Should Fumbles be treated differently than Interceptions?

It's hard to refute the significance of turnovers to which team covers the spread, both in terms of the "turnover difference" theory projecting which team is the better bet (57% over 20+ years), and in terms of how often the team with fewer turnovers in the game itself covers (75% over a typical five year span).

However, many people have argued that while fumbles may involve a certain amount of luck, interceptions are more dependent on the skill of the players and schemes involved. So our first tack will be to examine each factor individually, starting with fumbles, in our T/O Difference angle -- look for matchups where the net fumbles per game difference between the teams is at least 0.5 or more and play the team with the WORSE fumble record.(e.g. Team A is -1.0 net fumbles per game, Team B is +0.5, so difference between them is 1.5 and play the worse team, Team A).

So the overall record against the spread using only fumbles and discounting interceptions is a little worse than when we use both. The underdog plays hit a solid 55.5%, but that rises to 57.5% in games where the underdog is getting 5+ points -- which is very similar to the full (fumbles and interceptions) T/O results.

2001-2003 UPDATE:
The Net Fumble turnover difference plays have posted a good record of late, that has exceeded the long term average by roughly 2%.

Home teams have been the best wagers under this scheme over the past three seasons, accounting for a 52-34 (60%) mark, and home or away favorites giving a touchdown or less on the line are an even better 35-19 (65%), showing that yes you can still bet certain NFL favorites profitably.

Let's move on to look at Interceptions alone:

ANALYSIS: The overall results drop a little further: 54.0% for Interceptions alone, 55.0% for fumbles alone, and 56.6% using a combined net turnover number.

However, some interesting things happen with the "interceptions only" approach, namely that the bigger the difference the more successful the plays have been -- requiring at least a 1.0 net difference (instead of the customary 0.5) between the two sides in a matchup yields a healthy 375-289 record (56.4%), and 170-121 (58.4%) if you can get 1.5 or more.

So in conclusion it appears that both fumbles and turnovers have independent predictive value, although the turnover difference theory is strongest overall when you use both. A fumble is arguably more significant (or more damaging to a team), but the interceptions may be worth following alone for instances of wide variance between two squads.

2001-2003 UPDATE:
Consistent with the 18 year sample, the turnover difference theory using only interceptions hit 54% overall in this span. The truly big discrepancy plays (1.5+ net INT per game difference) were 23-18 (56%). In general of course we prefer to avoid excessive slicing and dicing of results -- stick with the basic tenets and prosper!

The Turnover Triad

From 1983 to 2000 when a game meets the fumble requirements (1/2 net fumble per game difference or more) and the turnover requirements (1/2 net turnover per game difference) without the net interceptions being under the 1/2 per game threshold, then the results are impressive: 61.4%. When the selected team is an underdog it's been even stronger at 66%!

2001-2003 ReTest: when the Fumble + overall T/O picks agree without the same team being selected by the INT numbers, the results come in as 34-17 (67%) over the most recent three seasons. Surprisingly though both favorites and underdogs were excellent plays. Conclusion: FUM + T/O is indeed a strong combo!

On the other hand, when the interception and total turnover agree, but do not have the fumble support, it's a 55% spread win history. If the fumble disagrees by picking the opposite side (rather than a "no pick"), it's strictly 50/50 territory.

2001-2003 ReTest: INT + T/O without Fumble support for the team was a handy 57-40 against the line over these three years, and underdogs were a blazing 44-25 When the fumble difference actually went for the other side it was again a 50/50 scenario (6-6), meaning it's actually been 51-34 with INT + T/O and a FUM "no pick" call.

More 2001-2003 outcomes:
All three agree (INT + FUM + T/O) = 36-29 (55%)
T/O only (neither INT nor FUM reach the threshold) = 8-9 (47%)

Parting Words:
While in any given season the fumbles alone or interceptions alone can go on quite a run -- witness the amazing "Fumble Difference" year of 2002 with a 31-13 (70%) mark -- the combined predictive power of both turnover types has over the long term been a stronger spread beating application. However, when the Fumble difference is enough to qualify as a play in its own right and agrees with the overall turnover pick, then you have a healthy 61%+ history behind you!