Ratings (Editing Question)

[mgadfly]

If I was going to try to create a roster update are there any guides to what type of ratings should be used?

For example, if there is a college player who shot 48% from three, 92% from FT and 54% from the field, what ratings might be used in the editor?

Likewise, what rating makes a great rebounder vs average rebounder?

Anyone know if someone put together a guide on this?

[PointGuard]

Try sending a PM to NCAAhoops. He hasn't been around much lately, but MAYBE he will see it. He created a mod with real players so had to figure this out. Also you might want to take a look at his mod for Greatest Teams: viewtopic.php?f=319&t=31672

[mgadfly]

Thanks PG.

In the meantime, I've figured out:

1. How to export a .db file from the game. (easy enough with in-game commissioner screen)
2. Locate the .db file and import it into my SQLite editor.
3. Export a CSV for all players in the game.
4. Import the CSV into a python editor and use Pandas and get the minimum, mean and max (as well as standard deviations) values created by default within the game for a good idea of how players should be rated. Note: I actually separated all these values out by percentile for positions played to get a real good idea of the distribution based on position.
5. I can edit those values quickly within Python or the CSV itself or the SQLite program.
6. I can export the SQLite DB.

But then ... the game won't import the SQLite DB.

I found this thread with a reference to MDB Viewer.

But, at least for me, MDB Viewer Plus doesn't recognize either my SQLite DB (edited) or the .db file the game exports initially.

I also found this from last year:

Another thread about MDB Viewer Plus and Access

I think it is very possible to export the .db file. Automate a lot of the editing by scraping NCAA statistics from the Web and writing a python (or whatever) script to convert those to DDSCB consistent ratings files. There would be huge drawbacks to an automated conversion process, but it'd be helpful in at least getting an update started and then people could manually tweak their teams/conferences/whatever. However, I don't have Microsoft Access because I code as a hobby and use my Mac, Python and SQLite editor for that.

If anyone happens across this and can give me an idea of how to actually get an SQLite database file imported back into the game, please let me know.

Thanks again for trying to assist PG.

[Rizzo]

I am not versed in this process at all but if you are able to pull something like this off, it would be pretty amazing. My only thought is how will the script recognize that the stats from a player from a tiny school averaging 20 points a night against bad competition should not be rated close to a player at a major school averaging 20 a night?

[mgadfly]

I am not versed in this process at all but if you are able to pull something like this off, it would be pretty amazing. My only thought is how will the script recognize that the stats from a player from a tiny school averaging 20 points a night against bad competition should not be rated close to a player at a major school averaging 20 a night?

There would be no perfect way to do this since it is somewhat hard to determine what players translate well in real life from smaller schools to larger schools, but with the number of transfer students in the past 5 to 10 years skyrocketing there might be enough data to make some guesses.

Take even last year at Gonzaga for example. They had two guards transfer in. Admon Gilder from Texas A&M and Ryan Woolridge from North Texas. They both played pretty well. But Ryan Woolridge played closer to his North Texas stats than Gilder did to his Texas A&M stats. GU's SOS (+3.13) was a big step up from North Texas (-3.48) but a big step down from Texas A&M (+10.62). We might assume that it'd be Gilder that would benefit from weaker competition while Woolridge experienced difficulty adjusting to tougher games. Or look at Brandon Clarke who improved his PER from an already pretty great 29.1 at San Jose State (SOS = +1.74) to an astronomical 37.2 at Gonzaga (GU SOS in 19 = +5.01).

Gonzaga is a weird example because it typically plays a really rough non-conference schedule paired with, in recent seasons, a vastly improved WCC schedule of games. But they are interesting because they have guys coming "down" from power conferences and schools as well "up" from lower level conferences.

But maybe there is enough transfer data out there at this point to give us some expected improvement a player would have against lesser competition. The conversion formula could multiply a raw statistical score by a coefficient based on Strength of Schedule. For example (and without doing any data analysis at all), if 18 ppg at Kentucky ends up with a Scoring value of 42, with everything else being equal, 18 ppg at Eastern Washington results in a 35 after both are multiplied by SOS multipliers.

And one other thought, Ken Pomeroy tracks stats vs A & B level competition. There certainly should be enough data out there to get an idea of how much better major college players are performing against tougher teams than their cupcakes. Use that information to inform the SOS multipliers for ratings creation.