Rx2, Rx3 vs PTO, MLO

[mick]

These are predominately mid-to-low-level claimers on dirt with a few S Alw, OC and Alw. No turf, mdn, mdn clm, 3YO or stakes. The races are recent, all since the first of the year. They are from PRX, MVR, OP, FG, GP, TAM and LRL. Field size is 6 or greater.

No pacelines were changed. All are RDSS default pacelines.

43123

[mick]

http://i63.tinypic.com/2w6tr83.jpg

[Bill V.]

Hi Mick

I have no idea what these charts are showing
Can you explain ?

Bill

[mick]

Hi, Bill.

I posted a reply several days ago but apparently it’s in the land of the missing, so let me try again.

These charts are an attempt to show “flat bet” performance of Rx2, Rx3, Morning Line Odds and Post Time Odds rankings. I was particularly interested in comparing the two Rx measures with MLO and PTO.

For example, the first column of the first chart indicates that post-time favorites (i.e. PTO # ranked #1) won 68 of 194 dirt sprints in my database for a hit rate of 35%. The average payout was $5.15 with a $1 ROI of 0.90.

Drop down to Rx3 and you’ll see that the top-ranked Rx3 horse in my 194 sprints won 54 races for a hit rate of 28%. But the average payout for those winners was $6.94 with a $1 ROI of 0.97. So, on a flat bet basis, the top-ranked Rx3 horses almost broke even.

The right-hand charts are filtered. I removed races in which the win payouts were less than $4.00 (i.e. odds-on) and greater than $40.00. I wanted to remove the extremes that skew/distort the overall results. I tried several “orthodox” statistical methods for eliminating the outliers, found nothing I thought was satisfactory, and so I devised my own. (My rationale was simple. You don’t need a computer program to bet odds-on favorites and no computer program will help you find many $40+ winners, so I filtered out both groups.)

I think these numbers underscore things we already knew and principally, that the value is in the lower tiers. I apologize if I presented the statistics in a format that was difficult to interpret.

[Bill Lyster]

Mick:

Nice work, but having done so much of the same, I have a couple of questions and observations.

I am always amazed at the accuracy of the ML makers. 90% for sprints is pretty good. A couple of seasons back the ML at GP was over 90% for ALL races, no exclusions! Of course not all ML makers have the same criteria. I always thought it would be interesting to ask the ML makers what factors went into their selections and odds determinations to see if there was an area of evaluation that we had not tapped as yet.

The other intriguing aspect of your research is the ROI of horses ranked 6th or higher. Have you attempted any research to see if there are any common bonds with those higher ranked horses? Also, since the ROI's are so large, is it possible that you only used the winner of the 6+ category and not considered that you would have to bet on all of the horses ranked in the 6+ category.?

Occasionally there are ranking ties. How did you resolve that? If two were tied for 2nd, did the 4th horse get a 3 rank or a 4 rank?

Default pacelines, do I presume correctly that you are using best of last 3 Perceptor or did you use one of the other selection mechanisms?

This info, as presented, dovetails nicely with my research in many ways. Keep up the good work.

[mick]

Mick:

Nice work, but having done so much of the same, I have a couple of questions and observations.

I am always amazed at the accuracy of the ML makers. 90% for sprints is pretty good. A couple of seasons back the ML at GP was over 90% for ALL races, no exclusions! Of course not all ML makers have the same criteria. I always thought it would be interesting to ask the ML makers what factors went into their selections and odds determinations to see if there was an area of evaluation that we had not tapped as yet.

The other intriguing aspect of your research is the ROI of horses ranked 6th or higher. Have you attempted any research to see if there are any common bonds with those higher ranked horses? Also, since the ROI's are so large, is it possible that you only used the winner of the 6+ category and not considered that you would have to bet on all of the horses ranked in the 6+ category.?

Occasionally there are ranking ties. How did you resolve that? If two were tied for 2nd, did the 4th horse get a 3 rank or a 4 rank?

Default pacelines, do I presume correctly that you are using best of last 3 Perceptor or did you use one of the other selection mechanisms?

This info, as presented, dovetails nicely with my research in many ways. Keep up the good work.

Bill,

Thank you for your kind words. I have reviewed your research and having a sense of the work involved, I find it humbling. Compared to your efforts, mine are rather feeble. But thank you.

I, too, am amazed at the ML makers. I know they're on the premises and they know what's in the stalls on the backstretch and they have access to the clockers, etc. but still, those are astonishing statistics. In my dreams, I'm that good, but not when I'm awake. :rolleyes:

I have not attempted any research specific to those Tier 6+ winners and their hefty payouts, but I have been thinking about it. Wouldn't it be great to have a combination of factors that might tip off a big effort by a high odds horse? Perhaps it isn't possible but I continue to doodle. (An acquaintance Randy Giles wrote a book about doodling "pace pictures." I've doodled those and other things, trying to make mental connections.) It wouldn't have to work that often, would it?

Regarding ties, I can only respond to the values that I manually added, i.e. PTO#, Rx2#, Rx3#. (What Ted's export algorithm created for the winner in Excel is what I copied and pasted into my Template spreadsheet. I didn't dig any deeper with those values.) In your hypothetical, I would have assigned the 4th horse a "4" rank, not a "3". So, it would be 1, T-2, T-2, 4. I've played in a lot of golf tournaments and that's how the prize money was distributed, but is that how you do it? If you're doing it differently, then I need to conform to your method.

Regarding pacelines, yes, they are all Default, Best of Last 3 Perceptor. Occasionally, I would change a line but for purposes of the research, I would switch it back for the export. And in all honesty, I'm not as good as Mr. Default. I've tried picking my own lines but it doesn't improve my hit rate or my ROI. For the present, I'm letting him do it, except when it's obvious that the horse won't repeat it.

Thanks for your encouragement, Bill. I'm still thinking about those 6+ horses.

[Bill Lyster]

All my research was best of three perceptor and as I became aware of how the machine rated ties, I converted to the 1-T2-T2-4 ranking.

The only problem I had with some of the perceptor lines was when they would use a race with an adjusted speed rating of say 85 from something other than line 1 (young/improving horses given more leeway), where 77 was the previous high from the last 10 races. I just assumed that the track variant from that day was off. In my own handicapping I try to use a line closer to what the horse usually runs.

Your observation on last two being best for older horses and not repeating because of likely form cycle issues (my interpretation of your words) is interesting. I will have to look at that relative to results to see how it stacks up. Nice angle.

Bill

[mick]

The lifetime "paired tops" for older was a concept that originally came from the Sheets and Thoro-Graph guys. Jim Quinn discusses it in his lastest book, The Complete Handicapper.

I would give you the page number but I'm traveling and away from my books. One of my oldest friends, we met as a high-school sophomores, is getting married for the fourth time. :rolleyes:

[Lt1]

Nice work Mick When it comes to horses 5 &up I also do not use lifetime best races as my paceline. I haven't done any survey but I know based on my capping they rarely repeat for win. I do however keep them in the mix for exotics.
Tim G

[Bill Lyster]

The lifetime "paired tops" for older was a concept that originally came from the Sheets and Thoro-Graph guys. Jim Quinn discusses it in his lastest book, The Complete Handicapper.

I would give you the page number but I'm traveling and away from my books. One of my oldest friends, we met as a high-school sophomores, is getting married for the fourth time. :rolleyes:

I always enjoy a guy who knows how to throw a good party every couple of years, don't you? I've just had one, a little over thirty years ago.

If you could screen shot the pages of Quinn's book, I'd love to read about it.

I will PM you with my contact info.

Talk to you soon.

bill

[Bill Lyster]

Mick:

I found on line articles and portions of the book I was able to download and print. Don't need page copies of what you referenced.

Bill

[mick]

For anyone following this thread, you can find "paired-tops" discussed on pages 44-48 of Quinn's book The Complete Handicapper. I think it's a fine book and recommend it.

If you want to read the pages but not buy the book, you can find them here:
http://books.google.com/books?id=Au-efGdphpgC&pg=PT75&lpg=PT75&dq=paired+tops+jim+quinn&source=bl&ots=vtH6wHO7yn&sig=nMcCRiQeqgS8RJD9VURP3QQ5tA0&hl=en&sa=X&ved=0ahUKEwiFpafqlsfTAhVl7IMKHYwDBrsQ6AEIUjAL#v=onepage&q=paired%20tops%20jim%20quinn&f=false

And finally here's a DRF article Quinn wrote on the topic some years back that focused on Big Brown:
http://www.drf.com/news/what-follows-good-pair

[Lt1]

Thank you Mick.
Tim G

[mick]

Here's an example of paired lifetime tops from my play today. "Being There" was the 4-5 post time favorite. As an older gelding with paired lifetime top speed figures, I anticipated a regression and bet on the 5-2 second favorite who won. Being There ran third.

[Bill Lyster]

Nice example Mick, actually looks like a pair of them, looking at L4 and L5, so just curious, how far apart (separated) can the ASR's be to still be considered "paired tops" as opposed to little bits of improvement or regression?

So for the sake of argument and using the most recent lines in a theoretical example, if a horse ran 84, 82 is he getting better or is that a "paired top" and vice versa if the horse ran 80, 82 - horse getting worse or "paired top"?

Thanks as always for you input.

[Bill Lyster]

Thanks for the example Mick. Was wondering, what the max spread between the two last races before you consider improvement or decline. Would 83-82 be paired, would 84-82 be paired or show improvement, would 80-82 be paired or show a decline?

Your example actually shows two sets of paired performances, nice!

[mick]

Hi, Bill.

On pages 45-46 of The Complete Handicapper, Jim Quinn states, "[T]he paired-figure patterns do not require identical speed figures, but figures within a length of one another. The figures will span within two, occasionally three, points of one another."

I do not know what an ASR point equals. (I also do not know what a TrackMaster SR point equals either, despite considerable research.) I know RDSS uses 8.3 feet for a length and since almost everything else in the program is velocity adjusted, I suspect that the ASRs are velocity adjusted too. It's one of those things that our friend is reluctant to talk about, but we do know that the faster a horse runs, the more lengths he covers in an increment of time. None of this, however, is answering your questions.

When I'm looking at the ASRs, I'm mentally using a Beyer-scale approximation, i.e. about 2 points per length in a route, about 2 1/2 points per length in a sprint. I may be completely wrong about that but I made my own speed figures for a time and used the Beyer scale as my starting point, so it's seared into my brain.

What I saw with Being There was paired tops framed in a new form cycle after returning from a layoff. Are they lifetime tops? I couldn't be sure but they're at least two points (one length, two lengths?) faster than anything else in his last ten races and more than that in most of those races. I didn't expect the older gelding to run another 84 and bet away from him successfully.

As far as fluctuations of a point or two from race to race, I don't know how much a player can read into that. Setting aside form, a point difference might be due to rounding in the tenths or even hundredths. (I adjusted my speed figures with Trakus data and the distance delta of the trip might move the needle two or three points.) It would be helpful if I knew exactly what an ASR point represented, but I don't so I'm flying by the seat of pants on this.

Here's another example, but this one represents the paired figures below the lifetime top. (Quinn, pg 46.) I thought "Justification" might have another good effort in him and he did, unfortunately. I bet him and the Place horse, that he caught at the wire and beat a neck. He was at even money while the Place horse was 8-1. Perhaps a two-point difference is significant, or marginally so.

[mick]

Having posted my confession that I didn't know the value of an ASR point, I decided my penance would be to find out. So I created a spreadsheet and populated it with about 150 races that had a DTV = 0.

The following pertains solely to beaten lengths at the finish and all my information is derived from the RDSS Original screen. BTW, RDSS uses 8.333 feet for a horse's length while TrackMaster uses 8 feet (or so I've read). Of course, the number of lengths a horse covers in an increment of time depends on its velocity and a horse's velocity will change during a fraction or a furlong or even a few strides. Some of Doc Sartin's algorithms deal with that. (I think he had a JPL mathematician's help.) I make no such pretense. Consider me Fred Flintstone at the Rock Pile while he's launching moon shots. But this is what I was able to extrapolate from published and readily available information.

Both TrackMaster and RDSS peg low-level claimers around 80 on their respective SR scales and that's about all they have in common. TrackMaster is Beyer-esque in that the value of a length varies with the distance of the race while the RDSS Adjusted SRs are consistent with the TPRs.

RDSS ASR - 1 beaten length equals 1 point at all race distances.

TM SR - the value of a beaten length varies as following:

5.5f = 2.4 points
6f = 2.2 points
6.5f = 2.0 points
7f = 1.8 points
8f = 1.6 points


Caveat - These are only estimates based on what I could coax out of the information. I think they are workable but I can't vouch for their absolute precision. If anyone has something more accurate, please post.