Evidence or observation is only of value as it relates to a falsifiable claim
Traditionally, a "theory" needs to have two main properties:
1) It explains the observed facts
2) It makes predictions (often formulated as "it is falsifiable" - an equivalent statement dealing with incorrect predictions)
RE: Market share / saturation. The main contention is irregular traffic pattern within the period. So, you might top out at 6000, and a typical day (in three-hour chunks, for brevity) looks like this:
Then one day you get
Obviously, the later traffic should be independent of the earlier traffic, whereas proponents of Throttle Theory argue that an early spike causes
The problem, as Nick points out, is that Throttle Theorists contend that sometimes the throttle is "switched off" and they get "normal" traffic with "natural" variation. Rather than seeing this as a challenge to the theory, they see it as evidence that Google is deliberately and maliciously "siphoning off" traffic.
Further muddying the waters is the matter of traffic ceilings. Say I posit my ceiling is 6000, but my series runs
Showing that to someone is unlikely to convince them of a traffic ceiling.
Now, the most credible report of throttling is where a site has natural traffic patterns way below the ceiling, but a spike takes them towards the ceiling and then traffic is lower than usual for the remainder of the day. The ceiling is only ever approached on a Spike day, and traffic is always lower in subsequent hours to maintain the quota.
I have not seen conclusive evidence of throttling for many years (there used to be more examples circa 2008, time of the "buzz-cut"), but I am inclined to think it exists (I am unaffected) as an unintentional artefact.
If you are interested in reading less polarised discussion on the topic, I recommend dipping in to some older threads: Tedster discusses basis in patents
[webmasterworld.com] internetheaven provides case study
[webmasterworld.com] Possible mechanisms
[webmasterworld.com]Though why remains unaddressed Yo-Yo Effect
[webmasterworld.com]The most popular explanation, now superseded by Personalisation