|
Grade 1
Join Date: Jul 2008
Posts: 311
|
Still doesn't make sense
Quote:
Originally Posted by Dorianmode
Thanks for posting this. If you... one of the leading lights here, have these questions, it makes me think I have done a very poor job explaining myself.
"the 95% confidence interval is a fairly useless way to measure the value of a horse's performances, if you're only looking at speed figs"
Correct. I don't just look at speed figures. One dimensional handicapping is not for me. I use NP to make SOME easy eliminations. The range may be wide, (and often is), but statistically, a horse with a significantly different range than another, requires that BOTH, (ie the probability is multiplied).. have to run outside their (95% SD) range, (if something "different" is going to happen). I only use it to help identify the non-contenders. I DO understand Mr. Schwatrz's reasoning, it does not appeal to me, entirely, using it the way he does. The reason I suggested there needs to be the calculation of a Confidence Interval, and a Neyman Construction, is to complete the probability picture, (as I said in one of my first posts. I care MORE about my win percentage, and frequent wins, than I do about high priced, less frequent wins.
Your elimination method basically seems to involve throwing out all the non-NP horses and then checking the remainder vs. RDSS.
No. I do not even look at whether a horse is a "NP" horse. I only look at the ranges, and the total range picture, to get a feel for the competitive nature of the entire field. Many times NP has as one or more of it's horses, an animal that could not possibly win the race. I would NEVER bet a dime on many of them.
"Why you or anyone else would do this mystifies me, since NP is a much weaker handicapping factor, with an average hit-rate of .18. The top two choices hitRDSS (either TE or V/DC) hit 60% of winners, whereas the top four of NP (which offers no rank-order) hit only 74%."
Exactly. That's why I NEVER use NP to pick winners. After I have made the eliminations, I ONLY use RDSS. The fewer contenders I have, (3 or 4 at the most), the clearer the RDSS picture is.
"To put it another way 97% of NP winners are already contained within RDSS. So, if you are going to go to the trouble of handicapping with RDSS, why bother to 'confirm' it with NP?"
Exactly backwards. I do not use NP to confirm anything. I use RDSS to pick winners, AFTER I have looked at a number of things to pick the the contenders, and eliminate the others, (including NP). I do not use it the way Mr. Schwartz suggests, for the reasons I have written about above. It's statistical validity, as far as it goes, makes sense, and I'm sure he will improve it. For now, all I know is, it help me eliminate non-contenders, quickly and consistently. 
|
DM,
What you say here seems to contradict a number of your handicapping examples in this thread. You open many of the with "a,b,c,d, horses out by NP elimination". So you begin by eliminating non-NP horses, and then cut out one or two more by traditional handicapping criteria (odds too high, bad PP's, long layoff), and then check the NP contenders vs. the RDSS horses. My point is that, if you've already handicapped the race with RDSS, you already have 97% of NP winners plus an additional 10% of winners that NP would have elminated. There's no contest between the two. And because it is rank-ordered, RDSS is much stronger against a broad range of races. So why not just use RDSS?
In addition, you make no response to the value of Dave's weighting of variance (form, form cycle) as a factor, even though you say you understand what he's doing. Also, if you hve the background to do Neyman pseudo-experiments to provide rank-ordering for NP, I think most people here would find it helpful.
Cheers,
B Jennet
|