Gamer Floater Theory
This concept by W.A. Hawkes-Robinson began to solidify around 2004 shortly after moving to Spokane, Washington, USA. A draft discussion was posted in various fora over the years, with some drafts on older rpgresearch.com websites around 2015. 
Updates to this theory can be found at the page "Regarding Gamer Floater Theory": https://www.rpgresearch.com/gamer-floater-theory
The original summary of this theory is that 90% of the gamers seen by the public are less than 10% of actual active gamers. Furthermore that the players and game masters that stand out most in the public’s collective memory and perception of gamers, are actually the most dysfunctional members of this population, though they are less than 5% of the total active gamer population in any given region. This small, but highly visible minority, stands out due to their extremely dysfunctional behaviors, and so are unable to remain in any long-term gaming group, so they “float” from group-to-group, location-to-location, and event-to-event, craving the gaming experience.
This number is based on over 45 years of direct observations of tens of thousands of gamers, direct interactions and interviews of thousands, viewing of audio and video recordings, and game session log notes. A more formal study is desired but as of 2020 does not yet exist, and due to the nature of the majority of actual gamers never being seen in public gaming settings, very difficult to organize definitive samples.
The percentage numbers have varied broadly by different locations, but over time the number has been increasingly showing stability. Early versions of the theory posited that anywhere from 4% to 15% of the population might fit the "floater" theory, but as more in-depth observations took place over the decades, especially getting past the floaters in each location, and getting to know most of the active, and many inactive, gamers in a region, it helped revise these percentages downward significantly as far as the percent of players that would qualify as a "floater", but also further reinforced that the floater group is publicly even more visible more frequently seen by the public as a reinforcer of the negative stereotypes about gamers being dysfunctional.
This theory's general concept seems to hold up across a wide range of cohorts, locations, environments, and demographics, but the specific percentages are still difficult to pin down. Some version posit that less than 1% of role-playing gamers are actually visible to the public (perhaps less than 0.035%), and that about 1/10th of those showing themselves in public account for 90% of those seen by the public. These less than 0.0035% of gamers are the ones struggling with functioning and role-playing gaming is their primary social recreational activity, but unfortunately for various reasons they are kicked out of most groups and so “float” from group-to-group, gamestore-to-gamestore, and gaming-convention-to-gaming-convention.
The numbers are regularly being revised as further data is accumulated. As of 2020, the current revision of this theory postulates that anywhere from the higher end of approximately 3.2% (2.2% to 4.2% (+/- 1%)) ((1 per 33, 16 (+/- 5) per 500, 32 (+/-10) per 1000, 320 (+/- 100) per 10,000 gamers. ) (to less than 0.035% on the low end) of active gamers are highly dysfunctional and highly (negatively) visible “gamer floaters”. This can be per region or per large event, and is the percentage chance that any given new or pickup gaming group may have 1 or more of these highly dysfunctional participants. Note that, as the theory postulates, this number does NOT hold in most longer-running established groups, as per the theory’s understanding of the selection and removal process for repeatedly disruptive groups. Groups that either contain a larger than average number of this population, or that attempt to hold on to 1 or more in a group, generally end up dissolving withing a few weeks or months, if they do not remove the disruptive factor. In the case of game masters that fit this population, this shows up as a GM that shows an exceptionally high turn-over rotation rate of players.
The theory proposes they are getting bounced from group-to-group, and event-to-event, reinforcing the negative stereotypes about gamers even though research shows none of the stereotypes are true.
This theory has not yet performed an analysis of gamers in live-streaming on Youtube, Twitch, and similar services. This number is based on physical real-world locations at game stores, gaming conventions, schools, neighborhoods, etc.
However, the gamers that most stand out to the public are the most dysfunctional gamers leading to ongoing perpetuation of the negative stereotypes about RPGers. Through decades of observations in multiple regions.
Locations where more in-depth observations have taken place:
Observed locations include: USA: CA, WA, OR, UT, NV, AZ, ID, MT, WY, NM, CO, TX, MN, IL, IA, GA, FL. Canada (BC) and Alberta. Europe: England, Ireland, Scotland. Brazil. Russia. Bangladesh.
Lengthier and more detailed observations have been at the following locations:
• Bay Area California (Peninsula and East Bay)
• Greater LA region
• San Diego region
• Tucson region
• Pheonix region
• Norther Utah (“Intermountain West”)
• Greater Spokane, WA region and Greater Post Falls and Couer d’alene (“Inland Empire”) region
• Seattle/Tacoma region
• OKC region
• Dallas region
• Portland region
• Indianapolis region
• Chicago region
• Upper peninsula MN region
• Atlanta, GA region
• London, UK
• and others,
This may be biased by more in depth observation in countries and cultures that are more individualistic vs. less socialistic (such as USA vs. China & Japan). It will be interesting to see if the social conformity rules of more socialistic cultures alter this number in any way.
Note: Mental health and other issues do not automatically equate to a “gamer floater”. There are many gamers with life challenges, mental health, and other issues, that do not display the highly dysfunctional and disruptive behavior of gamer floaters. I really want to find a better name for this population, and welcome input.