Question on using compiled factors...

[mikesal57]

Hey ALL....


For you homemade builders......

Say you created 6 factors....

The first one is 10 % better than the second one
and
between the second and sixth one its no more than a 2-3 % difference between each.


Would you include the #1 knowing that most likely it will impact the total figure rendering the rest as less significant ...or...toss it and use 2-6

Mike

[Tom]

Why would you toss the best factor?

[mikesal57]

Originally Posted by Tom

Why would you toss the best factor?

Ahhh...someone answered...

This is the thing....

Would that top factor bias all the rest....?

so that factors 2-6 does nothing ...in changing the order..get it?

[mikesal57]

Say that TOP Factor is Bris Prime....

Now everyone has that....

Do you toss that hoping you get an over looked horse?

[mikesal57]

I have a thought....

I will leave out TOP ...compile with rest of factors and then compare to
Bris Prime....

if same....favorite

Different gives me a "value" play

????

[NorCalGreg]

Turn down (if able) Bris Prime to 50%...and experiment with some other factors @ various levels--one of my favorites is EPS as a % of today's purse. It's easier to work with than straight-up EPS.
Or avg of last 3 SR's, %Early compared to FF, etc.

-NCG

[UnifiedTheory]

Ah yes...the key choice of weighting a method to create a near perfect odds line for win contenders and then watching the odds for overlays right up until they're loading into the gate...or weighting a method to pick a decent percentage of winners that are most likely overlays thus avoiding the time/hassle of betting at the last moment and at times still getting poor odds after the final bets.

I love the idea of creating a near perfect odds line for the top 2-4 horses, but I also love the idea of not having to bet just before the race starts.

When I'm creating or refining a Handifast analyst, I basically split the difference between the two weighting ideas to hopefully get the best of both ideas.

For instance, I assign heavy weight to Bris Prime Power but I also assign heavy weight to other factors even though some of the other factors are less powerful then Bris Prime.


For Handifast, I handicap every race using a primary analyst and a secondary analyst. The main analyst makes my picks and the secondary analyst confirms the quality of the picks. I only bet without hesitation if the two analysts agree. If the two analysts disagree but not by too much I make a judgement to pass the race or only bet if at high odds.


My secondary analyst is very basic but effective. Just three factors weighted as follows Bris Prime 250, Fast 200, Mitchell Power 150.

I think it's all about relative weights. It's like if I bake an apple pie and I can't decide whether cinnamon should account for 1% or 3% of the recipe and I decide that 2% one way or another can't make much difference. But is 3% just 2% more than 1% therefore doesn't really matter, or is 3% 200% larger than 1%?! So it's all about assigning the proper weights, relative to each other and relative to the whole.

That's my two cents.

[thaskalos]

Here are some numbers that I came up with for your 6 factors...and their respective handicapping values:

1....29.5%
2....19.5%
3....16.5%
4....14.0%
5....11.5%
6.....9.0 %

So...your standout factor is worth 3.3 times your worst factor...and about twice your two "average" factors. That's about what it should be, IMO. As Tom already stated...there is no way that I would exclude a handicapping factor that I consider the most predictive of them all.

[mikesal57]

So the consensus here is to leave TOP factor in.....

Thxs

[thaskalos]

Originally Posted by mikesal57

Would you include the #1 knowing that most likely it will impact the total figure rendering the rest as less significant ...or...toss it and use 2-6

IMO...the whole point of having a "best handicapping factor"...is to have this factor impact the total figure. And...since this is our "best handicapping factor"...then the rest of our factors can't help but be rendered "less significant". The REAL injustice is to have the LESSER factors impact our total figure alone.

[jay68802]

Originally Posted by thaskalos

Here are some numbers that I came up with for your 6 factors...and their respective handicapping values:

1....29.5%
2....19.5%
3....16.5%
4....14.0%
5....11.5%
6.....9.0 %

So...your standout factor is worth 3.3 times your worst factor...and about twice your two "average" factors. That's about what it should be, IMO. As Tom already stated...there is no way that I would exclude a handicapping factor that I consider the most predictive of them all.

I do the same, but break the races down by the # of E horses and speed points.

1 E horse

2 E horses with at least 9 speed points

3 or more E horses with at least 17 speed points

Funny how much the facctors impact value changes.

[UnifiedTheory]

Hmmm my book-like post clearly didn't do much if anything to answer your question. Apparently I wrote a text book example of ego driven self-indulgence.

On the bright side, the mental stimulation of reading other members posts on this topic helped me figure out what was wrong with the weights in my primary analyst so now I no longer need my secondary analyst.

[mikesal57]

Originally Posted by thaskalos

Here are some numbers that I came up with for your 6 factors...and their respective handicapping values:

1....29.5%
2....19.5%
3....16.5%
4....14.0%
5....11.5%
6.....9.0 %

So...your standout factor is worth 3.3 times your worst factor...and about twice your two "average" factors. That's about what it should be, IMO. As Tom already stated...there is no way that I would exclude a handicapping factor that I consider the most predictive of them all.

Thas...

I'm now gonna throw a twist in the mix now.....


In your chart above.......Suppose #1 #2 #3 are highly correlated with each other.......what does 4-5-6 do to this now....
And if you take out 2 & 3.....Now you have rank #1 2 times more than the next factor....

???

[Mulerider]

Originally Posted by mikesal57

Thas...

I'm now gonna throw a twist in the mix now.....


In your chart above.......Suppose #1 #2 #3 are highly correlated with each other.......what does 4-5-6 do to this now....
And if you take out 2 & 3.....Now you have rank #1 2 times more than the next factor....

???

I wonder about that too. I don't use Bris Prime Power (except to see how my final rankings compare). I do use last out SR, and the Bris last 3 average SR.

I'm convinced those two factors are weighted pretty heavily in Prime Power (just as "late pace" is weighted more heavily in turf races). So, in effect, if I used Prime Power in conjunction I'd be "double-dipping" those two factors, so to speak. Or am I missing something?

[mikesal57]

Originally Posted by Mulerider

I wonder about that too. I don't use Bris Prime Power (except to see how my final rankings compare). I do use last out SR, and the Bris last 3 average SR.

I'm convinced those two factors are weighted pretty heavily in Prime Power (just as "late pace" is weighted more heavily in turf races). So, in effect, if I used Prime Power in conjunction I'd be "double-dipping" those two factors, so to speak. Or am I missing something?

You got it!!!!!

I know from regression analysis, that Bris Prime consists of many factors in ways of final time calculations.....

I'm working with a software that has about 8 different factors that pertain to final figures....Bris Prime out performs all of them ......

After that , there are factors that are related to jky trn class form early late and none of these come closer to 20% to Prime B.....

So you now have the dilemma here that I posted....

You can leave BP in

or

Leave BP out....