Re(4): What do y'all think of THIS pattern? Posted on April 22, 2019 at 01:22:35 PM by Tiger
Tough to do in 25 words or less. Problem is, every time I look at it, I realize there's more. No matter how deep you go, it can always be better.
But yes, I think the perfect system (rating + simulation) can be made. Assuming your 'ratings' are exact (more on that later), the simulation takes all post position variability out of the picture - doesn't matter who's in which post. Of course, your simulator must process those ratings correctly on a per-point basis. And your output is still gonna be only a probability distribution, not a magic bullet.
And, we haven't even mentioned where your wagering strategy comes in (pool sizes, bet dilution, spreading your bets, what Joe Public is doing, etc.).
Your 'rating' has to take into account all the factors we've discussed here multiple times:
Hot or cold (how much emphasis on recent history, and how recent is 'recent'?)
Serve/receive ability (overall and vis-ŕ-vis current point opponent)
Style (e.g. strengths/weaknesses) vs. this point's opponent(s)
(could put in a few more)
It should be obvious that ratings must be recomputed with each point.
The Skiena model doesn't look for any of this. Mine do.
It's important to note that most of my mechanized system(s) are designed as "no-see-ums." That is, they were made to operate with only published results and entries (as adjusted for scratches). That's because it was difficult to get to the fronton or otherwise view the live action. Replies: