Which stat source is less wrong?

[jane2geo]

I use forced fumbles in my NFL model. It’s a very small factor for sure but does anyone know
who’s stats are less wrong? When I scrape PPD I get 439 total FF’s. Of course these are real time and are often adjusted after the fact. Espn, Yahoo, and NFL.com have concurrence on only 6 teams and list totals of 500, 407, 453 respectively for the 2014’16 game regular season.
Thoughts or insight are welcome.

[a4u2fear]

I'm not sure which are right or wrong, but I use ESPN for most of my stats and while I can't say they are 100% right, they are helping me beat the lines...

Just a suggestion - do not use fumbles recovered but use total fumbles. It's been proven that recovering a fumble is extremely random but if you're forcing more there is a good chance eventually you'll get one or two recovered.

[jane2geo]

- do not use fumbles recovered but use total fumbles. It's been proven that recovering a fumble is extremely random but if you're forcing more there is a good chance eventually you'll get one or two recovered.

Agreed, just looking at potential recoveries from FF’s.

I think I’ve resolved my quandary. When comparing player FF rather than just team forced fumbles I found that the higher numbers from ESPN include FF’s from offensive players, ie. Geno Smith had two FF, …really! Not exactly the type of FF that I want in my predictive numbers.

Thanks a4u2 and anyone else that gave any thought to my question.