Thursday, December 17, 2009

Madden Data Project #2

I'm still in the early stages on this, but what I've done (and thanks to the help of a few volunteers) is compile a play-by-play database for the last two minutes of the first half for the first ten weeks of the NFL season.

Here's what the main play description looks like:
Ben Roethlisberger Pass to Heath Miller to Pit32 for 5 yards.

There's also another column of data for field position:
1st & 10 at Pit27

So what I did was extract bits of information out of those descriptions, and create play type identifiers (run, pass, sack, etc.) based on the extracts. Doing that lets me compile ten weeks of plays into categories in seconds.

Excluding kicking plays, it's a total of about 1,800 plays.

What I'm trying to demonstrate is how incredibly aggressive NFL teams are in the last two minutes of the half, because in doing so it will illustrate how very passive the CPU is in Madden in comparison.

I'm hoping to do lots of different things with this data, but here's a sample. Take a look at the run-pass breakdown in the last two minutes of the first half by field position:


As a note, the numbers over 50 represent field position in the opponent's half.

Basically, if a team is between their own 20-yard line and the opponent's 10, they're throwing the ball at least 80% of the time, and incredibly, if they're between their own 10 and 20, they're still throwing the ball almost 60% of the time!

Like I said, teams in the NFL are incredibly aggressive in the last two minutes of the first half. I wasn't able to attach a game score to the play-by-play (that's not how the data appears), but with overall passing rates of 80%, there's no way that teams who are ahead are playing conservatively--it wouldn't be possible for the overall passing percentages to be that high if they were.

Why is it like this? It's a simple answer: in the NFL, you're either at an advantage or a disadvantage, and teams are absolutely merciless. There's almost never a situation in the last two minutes of the half where one team runs out the clock and the other team lets them, unless they're out of timeouts. No matter the score, there's still an entire half to play, so no one sits on a lead. Ever.

I'm going to keep looking at this data, slicing it in various ways, but I'm also going to start keeping a database of CPU play calling in the last two minutes of the half in my Madden franchise, because the difference in aggression between Madden and the NFL is huge, and I'd like to be able to demonstrate it very clearly.

Now if you want to contribute to the project, it's easy. If you're running an offline franchise in Madden, just use this format to record information in the last two minutes of the first half when the CPU is on offense (just record pass or run in the "play" column, and a sack counts as a pass for play calling purposes). Also, please let me know the game score as well. Send me the information and I'll put it into the database.

I'm hoping to get a database of 50+ games, at least, which should provide a good data sample. And I'll keep the games I'm playing separated, which should tell me if sliders/difficulty settings have an effect on CPU playcalling.

Site Meter