I'll try to ignore your usual uncalled-for belligerence and continue the civilized conversation I thought we had (for a few posts).
It doesn't matter what you think, feel, or can imagine. I can say from direct experience from running online communities, and from projects I worked on related to security, that people will absolutely put in ridiculous amounts of time and effort to be underhanded in order to gain an advantage over other people in any sort of competitive environment especially if money is involved.
Regardless of what we do or do not believe people will or won't do, the safe course of action is to simply avoid implementing systems that can clearly be gamed. Then its not about what we think or feel, but what we know people won't do.
So, it is like an anti-doping approach that is mainly directed at the already good players which just want to heighten their chances to enter the Playoff-phase. I'm sure that any number of people will try to game any system, but it doesn't matter as long as they don't have a chance to succeed in the end. That, of course, still is only likely for actually skilled players unless the gaming-advantage is so huge that the skill-aspect suddenly loses its predictive power. That means it's a security consideration against those where the gaming would be especially obvious if it happens.
Doping, of course, is an issue, I agree, but it's a question of whether or not it is a relevant issue or if some ways of gaming the system that are possible aren't more relevant.
Cool. Maybe someday you'll be curious enough to actually work with the data yourself instead of perpetually suggesting other people do more work to satisfy your curiosity about random shit only you care about.
I have not suggested you do any work. Far from it. I have asked if you already HAD done so and what the results were and how you came by them.
If you would follow my suggestion to just publish the list of things you have considered, no one would ever need to pester you with their curiosity about that again.
Regression is a method by which a line of "best fit" is created for a set of data such that the sum of the distances between each data point and the line is lower than it would be for any other line. Logistic regression specifically creates a probability function derived from that method to determine the likelihood of two different outcomes.
One of the main things done during regression is step-wise exclusion of contributing variables with checks to see if exclusion significantly impacts the final equation's ability to predict which of the two states is produced. If it does not then the excluded variable is dropped because it's not worth including.
Thank you for that general explanation. Now, of course, I'd like to know what the different states you are trying to separate were? I assume the input data-points are the teams with all their match-history, but I am unsure, yet what distance between the matches/teams you are/were measuring and minimizing. Is it the distance of zsums?
That approach, of course, would ignore the whole other history of the coach, so that would answer my other question.
In a scenario where the coaches don't know that it would be to their benefit to lose somehow (like, using the data we already have for which the past-performance was irrelevant), would including the whole coach-history (or maybe its tail because of course skill can get better over time) not possibly improve the post-hoc prediction?
It was not statistically significant when I ran the regressions.
That was all the answer I needed for that particular question.
Have you considered other machine learning algorithms than regression learning?
Lets not. The issue is that you have a real problem with personal incredulity, and it interacts badly with your absolute lack of personal initiative.
You're far overstepping your mark here. I'm still in the phase to understand what it actually is that you are/were doing and I have a completely open mind towards it. That is my current personal initiative.
Of course, I have my personal experiences to guide me to questions but I have no problem doubting their generality. But, of course, I also don't blindly accept anything that clearly contradicts my personal experiences without even trying to find out what the reason for that contradiction is. And for that, of course, I first need to find out how these things are produced if I would ever have a hope of doing my own research to find that reason.