Wksc = Form Cycle…

[DanG]

Barbaro’s last three Cramer, Per, Thoro-Graph figures & HTR WO scores…

Before I begin. This is not meant to be a ratings comparison. Just a simple illistration to point out some potential by cataloguing past HTR workout ratings.

Cramer, Per & Wksc highest = best. JBTG…lowest is best. The (-2.5) # being a negative #...BTW: The fastest ever assigned to a Derby winner by this source.)

(Left being most recent.)

Cramer…110…102…104
PER……..109…107…108
TGJB…(-2.5)..1.5...3.75
WKSC….85…..86…..89

The only rating of the four to have a strictly numeric declining trend is the HTR WO rating. Food for thought for those of us applying this data. This is not an isolated example I can assure you. There are powerful subtle patterns within this rating that I’m sure many of you are exploiting.

Best of luck...

[Donnie]

Dan-
Could you share an example of a race recently where the horse did not break down. I am in Henry's camp...the horse never got to run his race. I do not feel you can "rate" a horse that breaks down. As your other post stated by Hillenbrand, it happens about 1 in 700 runnings for each horse. I don't feel comfortable pointing to a number and saying "See! You could tell he was gonna break down!" I know that is not the way that this post is meant. But I seem to get the feeling people are looking for a reason that you shouldn't have been on Barbaro. Interesting species, this human being. We judge them against the clock. Something so cold and mechanical. But yet all they need to do is beat what is thrown against them. And up to this point that is what Barbaro had done. I didn't see anyone with any better horse up to that point. Every year we seem to hear "What a lousy crop of 3 yr olds this is!"
Now had the race been run, different story. Do you mind sharing more on this topic of work scores?? If not, that's cool!

[DanG]

Dan-
Could you share an example of a race recently where the horse did not break down. I am in Henry's camp...the horse never got to run his race. I do not feel you can "rate" a horse that breaks down. As your other post stated by Hillenbrand, it happens about 1 in 700 runnings for each horse. I don't feel comfortable pointing to a number and saying "See! You could tell he was gonna break down!" I know that is not the way that this post is meant. But I seem to get the feeling people are looking for a reason that you shouldn't have been on Barbaro. Interesting species, this human being. We judge them against the clock. Something so cold and mechanical. But yet all they need to do is beat what is thrown against them. And up to this point that is what Barbaro had done. I didn't see anyone with any better horse up to that point. Every year we seem to hear "What a lousy crop of 3 yr olds this is!"
Now had the race been run, different story. Do you mind sharing more on this topic of work scores?? If not, that's cool!

_____________________________________________

Donnie,

Your right…This was NOT meant to be a post that said “see” he was about to be injured. It had nothing to do with it. Just an illustration on a reversal in trend that conflicted with final time ratings.

As far as further examples of this illustration. I don’t have specific examples at my fingertips because it’s not the way I apply the WKSC as a rating. (I was just trying to encourage research in this area without getting into specifics. I thought Barbaro was a logical place to start being so fresh in everyone’s mind)

Now…This brings up something else which will get several people cursing me but here it goes…I use a rather complicated method of establishing a horses “live” threshold within many HTR ratings. WKSC being a paramount component. After this threshold is established allowing for a “maturity adjustment etc” I have simple on / off switches that indicate a horse’s likely hood of firing a competitive effort.

I have spent a tremendous amount of time on this and would rather not get into specifics, but these benchmarks and subsequent trend lines when applied correctly have improved my game immeasurably. I’m not sure my specific method would help anyone to be honest. Many of you could program what I use in a matter of hours and it has taken me months to create what I now use.

Hundred’s and hundreds of lookup tables, a series of 25 checkpoints pacelines are evaluated on etc…etc…

Sorry if the original post was misleading. I was just trying to spark the HTR’s player’s imagination with an alternate approach to WO scores and their predictive value in today’s race.

[Donnie]

Dan-
Never meant for that post to be derogatory in any way. I am glad you did not take it that way. In your follow up, I sat up and took note of your line After this threshold is established allowing for a “maturity adjustment etc”. Tis an angle I think many handicappers ignore....age and or experience.

Concerning the WKScore...I hope Henry chimes in here...he was doing a db that did just as you propose....track the wksc, but he also was tying it into trainer intent. Very new stuff. I am excited to see what Ken is going to reveal at the seminar. He always seems to have something up his sleeve going in.

I'm all for spraking imagination as well. We have some of the most talented and inventive minds on this board. Some of the things some of you people do rival rocket science. It is amazing. Thank God we have good numbers to start with.......

BOL!

[DanG]

Dan-
Never meant for that post to be derogatory in any way. I am glad you did not take it that way. In your follow up, I sat up and took note of your line After this threshold is established allowing for a “maturity adjustment etc”. Tis an angle I think many handicappers ignore....age and or experience.

Concerning the WKScore...I hope Henry chimes in here...he was doing a db that did just as you propose....track the wksc, but he also was tying it into trainer intent. Very new stuff. I am excited to see what Ken is going to reveal at the seminar. He always seems to have something up his sleeve going in.

I'm all for spraking imagination as well. We have some of the most talented and inventive minds on this board. Some of the things some of you people do rival rocket science. It is amazing. Thank God we have good numbers to start with.......

BOL!

_________________________________

Donnie,…

It’s funny you should mention Henry tying in Wksc etc…with trainer intent. That’s what I’ve been tinkering with recently.

Tidbit… (Maybe not the greatest example, but a TRN we are all familiar with.)

HTR WO grade of most recent race. Todd Pletcher…All Burger…

Grade…Sample…Win%...ROI
A……..323………29%.....$1.13
B……..288………24%.....$0.89
C……..514………24%.....$0.92
D……..219………27%.....$0.88
F……….88……….28%.....$1.08
Null…..38……....24%.....$0.79

(Do these things ever format the way you want?...lol)

“Thank God we have good numbers to start with.......”

How true is that…That’s the foundation of everything. Without accurate data underneath our individual approaches we are all flying blind.

BTW: I think I read where Ken is thinking about related last out ratings… (WO, PED etc…) to today’s circumstances. You would think people would cringe that Ken is promoted their individual methods, but it’s really just the opposite. I have e-mailed Ken with a couple theories of mine and he always responds with an idea that winds up improving upon it. I’m sure this will be no exception.

[MVM]

I use almost all of the HTR ratings in different contexts, depending upon what a logical grouping for that rating is.

IMO opinion, the best ratings in HTR are Wk, Ped and Trn (when I say "best" I likely mean this in terms of ROI, as they often differ from conventional wisdom or the ratings and opinions available to the general public).

In the case of Wk, I view and rate it in the context of Circuit, Class, Trainer and Horse. What this means is that I take all of the workout ratings for starters within a grouping defined by the definition listed above, and determine the distribution characteristics for that collection of ratings. Basically, all I do is calculate the Mean, Standard Deviation and Skewness for every Wk rating on a given Circuit (ex. SoCalMajor is SA, Hol and DMr), within a given Class level, for every starter from a particular Trainer, and finally all Wk scores for a particular Horse. I then scale the ratings accordingly, and view each "facet" of the Wk rating through what is similar to an A-F grade curve.

By doing this I know that an 89 Wk at TuP is much stronger than the same score at EmD, or more intuitively that a 90 Wk at the 85 EPR level is a much bigger deal than it would be in a race with a 98 EPR. In spite of the small sample sizes though, I have found the Trainer and Horse contexts to be much more valuable. For a trainer whose horses commonly have Wk scores in the 90s, there is nothing to get excited about with a big rating, there is however a reason for concern when one of their entrants shows up with a rating in the mid 70s. Vice-versa, when a trainer who normally runs horses with lower Wk ratings (lower meaning below avarage for the Circuit and Class) shows up with an entrant sporting a strong Wk score, it is important. The same basic logic applies to a horse. Horses running 1st or 2nd out off a new lifetime or yearly high Wk score show a much higher ROI than other horses sharing the same workout score (ex. horses with a "fresh" 90+ Wk score
who have never raced with an 87 or above prior show a significantly higher ROI than all 90+ horses without the "prior" characteristic).

It is not overly difficult to do the same thing with the Trn ratings. In the Trn ratings, I primarily look at the rating in the context of Circuit. A 293 is much better at EvD than it is at GG, and a 50 is not as bad. I also look for upward movement in the Trn rating based on the last 30 starters.

As for Ped, this rating (as most have probably figured out) is gold. My experience shows me that it is important to look at the different Ped categories side by side. For example, a young horse with a high DS Ped moving to a TR and sporting a similarly high rating is not nearly as good a bet as a horse with a poor DS Ped that also sports a high TS Ped. The problem is that the DS, DR, TS and TR Peds are not scaled equally. You will see few if any 900+ DS ratings, but they are all over the TS and TR categories. Again, look at the entire collection of ratings within a given grouping, and using basic statistical tools determine the appropriate scaling to use.

[DanG]

MVP,...

Brilliant post!

You explained the basis for this premise far more coherently than I ever could.

Very interesting while many of us our taking a similar approach we each have our own particular way of applying it.

[Donnie]

...see what you started again, Dan??? LOL :D

Now we got more stuff we gotta look at!!

(in total agreement....excellent post Mike! Thanks! We still gonna see you and Kat at PRM in June?)

[MVM]

Actually Donnie, I've been in Oz for the last month, returning home on Friday.
If I get home and tell Kathy we are gonna head out to DesMoines...well, I don't want to think about what would happen.

[DanG]

...see what you started again, Dan??? LOL :D

Now we got more stuff we gotta look at!!

(in total agreement....excellent post Mike! Thanks! We still gonna see you and Kat at PRM in June?)

________________________

So true Donnie…

Here I thought I took this rating to its next logical progression and Mike is already out of this galaxy and on his return trip!

I’m surrounded by people much smarter than myself which to be honest…Is how I like it!!!:)

[Donnie]

Mike....Cool!!! We'll miss hooking up with you guys again, if you don't make it, BUT I like the thoughts of diminishing competition! Remember, though...the invite is open....and if you need a place to stay....just holler....but ya better LIKE dogs! ALOT!! Just ask Tommy!!

Dan.....ditto!

For anyone else....anyone brave enough to make the trip to PRM in June for our NTRA tournament automatically qualifies for a free steak dinner at our house....but I gotta know before hand....there is a secret marinade I use that really makes horseplayers all LOOPY. But it sure tastes good!:D

[overdog]

Hi guys and gals;

I LOVE this forum. Almost every post regarding how you all take different approaches to a specific rating jumpstarts my imagination and I'm sure its the same with the rest of you.

Above, MVM said".....The problem is that the DS, DR, TS and TR Peds are not scaled equally. You will see few if any 900+ DS ratings, but they are all over the TS and TR categories. Again, look at the entire collection of ratings within a given grouping, and using basic statistical tools determine the appropriate scaling to use...."

In my day job, I do data mining, then take the results and build mathematical models for clients. In all projects, the data "comes together" or falls into place best if its scaled properly. Leaving data in its original or logical format often causes me severe problems. I did a project for a New Jersey residential real estate Multiple Listing Service." Goal was to predict the selling price of a newly listed house.

In this case the most expensive houses were in the $1.8 million range. I also converted lot sizes to square feet. (Acre is 43,000+) Number of bedrooms could be as low as 1 or 2 with cheap houses. Most data mining techniques crash or give crazy results when variables are orders of magnitude differently scaled. The main method to scale numbers is called normalization.

Fastest way to do it in a spreadsheet is to divide each individual item in a column representing a specific variable(i.e. workout rating or K-rating or ??) by the category maximum. This makes everything score somewhere from 1.00( the best) to 0.00 for the worst. The average score should be somewhere near .50, etc.

Most modelers would then use an adjustment to make all scores fall into the .90 to .10 range. This compresses the data somewhat and has the great benefit of dramatically reducing extreme scores at the very high or very low end.

Here are the two formulae:
First one scales each rating from 1.00 to 0.00(Best to worst) All others will score from .99 to .01

Normalized = (horse score - minimum score) divide by (Maximum score - Minimum score)

(Max-Min is just the range.)

Compressed = (((0.9-0.1) Times (Horse score - Minimum)) divided by
((Maximum score - Minimum))) plus 0.1

STRONGLY recommend the latter!

Hope this helps,

Good luck Barbaro and to you all, too.

Fraser Rawlinson

a.k.a. overdog

[DanG]

Fraser,

I could have used your skill set and grey matter many, many times in my life.

Very well explained. Thanks very much for sharing your expertise. I look forward to using these formulas!

[dehere]

Good news, bad news.

First the good. I was working today and never checked the board until now so I got some work done. Had I happened upon this thread I would have been all over it.

Second good news - Taylor Hicks is probably gonna win. Not that I'm an Idol nut, but I picked him as soon as he walked on the stage with his harmonica and its kinda like picking the derby winner in December. 'Cept i might have made money had I actually been able to bet on him at that point - so maybe not such good news.

Bad news. I have now seen this post and feel I just gotta respond. God, how I love HTR. This is bad news because I tend to be kinda lengthy in my posts.

First, let me say that it absolutely amazes and delights me to see that someone other than me is playing with historic workout score data. I thought I must be way out in left field as no one else seemed interested in this sort of thing.

And MVM (sheesh, i wish my alzheimers would let me remember actual first names here) really brought up something I had not even considered with respect to different workout scores based on track. That is something intriguing to look at.

For anyone interested, there is a workout score thing in the PP that I posted the other day. I will repost it here for ease of looking at it. The workout history shows the history for the trainer and the horse. If you look at the part of the PP which is highlighted in green for Like Now, for example, you can see that the current score of 82 is some eight points higher than the trainer typically works his winning horses and 7 points higher than Like Now has worked when he won in the past. The impact of that has been to move up horses in my opinion.

The thing that kills me now is that if you look at Hemingway's Key you see an 88 score as compared with a 77 when he won in the past. (Incidentally, offspring of Notebook are ITM 64% of the time at Pimlico on a fast track (and 59% in route races on a fast track) another example of how redboarding could have led to a bet on Hemingway).

What is also relevant is looking at Platinum. While his 79 workout score is okay by itself, when you look at his history it seems he generally works faster or better or whatever. Thus, I would downgrade the horse in today's race even if there weren't other reasons to do so.

That is enough for the workout score for right now. Now for a brief comment about Dan's original post. Dan, what I found most interesting about that is that both the Cramer number and the PER rating indicated an off race in the Florida Derby for Barbaro whereas TG had him making a nice forward move. I happen to think that TG nailed that, but would be interested to hear alternative views.

I'll throw together some other comparisons using other horses pretty soon - maybe it would be good to look at some other examples for this sort of stuff.

Anyway, another great thread for HTR.

[DanG]

Henry,

I have a feeling there are people out there doing many things with Ken’s data that would blow all of our minds.

Yes, a historic DB to “re-create” each race within a horses PP’s I have found to be EXTREMMLY rewarding.



That is enough for the workout score for right now. Now for a brief comment about Dan's original post. Dan, what I found most interesting about that is that both the Cramer number and the PER rating indicated an off race in the Florida Derby for Barbaro whereas TG had him making a nice forward move. I happen to think that TG nailed that, but would be interested to hear alternative views.

.

___________________________________________

As far as JBTG having Barbaro’s Florida Derby rated higher than Cramer & the Per I have strong opinions on the Sheet’s ground loss, weight carried and wind adjusted methods that will pull this thread of course, but what the heck…

I personally have minimal input of final time in my current approach as I have said often; I think it has been driven into unprofitably. But as someone who has made figures for many years… (Thoroughbreds and Greyhounds with my own hand times)…I MUCH prefer figures that do not incorporate ground loss, wind and weight. I found when I incorporated ground loss I lost confidence in my application of the data. With EVERY adjustment made to race timing the potential of additional error is introduced.

Let’s use Tampa Bay Downs as an example… (A track I’m very familiar with)

Last year (2005) I had a terrific run at Tampa Bay Downs. I can’t imagine using ratings that give significant extra credit for “lost ground” when 95% of the time it is the preferred path on dirt. Show me a Tampa horse who dueled on the rail at 7f and held well and I’ll show you an extremely fit animal. This animal is discredited using the sheet approach.

In addition…Do they use “live” wind?...I can tell you, TBD is a notoriously windy and the surrounding airports have NOTHING to do with the unique location of the track. If any sheet source is using airport readings for wind in Tampa IT IS WRONG MOST OF THE TIME!

Weight…Now…I’m not one of these guys who says “weight means nothing”…Of course it has meaning. If horse A was made to carry a 1,000 safe on his back it would slow him severely. So by that logic if 1,000 lbs makes a difference, so does 1 lbs. But I have no concrete way to measure what adjustment to make and have MY suspicions of those who do. Supposedly Ragozin scientifically worked out his weight adjustments many years ago and they are on sound footing. I just don’t risk my money on what I don’t understand.

When I apply “speed”…Give me a raw clocking, quality variant and I want my figure maker to stay on top of their track to track adjustments. That’s Cramer and Massa. Then I can objectively say…Ok…Barbaro Fla. Derby...

A)...He was sent hard early from the 10 hole…(Todd Pletcher was quoted as saying “At GP at 9f a wide post is a death sentence)…
B)...He was in the two path on both turns.
C)...He carried about 7 lbs of ‘dead weight (led)…
D)...He had a taxing stretch duel and was headed 3 times and fought back like a grade-1 horse. Etc…

Every time I visit Vegas I check out the sheets that one of my partners uses. He is always amazed when I point out how certain tracks, biases, pace situations could have completely altered the ratings thus producing contradictory patterns.

I'll stick with my current approach, but there are MANY valid approaches to this game. And that is only a good thing. In the end we are all trying to do the same thing weather it’s by the sheets, HTR, data mining etc…Trying to find the “live” horse who will outrun their odds.

[OPM]

There are alot of other ratings in HTR that are sheet like or better, you have to make your own adjustments but it's there. I'm working on something similar but like Dan, I don't believe in ground loss and wt. It's going to take some more time but everything is there you just have to find it.
I would love to discuss this in person but won't be able to make the seminar again this yr. Good luck.

[dehere]

That's cool. Having never gone through the exercise of creating my own figs I cannot speak to a lot of what you say Dan. I won't pursue any speed figure discussion here - that issue has been amptly dealt with many times in other threads.

For me what was interesting about this thread was MVM's use of historic data in a workout analysis and Dan's thoughts about looking at historic workout scores in some sort of pattern analysis in the first post of the thread.

I would like Ken to jump in here at some point regarding the mechanics of the wksc. Is a consideration of historic wksc data somehow redundant? In other words, does the wksc already take historic workout data into consideration so that by plowing through history are we somehow giving double credit to something? I still think that a consideration of trainer data insofar as their workout history remains relevant, but I wonder about using workout scores in a pattern analysis. Any help Ken?

I am also interested in how MVM incorporates class and circuit data in his consideration of historic workout scores. Is this something that you have automated or is it something you do manually or simply by observation when you look at a race? Your thoughts regarding implementation of standard deviation made me knock my forehead - a why didn't i think of that sorta thing. I worked on incorporating that into my access run last night after reading your comment. It should give a bit more context to the historic data that shows up for each trainer. By the way, what is "skewing"?

Anyway, I hope this discussion continues.

[DanG]

I would like Ken to jump in here at some point regarding the mechanics of the wksc. Is a consideration of historic wksc data somehow redundant? In other words, does the wksc already take historic workout data into consideration so that by plowing through history are we somehow giving double credit to something? I still think that a consideration of trainer data insofar as their workout history remains relevant, but I wonder about using workout scores in a pattern analysis. Any help Ken?

I am also interested in how MVM incorporates class and circuit data in his consideration of historic workout scores. Is this something that you have automated or is it something you do manually or simply by observation when you look at a race? Your thoughts regarding implementation of standard deviation made me knock my forehead - a why didn't i think of that sorta thing. I worked on incorporating that into my access run last night after reading your comment. It should give a bit more context to the historic data that shows up for each trainer. By the way, what is "skewing"?

Anyway, I hope this discussion continues.

___________________________________________
Henry,

I knew before I started typing I was going to ramble on about figure making. Like most current or “rehabilitating” figure makers I get very passionate about it…anyway…

I agree completely that MVP’s excellent post has much promise. I can vouch that I’ve found significant correlations among what he discussed.

As far as the algorithms of the WO rating and are we pursuing redundant data conclusions. I can categorically say…absolutely not.

I have too large a current sample size for this to be an aberration. But, in all honesty I may be tweaking things a little differently than some. (But aren’t we all?).

“The law of large numbers” Bill Benter is fond of saying. In reading Ken’s philosophy he subscribes to the same mantra. Ken is adamant about establishing “National Benchmarks” (such as PED 450, WO 85, TRN 350 etc…) that have a strong foundation. Then he encourages the individual user to ‘drill down 2 or 3 levels to local or categorical significants.

I too would love to hear Ken on this subject. I have a very complex maze of queries and formulas to arrive at my conclusions and Ken often simplifies (and improves) my approach by cutting to the chase. Something he is very good at.

I use a combination of…

• Workout score…
• Days back and frequency of workouts…
• Grade…
• Distance…
• Form cycle of animal in question. (I.e.….3rd off layoff, 2TS, and 180+…etc…)

All weighted by a combination of…

• Horse maturity…
• Track profile…
• Individual trainer intent… (When sample size dictates…very important!!!)
• Among other bells and whistles…

Now, I know you’re thinking…”He’s being redundant using many of the items that are already in the WO rating.”…
Yes, this is true. However, I found there is a fascinating inter-play between each factor when weighted properly that is undeniably effective on most circuits.

In fact I went so far as to write Ken and plead with him not to chance a hair in his algorithm in this rating and if he does to please buy time on “American Idol” to let us all know he’s done so. :D (Having said that I LOVE the fact that Ken is always tweaking and improving individual ratings. As we all know, this is not a static game. Like a shark that stops swimming, this game kills those who stand still.

BTW: That PDF report you post continues to blow my mind. What a serious piece of work that is!!!

[hurrikane]

Damn, I love this place!

when I have more time I'll toss in 2cents.

You folks are great!

[DanG]

Damn, I love this place!

when I have more time I'll toss in 2cents.

You folks are great!

__________________________

No excuses Kane,…

Just because you’re putting 14 kids through college is no alibi!:D

BTW: Did your son read the Andy West link I left on another thread? “HTR Beta” I think?...Right up his alley as a fellow bass playing / computer monster.

[km]

good thread

My reply as to whether increasing or declining Wk numbers mean anything? =

i don't know

Obviously the Wk is accumulative and the previous score will include much of the same information as current rating. But there is a big difference between the Wk and speed figures in terms of pure performance prediction.

We know for a fact, that there is a greater impact in performance with a higher Wk rating. In other words, 88's definitely get better results than 85's; and 85's predict more winners than 82's, etc. The same is true of the JKY, TRN, PED. The higher the number, the higher the probability of a "live" or winning effort.

We can't determine that same type of correlation with speed figures. If the horse ran a 99 previously, then ran a 104 last time, the impact on performance is ?. Perhaps the 104 is a peak rating (top) and the horse is almost certain to decline from it. Maybe the 104 was achieved with a softer pace than the 99. So 104's do not at all relate to a better performance next out than 99's. There is no definitive way for us to look at a series of speed figures and predict that peformance is likely to go up as the speed figures go up.

There is far greater complexity to the variables involved in a speed figure (trip and trouble, dist/surf preferences, track condition and weather, class changes, etc etc) so the analysis is much more challenging.

People have written volumes about speed figure sequences and patterns. Nothing much has been tested with the changes to the Wk, PED, TRN, JKY, but a new screen in HTR2 this summer will help our cause.

[MVM]

Just a couple of random thoughts/answers to questions here (my last post from the southern hemisphere):

1) Two rules when it comes to research/data mining/correlation..whatever you want to call it.

Rule #1 - Never use a subjective measurement as your outcome. Finish position, payout, ROI, that is the purely objective data you want to use as your outcome. It may sound like I am stating the obvious, but it is very, very easy to fall into this trap, especially when working with numeric ratings that represent subjective opinions (i.e. speed figures).

Use a WK rating as a predictor for a finish position, and you've got a great little research project slated. Pass go, collect $200, etc. etc.

If you use a Wks as an explanatory variable and a Cramer Fig as a dependant variable (use the workout score to explain the increase/decrease in the Cramer fig), that's a venial sin. Ya, they're both subjective ratings, but it's easy to get fooled and think of the CramerFig as an objective measure, and besides it is an apples-oranges comparison, which is in this case actually a good thing. Say 2 Hail Mary's and let me know if you found something interesting.

If you decide it would be useful or interesting to embark upon a research project that relates speed figures to speed figures, then I would suggest you click on the following link http://en.wikipedia.org/wiki/Self-fulfilling_prophecy. After reading the article, open an Excel spreadsheet and in Cell A1, type the following =A1. Hopefully this wasn't too subtle.

Rule #2 - See Rule #1.


Dehere question "What is skewing?"

This would be so much easier if I had an overhead projector, so I am going to have to do this from a conceptual standpoint.

Obviously you understand Standard Deviation, and realize that the StDev is a representation of the spread within a collection of data. Well, if you think of a distribution of datapoints (ex. all Wk ratings at Pim this year) as a shape, there are a number of items that describe that shape. Standard Deviation is one of those descriptors, and it relates to the internal spread of the values. Skewness relates to the "tail" of the shape (distribution). If my average Wk rating is 80, my highest is 95 (15 points higher than 80) but I've also got a few 60 and lower scores thrown in (20 or more points lower than 80), then I most likely have a negative skew or "tail", as my range of values extends much further to the right (lower) of the mean than it does to the left (higher).

One of the problems with excessive Skewness (as well as another "shape" descriptor, kurtosis) is that it nullifies the usefulness of Standard Deviation as a measurement tool. If anyone actually has an interest in this, perhaps I can elaborate at a later date.

One of the challenges in working with subjective data (like the Wk or Ped) is that it often has a "false" bottom or top value. In the case of the Wk rating, the actual numeric bottom is much lower than the point where it becomes an indeterminate rating (I think KM identified this around 75 or so). With the PED rating, the 999 as well as the lower bound (which I don't recall at this point) are (IMO) best left out of any distributions, as these numbers may actually represent higher or lower values and should be treated as outliers.

Anyway, I've got a plane to catch in about 30 minutes, but I have to admit that I have been really impressed with the quality of the threads lately, some really great, well thought out discussions going on here.

[dehere]

I'm out of town to do research today, but gotta say sumpin before I go. I feel as if I plow around through this stuff hoping to come up with a nugget here and there. Eventually I get a data set that I can work with and then what? I don't know, I just look at the data set and think about how it should be helpful in some way, but I can't quite get it to do what I want.

And then Mike (MVM) and Fraser (Overdog) and DanG comes along with clarity of mind and opens up a whole new world explaining just what needs to be done to actually make that data set actually do some work for you. And all I can say is thanks! Now i have something I can run through my mind all day as I'm stuck in traffic in northern Virginia wondering why the stupid legislature cannot get its act together and do something about this mess.

I happened to qualify for Vegas in 2003 at CNL and, after doing so, thought to myself that I really needed some handicapping tool to help deal with handicapping 8 or so tracks a day. HTR was out there and, after spending a little time wandering about this board and analyzing the cost of this thing, etc., I thought I'd give it a try. I had no idea what I was getting myself into. Thanks Ken - and y'all.

[hurrikane]

Dan,
Actually, I"m sitting in Sedona, AZ on a little sebatical. So my computer time is limited.
this is a great thread, giving me lots to think about.

One question to everyone. My understanding and from what I have found to date in research is that the WK rating is not linear. It seems a 69 is not necessarily worse than a 74. There does seem to be a threashold at which the number is effective.

That the WK number is cummulative may have some value in tracking developement/form cycle but it seems the FC number is pretty effective at that however more evident to the public.

My point is that it seem some of the direction being taken is treating the WK number in a linear manner. I'm just curious if you have found it to work that way for you for I have not.

Thanks for the input.

BTW, Henry, If you can solve the congestion problem in NOVA I'll vote for you ass for President. Not that you would want the job.

[hracingplyr]

I just wante to throw something out to the group here, Ken in particular and that is pattern recognition, for yrs jerry brown's clients over at thorograph were hounding him as to wanting to know how certain trainers horses did after they ran a top and he finally did so although it is a little more involved than just that but u get my drift. was just wondering if there was a way to incorporate in the horses pp's the wk rating, this way if we see a trainer who has consistent wk ratings in the high 80's or 90's and his horses do nothing as far as running ability we as handicappers do not get to excited when we see this horses last work rating a high 80 or 90.

bob

[Pizzaman]

For some horses, higher workout ratings are not indicative of a better effort. For example, Spite the Devil in today's third at Belmont has never won with a high workout rating or after 'A' workouts.

His workout score drops from 91 to 77 today - may be worth a second look.

4 Spite the Devil wt odds fin PER JK TR T30 T% ch lay fin wrk ML=4/1 7/2 VI=32
BEL 05/25/6 8.5F D +a40 $40 -6 97 6 Castellano Ja 332 Jerkens H. Al 323 25 16 116 77 C3b B7b B3h B7b
AQ 01-29-6 8F D +a60 $60 11 8.1 6 89 6 Santos Jose A 271 Jerkens H. Al 341 23 13 8 6 91 B6b A5h
AQ 01-21-6 8.5F D +a99 100 -6 24.5 7 98 8 Morales Pablo 218 Jerkens H. Al 342 18 11 24 7 91 B6b A5h A7b
AQ 12-28-5 8.5F D +a75 $75 -5 17.1 6 102 5 Santos Jose A 262 Jerkens H. Al 329 13 23 28 6 83 A7b B5b C4b
AQ 11-30-5 8.3F W +a65 $65 8 5.3 9 101 3 Gryder Aaron 304 Jerkens H. Al 310 34 24 11 9 76 B4b C3b B7b D5b
AQU 11-19-5 9F D +a99 100 -3 7.7 5 101 4 Hill Channing 255 Jerkens H. Al 330 36 22 28 5 76 C3b B7b D5b B4b
BEL 10-22-5 9F W +a99 250 6 9.2 1 102 5 Castellano Ja 326 Jerkens H. Al 323 32 19 133 1 78 C4b B8h B5h D4b
BEL 06-11-5 9F D +a99 250 -6 25.8 8 102 9 Smith Mike E. 293 Jerkens H. Al 286 24 17 41 8 84 B7b B7b B6h
AQU 05-01-5 9F W +a75 $75 4 2.4 2 98 4 Castellano Ja 323 Jerkens H. Al 298 40 23 29 2 84 B7b C5b
GP 04-02-5 9F D +a99 100 -4 22.8 6 102 5 Maragh Rajiv 229 Jerkens H. Al 346 29 21 161 6 86 B7h B7h B7h A6g
BEL 10-23-4 9F D +a99 250 1 15.0 1 105 1 Castellano Ja 322 Jerkens H. Al 307 26 15 43 1 77 B7b B7h D4b
BEL 09-10-4 7F D +a99 100 4 2.6 4 103 4 Prado Edgar S 368 Jerkens H. Al 313 17 12 26 4 79 B3h C4b C5b
SAR 08-15-4 9F T +a99 100 -3 8.2 12 104 4 Prado Edgar S 378 Jerkens H. Al 319 27 22 21 12 79 C5b A4h
BEL 07-25-4 8.5F D +a99 100 -3 2.9 1 99 5 Prado Edgar S 379 Jerkens H. Al 320 33 27 17 1 76 A4h D4b
BEL 07-08-4 8F D +a42 $42 2 1.3 1 98 4 Prado Edgar S 328 Jerkens H. Al 314 34 12 27 1 76 D4b
BEL 06-11-4 8F D +c85 $49 -1 4.7 4 97 4 Chavez Luis D 241 Jerkens H. Al 326 40 18 6 4 80 B3h B4b B7b
BEL 06-05-4 8.5F D +a58 $58 0 25.8 2 101 3 Fernandez Vic 152 Jerkens H. Al 342 40 18 22 2 80 B3h B4b B7b
BEL 05-14-4 8.5F D +a48 $48 8 15.1 4 102 1 Fernandez Vic 139 Jerkens H. Al 332 40 33 111 4 80 B7b B6b B7b
AQ 01-24-4 9F D +a46 $46 2 2.8 6 104 2 Cotto_ Jr. Pe 174 Jerkens H. Al 376 28 18 29 6 82 B7b B6b B3b
AQ 12-26-3 9F D +a46 $46 -7 5.0 2 105 1 Cotto_ Jr. Pe 170 Jerkens H. Al 377 11 9 29 2 79 B7b C5b
AQU 11-27-3 8F D +a52 $52 0 8.2 5 105 4 Chavez Luis D 292 Jerkens H. Al 419 25 8 40 5 79 D4b B6b C5b
BEL 10-18-3 9F D +a99 250 2 23.1 4 101 8 Chavez Jorge 283 Jerkens H. Al 397 36 19 20 4 86 B7b B5b
BEL 09-28-3 8F W +a52 $52 -4 4.4 5 101 5 Velazquez Joh 364 Jerkens H. Al 385 34 12 4 23 5 86 B5b A5b
BEL 09-05-3 7F D +a99 100 -5 6.7 3 101 5 Chavez Jorge 284 Jerkens H. Al 392 34 12 1 16 3 85 A5b B3h B7b B6b
SAR 08-20-3 9F D a99 150 -3 5.5 3 101 2 Chavez Jorge 293 Jerkens H. Al 403 25 20 25 3 85 B7b B6b B5b
FL 07-26-3 8.5F D a99 125 1 5.7 2 103 1 Espinoza Jose 133 Jerkens H. Al 389 32 25 28 2 85 B5b B7b
BEL 06-28-3 7F D a99 100 2 4.2 6 103 2 Migliore Rich 290 Jerkens H. Al 377 34 18 35 6 77 B6b
BEL 05-24-3 9F W a99 200 5 6.9 4 103 1 Castellano Ja 308 Jerkens H. Al 374 38 32 21 4 73 B7b B7b
AQU 05-03-3 8F D a99 150 -7 16.9 1 102 3 Chavez Luis 259 Jerkens H. Al 403 33 39 2 21 1 70 B7b D3b F5b
AQU 04-12-3 9F W a99 750 7 57.8 8 101 4 Luzzi Michael 235 Jerkens H. Al 336 23 22 27 8 68 C3b F5b C4b
AQ 03-16-3 8.3F D a99 200 -2 5.0 3 96 9 Luzzi Michael 219 Jerkens H. Al 331 21 14 29 3 76 C3b F5b
GP 02-15-3 8.3F D a38 $38 0 1.7 2 101 1 Bailey Jerry 406 Jerkens H. Al 337 24 29 28 2 84 B7b B6b B6b
GP 01-18-3 8.5F D a99 100 -1 14.9 10 101 2 Bailey Jerry 398 Jerkens H. Al 334 13 23 71 10 79 B7h C5b
AQU 11-08-2 8F D a46 $46 4 0.6 1 102 1 Castellano Ja 297 Jerkens H. Al 325 22 23 20 1 68 D4b
BEL 10-19-2 8F D a99 100 -2 8.1 2 95 2 Castellano Ja 293 Jerkens H. Al 310 31 13 20 2 70 C5h F4b
BEL 09-29-2 7F D a75 $75 1 12.4 2 93 5 Castellano Ja 282 Jerkens H. Al 309 24 13 22 2 70 C5h F4b D4b
BEL 09-07-2 7F D a47 $47 0 2.3 4 95 1 Carrero Victo 236 Jerkens H. Al 326 21 24 24 4 69 D4b D5b D3b
SAR 08-14-2 6.5F D a99 150 -2 10.5 3 95 3 Carrero Victo 251 Jerkens H. Al 322 22 9 20 3 72 D3b B4b C4b B6h
SAR 07-25-2 6F D a99 150 5 39.8 3 86 9 Carrero Victo 296 Jerkens H. Al 313 36 22 29 3 74 C4b B6h C3b
BEL 06-26-2 5F D msw $41 0 1.8 1 Carrero Victo 239 Jerkens H. Al 304 29 21 0 1 77 B3g B6b C4b B4h

[DanG]

This is the tough part…

You are involved in a train of thought and you want to participate and life / racing gets in the way.

Belmont has one of their “rare” significant carry-overs today and I have to burn the “morning-oil” to paraphrase the legendary Harvey Pack.

A few timely and short thoughts…

Pizzaman…How do you get your posts to format so perfectly?...Nice piece of work, thank you. But, couldn’t you have used one of the races involved in the pic-6…lol ;)

How timely is this…

I am NOT saying this to brag in anyway, just thought it was incredibly ironic considering this discussion. We took down the pic-6 at Hollywood yesterday along with several consolations.

West coast racing is two of my partner’s expertise and my main “contribution”… (Or, should I say Ken’s) was the key to the sequence.(Other than beating an overbet Mullins dropper in the 6th.) “Ellwood & Jake” in the 5th at $24--…3 guesses as to what put me on this FTS and the first two don’t count. 90+ WO!!! Not that we keyed this horse, but I was the only one of the four of us to lobby strongly for him and not by coincidence I’m the lone HTR user.:)

Thank you Ken and all of you for a fascinating discussion. Now it’s back to Belmont!!!

[Snagaltooth]

This thread got me to thinking about trainer intent. I’m not sure I know how to measure this in a numerical sense but I do know it is important. That’s only one of several reasons I use HTR.

In an attempt to be brief (never my long suit) just let me say that in my twenty years of enjoyment in this sport, I know that the trainer can make or break a horse and my pocket book. I remember a day when even a 30 day layoff meant a horse had to “have” a race before being a contender for a win. A route horse would be entered in a 6f race almost as if today’s race were a workout. Today, with feed, meds, and the knowledge level where it is, this is not true any more. We, as players, must now look to the workout cycle as a measure of intent or to determine current form. I concentrate my plays on the lower level races and feel the stakes level and FTS horse should have its own considerations for training cycles. In other words, one size may not fit all equations. A case or two to make my point:

1. Is a slow work time a result of the horses form or trainer intent? The number we see for a workout is just a number and says nothing about what the trainer is wanting to do with the horse on that day. Teaching vs form training can result in two different numbers.

2. Does the use of a handler/workout rider vs. the use of the jockey intended for the next race change the workout intent and number?


Now, along comes Ken with “Nothing much has been tested with the changes to the Wk, PED, TRN, JKY, but a new screen in HTR2 this summer will help our cause.” Maybe I should just shut up and wait…….lol.

[Rick]

DanG,

The trick for getting things to line up is to use the [ code ] (without the spaces) at the beginning of what you are trying to line up and [ /code ] (again without the spaces) at the end of it all.

It is best to use NotePad or something like it in a font like Courier New. Probably best to use spaces but tabs can work, it just may take a few test.

vBulletin drops all extra spaces in normal posting.

Another way is to put periods in between. That can take much trial and error.

It might help to change the font to Courier New when doing the period trick.

Whichever you use it helps to set it all up in NotePad first. Then copy and paste.

You can do the copy then click on the CODE icon and paste between the two codes.

You can find more on using codes by clicking on the FAQ, Reading and Posting Messages, Are there any special codes/tags I can use to markup my posts?

Then click on the Click Here.

Or just go to http://www.homebased2.com/forums/misc.php?do=bbcode

[OPM]

Hurrikane:
I love Sedona, are you going to Yavapai, it about 30 miles away in Prescott.
Another thought on this thread, you can have all the data you want but whenyou are looking at so much data, you tend to miss obvious things that any RF reader would pick up on.
Look at the ROI of Sal Gonzalez at SA this past meet, he has an avg trainer rating(maybe below avg) but had some real bombs, starting with Proud Tower Too in the Malibu. I would really fall into the trap of looking at too much data.
Yours question for every horse should be: is he in form? If he is in form, will his best win today's race? Now this opens up a whole new can of worms that tries to answer the above 2 questions but if you concentrate on horses that are in form(WK rating), your ROI will go up. That's what the sheet guys do, finds horses with positive patterns who are close to the best #'s in the race. But these 2 questions can lead to a lot of data mining but would you bet a horse that is in form who switches to a jockey with a 100 rating from a jock with a 300 rating?

Another problem with final time, pace #'s, etc is the distance to distance and track to track comparison's are about #'s, regardless of what anyone says. That's why meets like SAR, Kee are very tough to handicap. Just my 2 cents.

[Donnie]

Pizzaman-
Nice form......I really like it.
I don't think the #4 will be facing a pace of 117 today, so my take it is hard to believe he will have an impact in here. He appears to be the type of horse who picks up the pieces after a pace battle up front. That was the pace numbers he faced in his last 2 wins. I project the pace to be a mid-90. He doesn't do well in those races......But then again I could be wrong!
BOL...thanks for sharing!

[Jetcity]

4 workouts last 16 days. With a WO yesterday. I don't think I have ever seen this pattern b4. Will watch but not play


Jet

[DanG]

DanG,

The trick for getting things to line up is to use the [ code ] (without the spaces) at the beginning of what you are trying to line up and [ /code ] (again without the spaces) at the end of it all.

It is best to use NotePad or something like it in a font like Courier New. Probably best to use spaces but tabs can work, it just may take a few test.

vBulletin drops all extra spaces in normal posting.

Another way is to put periods in between. That can take much trial and error.

It might help to change the font to Courier New when doing the period trick.

Whichever you use it helps to set it all up in NotePad first. Then copy and paste.

You can do the copy then click on the CODE icon and paste between the two codes.

You can find more on using codes by clicking on the FAQ, Reading and Posting Messages, Are there any special codes/tags I can use to markup my posts?

Then click on the Click Here.

Or just go to http://www.homebased2.com/forums/misc.php?do=bbcode
_______________________________________

Thanks Rick...

I'm not making any promises in my next post, but I'm going to work at it.

BTW:I meant to say earlier we got a piece of the pic-6, not “take it down”…Wishful thinking!:D

[Donnie]

Gary-
If you don't mind me asking...what is the number just to the right of the Surface?
Thanks!

[dehere]

Kane - lucky you, Sedona is one my absolute favorite places on earth - dinner at the Enchantment hotel or whatever it is called is so nice.

Dan - congrats man. we can all blab about this game but you won where is really counts. way to go!

[Pizzaman]

Donnie

The value to the right of the race surface ('D'=dirt, "W'=wet) is the race class, where the '+' = older horses, the 'a' = allowance and 'm' = maiden and the number is the purse value or claiming price.

The output was created by software I wrote to get lifetime info from previous years HX4 files. The software was straightforward to develop, but the real effort is keeping this year's HX4 files current as any missing races will break the chain. I have had several requests, so I plan to put software in a form that I can share with those that want to try it out with their own HX4 database.

[tomcat]

Kane...have you taken the jeep tour yet? There is a Mexican restaurant down around the corner on the right among the trees that we really liked!

[Donnie]

Are they having hurricanes in Sedona, 'Kane? Hope you're enjoying some downtime while you are there....your busy season is about to start!

Gary...thanks! I for one would be interested in your "final" product. I put a report together that is very similar, only it is geared towards trainers. I like to see how the trainer has been doing with their horses. Only bad part is when you have one of these "super" trainers....all of their horses span quite a few pages...but very interesting to see that some trainers have only one or two productive horses in their barns! 20 horses, only 2 have won over the last 4 months!!

[hurrikane]

Nice report PM.

did the jeep tour yesterday. great adventure. Some horseback riding through the canyon coming up. Took Sue to the GC, she had never seen it before. Stunning. Moves me everytime I see it up close.
Hope to do Yavapai on Sat.

Tom, was the Mexican place Oaxaca? There are so many restuarants here you could be here a month and not get them all.

I have been tracking patterns the way Pizza is, by horse with trainer. haven't found anything earth shattering but still digging.

Nice job Dan. I had Ellwood as well, 90+ and Mitchell running at a 35% hit rate a Hol and PVal on top. Hard to pass on that one.
Made the win but unfortunately I missed my single in the p6. Needed your other 2 guys input I suppose.

Good luck everyone!

[hurrikane]

Here are some questions scurring around in my head. Maybe Ken can shine some light on the rating questions and the other data tinkerers can help me think this through.

The TRN, JKY ratings are scaled by trk, dist, level of comp, etc. Isn't building what might be considered 'Pars' for different tracks a bit misleading as that relationship is already considered in the rating?

the WK rating is blind to trk, dist, etc and comprises a score that says horses at this cls, lvl, dist, wk schedule show form on this scale 50-100(or whatever) and significance in the rating occurs above the 80pt mark with increased sig as the number rises.

I have found that a particular horse will win with 70s and 90s without any significant relationship to the number. I haven't found that with trainers but the jury is still out on the significance of the WK number to the trn intent.

How do you recitify in your mind that these relationships are significant. Any specific data mining efforts return a significant confidence interval to actually put it in play?

Any insight you folks have would be great.

See, I'm here on vacation and you got my head stuck on statistics and data mining. thanks

[DanG]

The TRN, JKY ratings are scaled by trk, dist, level of comp, etc. Isn't building what might be considered 'Pars' for different tracks a bit misleading as that relationship is already considered in the rating?


______________________________

Kane,

As far as I know the Trn / Jky rating are not scaled to reflect Dist / Trk & Level etc…


They include the 365 day weighted WPS formula.
TJ record
Record with animal in question.

[hurrikane]

hopefully Ken will jump in on this and fill in the blanks. I was under a differnet impression but that would not surprise me

[mr. posh]

Dehere, I'm getting to this thread late and I'm curious...The screen you showed on your attachmednt is it a custom? How can I duplcate? I also saw that it was page 49. Can you elaborate?
Thanks for your help.

[dehere]

Mr. Posh,

Its custom and really not very easy to duplicate because I'm so disorganized. I would be happy to share some parts of it but it really includes too much stuff to explain it easily. The notation to page 49 is simply that it was page 49 of the day's pp report.

Incidentally, in reference to this thread, others have probably already come up with this nugget but I was rather surprised with a Robot run using female FTS horses with a WKSC of 85+ and HTR1. A nice little ROI of $1.41, place ROI of $1.25 and show ROI of $1.01. The Robot is attached. This was an all-burger for 2006. Actually, I was rather pleased to see the ROI for all females with a lay-off of at least 30 days.