I probably said something about being more partial to median than average. Average can easily be skewed by a few outliers. — RanchWest

Boy, do I agree with that!

I stumbled on a Thanksgiving 2012(?) episode of Dave's "How to Improve your Game", yesterday. It was 90 minutes of handicapping wisdom with only an occasional DS commercial (you know how he is). One of the concepts that wowed me was the 90% Confidence Interval for speed figures (or any metric). Are you familiar with it or use it? Is it still viable 10 years later? If it is I will try to validate it on my program later today. No matter how good an idea seems on the surface I ALWAYS have to test it myself. — Biniak

I'm just finishing the 2022 HorseStreet Par Times.

The 90% CI is a critical part of the work - and especially the improvements.

The biggest improvement THIS YEAR is the class levels and at the very center is the 90% CI.

Since so many people like making their own figs, I'm thinking about building a small seminar on how I build those class levels. Of course, the secret to GREAT par times is having GREAT class levels.

No.
You must turn it into at least 2 (i.e. 0 & 1) systems.

Another alternative is...
Because System 0 must be the first one alphabetically, change the order of the system/object name.
or...
Find some other system/object - even one that does almost nothing or has no use - and slip it in front.

Then...
You can push all the other systems down.

THUS...
What USE TO BE system 0 is now system 1 and you have complete control over it.

You won't be sorry, I forgot just how many pages of data it contains, but if I'm not mistaken there is over 150 pages. — Jim Parker

Actually, there are over 2,400 pages!

Dave, in the materials, the tables can get them with all the ranked pace factors. — Bill T

Forgive me but I don't know what that means.

Click "Change."

It is really very simple but needs more functionality.
Ironically, in the decade since I built this, nobody has ever asked how to use it so I didn't expand it.

• Think of "Categories" as "Groups" or "Lists."
• By default, these Categories are by Factor Type.
• But you can edit the Categories any way you want.
• Example: You could create Categories according to Type of Race or Filter Name.
• Factors can have multiple Categories, so it an exist in multiple Categories.

You can Save & Load the current Categories with the appropriate buttons.


"Run Style" is Category #5.

Recommendation
In our software we actually do not try to show the TOTE ODDS.
We want to show the EXACT number, such as 2.52 or 2.99.

It works far better for predicting low odds horses.

Also, we like to show POOL PERCENTS.

I knew one pro player who did.
Claimed that it had great value.

He also tracked what he called "cause scratches."
Those were scratches where the horse was entered in another race within the next 6 days.

He said those were far more valuable.

FWIW, in our software we use these factors to predict the odds.

I actually did nothing but sweep the file for viruses and host it on our site.

Steven did all the work.

This is really good work!

Do you recall how Huey Mahl did this thing he called The Length Variant? This connects me to the next idea, and all of it connects back to YOUR WORK.

When I was doing work for some AUS a couple of years ago, I found some of their ideas quite intriguing. One was the idea of expressing "How They Ran" in terms of Odds & Beaten Lengths.

Thus, if you look at each odds range, and come up with the average lengths behind, one could say that a horse who ran (say) 2.5 lengths behind ran like he was (say) 5/2.

Your work is similar but expressed as a win percentage (I assume).

This is a really goods idea.

Is this the average of the best three (3) in the race? Seems like I heard this in a workshop but I don't remember it exactly. — Steven

It is how ever you choose to look at it, but better would be:

Best-of-Last-2 in sprints
Best-2-of-Last-3 in Routes.

Say Oaklawn's par time for 6 Furlongs Claiming $10k for Filly/mare statebreds is 1:10 flat and your adjustments (example, not the REAL #ers) are 3.5 for f/m and 0.5 for statebreds

Do I add the adjustments to the Par Time or the Actual Time I am computing to get a figure for?
And is it better to use class specific times or the $10k times? — Conley

If you want to know what the PAR FOR THE RACE would be, then:

1. Start with the CLASS PAR for the level (i.e. the par number like 103.4 or whatever)
2. Subtract the Female adj.
3. Subtract the statebred adjustment.
(Don't forget the N2L adjustment if applicable)
This number becomes PAR NUMBER for the race.

However, as I explained in Percentages & Probabilities, a more powerful approach is to look at the mix of horses in the race and build a STRENGTH RATING for the actual race and use that.

This way, the level of the race is determined by the actual HORSES rather than the WRITTEN CONDITIONS of the race.

I'm not sure I'm understanding "subsets". Are we looking at races where the best early speed number is 7-8, and then races where best early speed number is 5-6, and then races where best early speed number is 3-4 and we're doing this to see if 7-8 races produce more winnners (or more money) than 5-6 races? — Steven

Yes.
I'd expect that a 5-6 horse would perform better/differently in a field where there were no 7-8s.

This is overly simplistic - and subsets cut down your sample sizes - demanding far more races. Like a full magnitude more.

But the info you gather is huge, as are the ensuing improvements.

Gee, aren't you the greatest poster in the history of PaceAdv?

LOL

Welcome over here, old friend.

Can you break it down to where the top horse is 5 or 6 ? Then break that down to those with 5 or 6 which also have a 4 point edge? — William Zayonce

Yes, and please have this on my desk by tomorrow morning.


Geez, Bill. Maybe clip a couple of c-notes to this request. LOL

PS: Exciting, isn't it?

So that means a top horse "8" and a top horse "3" are all included.

And THAT means you can easily break out the 3s from the 8s and all the others to see if there are SUBSETS that are more powerful.

You should have a good time with that.

So, if you had a race where the horses were:

• 8
• 5
• 3
• 3
• 1
• 1
• 0

You'd play the 8?