There are two major problems with attempting to answer this question:
- the absence of a uniform consensus of what constitutes a mass shooting
- the lack of a consistent primary source of data spanning all incidents – attempted or otherwise partially / fully committed
That is to say, databases that track mass shootings record and publish data in uniquified ways, and thus there's not even a consensus on the number of mass shootings that occurred within a given timeframe. For instance, the most authoritative compilers of mass shootings – the Gun Violence Archive, Mother Jones, the FBI, and the Stanford MSA – have databases of "mass shooting" that wildly deviate from one another. For 2017, the total number of "mass shooting" victims enumerated by these sources ranged from 612 to 2200+!
Even when factoring and filtering the databases to unify onto a single strictest definition of "mass shooter" incidents results in discrepancies:
Data compiled into these libraries also come from various sources, including but not limited to: media stories, research data sets, the scientific literature, books, advocacy groups, press data sets, police records, FBI records, CIA records, court documents, the BJS, and accounts detailed by the offenders. There is no centralized, dependable, consonant record of every incident that led to 3+ or 4+ deaths that occurred in any year nationwide, let alone an authoritative entity that has record of every attempted mass shooting that could have potentially caused 3+ deaths.
My point is that any dataset that you view is likely to be flawed or incomplete, although the post above with the FBI's sequence diagrams is likely the closest you'll get to an answer.