to round up some prizes for the people who sim the most seasons.

If you're interested, just hit the "email me" link in red in the upper-right hand corner of this page. Thanks.

Site Meter