It was not announced as a title. It was announced as a confirmation that Halo, the IP, would make a splash on the platform.
This is incorrect. Earlier in the thread I posted direct quotes from the conference where the speakers were clearly talking about a single game (while also promising that more than one game in the franchise would appear on the platform).
...that does not eliminate the skew. Counting only games announced AND released eliminates the skew.
Both approaches address the skew, in different ways. What bias are we trying to remove? That Sony has announced games more recently than Microsoft, leading to falsely short intervals. By setting a
terminus ante quem at E3 2015, we are both eliminating the possibility of Sony announcing later in the same year (since both platforms present at E3), and chucking all games still in development for less than 2 years. Obviously, this strikes out every small number, and levels the playing field.
This matters little, though, because in fact I also presented results for only released games. This does not reverse the comparison.
I went through and found each individual game you listed for the platform with the smaller number of titles and used the info that I gathered myself. I did not use any existing data set. If you have a data file there is no rationale that makes sense for keeping it for yourself.
There is a pedagogical rationale.
You initially posted results with all manner of screwy methods skewing the data and then after others called you on it you provided more results.
This is an inaccurate recapitulation of events, indeed so wrongheaded that it's scarcely credible that you've offered it in good faith. The "all manner of screwy methods" was to include every game from every conference. The "more results" I provided were not expansions of what I initially presented, but invariably filtered subsets.
That is surely better than nothing, but it also suggests that your intention was to get a certain result to report back to folks here instead of doing good research for its own benefit.
It should be very easy for you to disprove everything I said then, by posting the "real" data.
The way you react when these criticisms of your methods are levied is quite telling, as your primary rebuttal always ends with a pronouncement that the 'results' do not change regardless of methodology. Of course the results change! By 'results' you seem to be looking at the ranking between the two companies only. The question is why is that your focus?
Because the competing platform is the focus of your criticism. Here are some of your points:
"...yet you later seemed to suggest that you abandoned this when deciding when to count Halo: TMCC as being announced."
"Starting at E3 2013 would seem to strongly favor one competitor here over the other..."
"Your methods skew the data to favor the platform with more titles recently announced."
"I checked the 18 titles you said you used in your Xbox One analysis [but not the Sony titles]..."
Instead of being concerned with what constitutes good methods you instead seem to care primarily about the ranking between these two companies.
Again, you seem not to have read the thread at all. You claim this despite the fact that many of the further conclusions I posted do
not compare the platforms. That comparison did feature in my OP--warranted by the surprising nature of the results--but only remained the primary topic later due to thread interest. Please observe that when bringing my own analyses, I discussed general topics like the individual character of press events without distinction as toy platform; further comparisons were wholly in response to others. Further, in my recap post only one of the four prior expectations I address is comparative. And I repeated my belief in the primacy of complete results over comparisons later.
If you'd like to discuss any topic besides the accuracy of the Microsoft/Sony comparison, I'd love for you to bring it up. I'd much prefer that.
If the research questions to be addressed is: How long do games on platform X have between announce/release compared to games on platform Y?
...then you should be including games that are announced, regardless of the venue. The venue has nothing to do with your research question.
You seem not to have read even the title of this thread. The question to be addressed is emphatically not "How long do games on platform X have between announce/release compared to games on platform Y?" The question I set out to address is literally the first question in the OP:
Exactly how long has it been taking games to come out after they're presented?
My "flawed" method was chosen to answer a research question you claimed to be addressing. Your methods do not address the question. If you want to understand how long games on each platform has between announce/release, you need to have release dates.
This glosses over a fundamental issue: the set of games we're looking at have not all released. There are three ways we can deal with that bias. Most conservatively, we could simply wait for them all to be released (or canceled). This is boring. At the other extreme, we can do as you (and
Anderson DL) suggest, and simply drop all the data that's hard to parse. This isn't an invalid approach, but if we're interested in future probabilities it has a drawback: it assumes the premise that currently unreleased games will exactly replicate released games in their timing and pattern.
I chose a third approach, which was to allow that real-world data sets are messy, but attempt to incorporate the full content of our current knowledge. Since that knowledge is incomplete, explicit disclaimers about the inevitable skews and limits of inference are inevitable. I chose to believe the audience could mentally accommodate those, and that discussion would be better served by a complete (so far) set of data rather than a carefully scrubbed one.
You are supposed to be picking a good methodology for a given research question and letting the data filter out as it will on its own. Not picking and choosing your methods based on which outcome you favor.
I'm at somewhat of a loss what you're proposing occurred here. You think I collected all the data, then analyzed it in many different ways until I found one that supported my bias? Since I'm a stranger whose character you don't know, this assertion must be based on the traits of the results presented. But I don't see any warrant for such a conclusion from that source.
For one, some of the alternate filterings suggested by you and others give an even greater disparity in Sony's favor. If I was pre-selecting, why wouldn't I have posted one of those? Second, the data set I chose was all games, from all conferences, at their first appearance only. Every proposed alternative is more restrictive in culling this data set, which suggests that I didn't overfit my initial scope. Finally, and most trenchantly, all those alternatives give results which are cardinally different, but ordinally the same. I think you'll agree that this strongly suggests the comparison to be a robust, pervasive trait of the entire set (given that so many subsamples exhibit it).
Ha! I'm fairly confident you understand that this is NOT what I was referring to.
It was an attempt at dry (perhaps not very funny) humor.
I appreciate that you saw it.
I was saying you need to include an effort to incorporate Bayesian analysis within your data if you want to claim you are addressing a future probability. ...It is a lot more work, so I don't blame ya for not wanting to do a rigorous analysis, but at the same time you can't honestly claim to be addressing the research question you had posed without it.
I will cop to a lack of scientific rigor here. If the only valuable approach is journal-publishable research work, my efforts fall short of that ambition.
However, I must reject the conclusion that what I've presented is fully meaningless with regards to assessing the pattern of presentation-to-release intervals. At the very least, it provides a halting start toward a fuller and unexceptionable analysis, and a productive foundation for discussion that can guide and shape the growth of such a putative grounded consensus. I don't believe a mere biased screed devoid of apposite cogency could serve as a target for nuanced discussions of inferential technique, for example.
I am, as are we all, crooked timber. Yet the things we build need not be straight in order to be useful, because the edifice is built through iteration. In the end, every involution is just decoration on its piled surface. I invite you to expand and improve upon my little work, if you have the inclination.
