PARS plus stdev

[stu]

Y'all,

After reading Super Crunchers, I was rethinking the use of speed figure pars in handicapping. One of the concepts in the book was to remind that the average is more useful when the standard deviation is known.

Many handicappers know the Beyer Speed Figure pars for the various Breeders' Cup races. I want to evaluate the races considering the standard deviating of those pars. The rationale is that 95% of the time the winning speed figure will be between the par and plus/minus two times the standard deviation. The goal is to throw out runners that appear in capable of throwing a number better than the bottom threshold.

In the 141 previous BC events, the winners' times fell within the range all but seven times (approximately 5%). Two of which were too low and the rest were faster than the average plus two standard deviations. The laggards were Gilded Time and Six Perfections. The swift ones were Artax, Fantastic Light, Inside Information, Miesque, and Tempera.

For the purpose of this post, we will assume that no runner will throw a lifetime best in the Breeders' Cup. We will eliminate any North American based runner that has never produced a time faster than the bottom end of the range for the respective race. (Theoretically, this will eliminate all of the losers and 2.5% of the winners.) We will also not eliminate any non-N.A. runners.

BC Juv Fillies
par BSF 95.0
stdev 5.5
7 of the 16 pre-entries remain

BC Juvenile
par BSF 99.5
stdev 4.5
8 of the 16 pre-entries remain

BC F+M Turf
par BSF 108.0
stdev 2.3
4 of the 13 pre-entries remain

BC Sprint
par BSF 114.2
stdev 3.9
13 of the 15 pre-entries remain

BC Mile
par BSF 111.3
stdev 3.0
8 of the 18 pre-entries remain

BC Distaff
par BSF 108.1
stdev 5.3
12 of the 15 pre-entries remain

BC Turf
par BSF 112.1
stdev 2.4
7 of the 12 pre-entries remain

BC Classic
par BSF 117.4
stdev 3.3
5 of the 9 pre-entries remain

BTW, the other three races are ignored because there is no statistical data as this is their first year.

stu

[stu]

BC Juv Fillies
par BSF 95.0
stdev 5.5
7 of the 16 pre-entries remain


Remaing are:

Backseat Rhythm
Grace Anatomy
Indian Blessing
Irish Smoke
Phantom Income
Proud Spell
Tasha's Miracle

[stu]

BC Juvenile
par BSF 99.5
stdev 4.5
8 of the 16 pre-entries remain


Remaining are:

Globalization
Kodiak Kowboy
Majestic Warrior
Pyro
Slew's Tizow
Tale of Ekati
War Pass
Z Humor

[stu]

BC F+M Turf
par BSF 108.0
stdev 2.3
4 of the 13 pre-entries remain


Remaining are:

Nashoba's Key
Simply Perfect
Timarwa
Wait a While

[stu]

BC Sprint
par BSF 114.2
stdev 3.9
13 of the 15 pre-entries remain


Remaining are:

Attila's Storm
Benny the Bull
Bordonaro
Commentator
Greg's Gold
Idiot Proof
In Summation
Kelly's Landing
Mach Ride
Midnight Lute
Park Avenue Ball
Smokey Stover
Talent Search

[stu]

BC Mile
par BSF 111.3
stdev 3.0
8 of the 18 pre-entries remain



Remaining are:

Excellent Art
George Washington
Host
Jeremy
My Typhoon
Rabatash
Remarkable News
Silent Name

[stu]

BC Distaff
par BSF 108.1
stdev 5.3
12 of the 15 pre-entries remain


Remaining are:

Balance
Bear Now
Ginger Punch
Hystericalady
Indian Vale
Lady Joanne
Lear's Princess
My Typhoon
Nashoba's Key
Prop Me Up
Teammate
Unbridled Belle

[stu]

BC Turf
par BSF 112.1
stdev 2.4
7 of the 12 pre-entries remain


Remaining are:

Better Talk Now
Champs Elysees
Dylan Thomas
English Channel
Red Rocks
Shamdinan
Sunriver

[stu]

BC Classic
par BSF 117.4
stdev 3.3
5 of the 9 pre-entries remain


Remaining are:

Any Given Saturday
Curlin
George Washington
Lawyer Ron
Street Sense

[Tom]

I would expect some will run new tops on BC day. It would be interesting to repeat the study with pre race numbers, if one had access to all the PPs.
I could do it for the last 6 or 7 years, using Cramer Figs......good project for this weekend.

[stu]

Remaing are:

Backseat Rhythm
Grace Anatomy
Indian Blessing
Irish Smoke
Phantom Income
Proud Spell
Tasha's Miracle

Depalma13 suggested in another thread that one can translate Racing Post Ratings (RPR) to Beyer Speed Figures (BSF) with the following approximation:

BSF ~= RPR -14

With that approximation, Annie Skates and Joffe's Run are rightfully excluded and we still have six remaining.

To Tom's assertion, this would be the race that I would be most worried about a new top.

[stu]

Remaining are:

Nashoba's Key
Simply Perfect
Timarwa
Wait a While

When adjusting the RPR to BSF, all of the following Euros get eliminated:
All My Loving
Argentina
Danzon
Lahudood
Passage of Time
Timarwa

That trims the contenders to 3 (Nashoba's Key, Simply Perfect, and Wait a While).

[stu]

Remaining are:

Excellent Art
George Washington
Host
Jeremy
My Typhoon
Rabatash
Remarkable News
Silent Name

Pacing Post Rankings knock Rabatash out of consideration thus leaving seven

[stu]

Remaining are:

Better Talk Now
Champs Elysees
Dylan Thomas
English Channel
Red Rocks
Shamdinan
Sunriver

The RPR knock out Champs Elysees and Shamdinan thus leaving us five remaining in the BC Turf.

[stu]

Remaining are:

Any Given Saturday
Curlin
George Washington
Lawyer Ron
Street Sense

George Washington's fig in last year's 2000 Guineas keeps him in the mix.

[rastajenk]

You could put 'em all in a Pick Six for only $163,800. :faint:

The Sprint and the Distaff are kinda dealbreakers, I suppose, although that's where the bombs could come from to make it worthwhile.

[WildGeneral]

Intriguing thoughts here...

Questions:

1) From where is the 'par value' for each race derived? Is it the mean of all entries or based on previous BC results for the same race?

2) If making the leap of faith that this 'system' is fool-proof, it is only predictive of winners, right? In other words, the horses eliminated based on these data can still finish in the money...(or no?).

3) Per another comment in this thread, has anybody tried this theory (based on par/st dev) for a retrospective results analysis? For example, in 2006 BC results, how did the 'pre-eliminated' horses fare?

Thanks to the OP for sharing...

[garyoz]

[QUOTE=WildGeneral]Intriguing thoughts here...

For example, in 2006 BC results, how did the 'pre-eliminated' horses fare?

QUOTE]

You should probably do the study for 2006 based upon data up to 2005. Also you are likely to have new tops for the 2 year old races based upon physical maturity. I would also think that you'll have other new tops--depending on the figure adjustments for MTH on BC Day. Most figs use a projection method anyway (incl. Beyers) and will probably adjust them down or up toward the averages after the race.

Also, front runners can have the escalator effect due to bias at MTH and get very brave. I'd expect to see some better than "average" performances for horses on the front end.

Personally I don't care for averages (or other measures of central tendency) in that it misses form cycle forward moves and new tops. I think the most misleading data in racing are moving averages (but that's off topic)

But still a very interesting post--good work. It will be very interesting to compare the numbers after the race to the pre-race Mean + s.d.

[stu]

[QUOTE=WildGeneral]Intriguing thoughts here...

Questions:

1) From where is the 'par value' for each race derived? Is it the mean of all entries or based on previous BC results for the same race?
QUOTE]

I believe that I calculated it the way the industry calculates it in other publications.

The par is the mean speed figure for the winner of previous BC races in which speed figures were published.

stu

[stu]

Intriguing thoughts here...

Questions:
2) If making the leap of faith that this 'system' is fool-proof, it is only predictive of winners, right? In other words, the horses eliminated based on these data can still finish in the money...(or no?).


The theory that I am working on doesn't address anything but the winner. Even then the 'ave + 2 stdev' elimination method by construction should miss 5% of the winners speed figure -- equally 2.5% too low and 2.5% too high. So using the bottom threshold as a cut-off and assuming no new tops, you will identify correctly the pool of potential winners 39 out of 40 times.

Without having the mean or standard deviation for the speed figures of the place horse, there is no profile to eliminate from the second spot on your ticket. Respectively iterate to last place.

Without proof, on instinct alone I would suppose that the average for the place speed would be lower than that of the winner. Additionally, I would wild ass guess that the standard deviation for second would be larger than that of the winner.

I don't necessarily believe in my theory but am putting it through the scientific method to test the hypothesis with this year's rendition of the BC.

stu

[stu]

Intriguing thoughts here...

For example, in 2006 BC results, how did the 'pre-eliminated' horses fare?



You should probably do the study for 2006 based upon data up to 2005.

I don't have last year's form available to check. I however have recalcualted the range excluding the 2006 results in case some else has pack-ratted their copy.

pre-2006 ranges

Classic 110.7 - 124.3
Turf 107.2 - 117.2
Juvy 90.8 - 107.2
FM Turf 103.1 - 112.9
Sprint 106.2 - 122.0
Mile 105.4 - 117.4
JF 84.3 - 106.4
Distaff 98.4 - 118.7

[stu]

Remaing are:

Backseat Rhythm
Grace Anatomy
Indian Blessing
Irish Smoke
Phantom Income
Proud Spell
Tasha's Miracle

The quartet of the JF Superfecta all came from the remaing horses.

[stu]

Remaining are:

Globalization
Kodiak Kowboy
Majestic Warrior
Pyro
Slew's Tizow
Tale of Ekati
War Pass
Z Humor

The quartet of the Juvenile superfecta were all from the remaing group.

[stu]

When adjusting the RPR to BSF, all of the following Euros get eliminated:
All My Loving
Argentina
Danzon
Lahudood
Passage of Time
Timarwa

That trims the contenders to 3 (Nashoba's Key, Simply Perfect, and Wait a While).

The elimination method failed here. Cost me a little since my wager was on Simply Perfect.

[stu]

Remaining are:

Attila's Storm
Benny the Bull
Bordonaro
Commentator
Greg's Gold
Idiot Proof
In Summation
Kelly's Landing
Mach Ride
Midnight Lute
Park Avenue Ball
Smokey Stover
Talent Search

Still not sure if the theory is a good one but the eliminated horse, Forefathers, finished last.

[stu]

Remaining are:

Excellent Art
George Washington
Host
Jeremy
My Typhoon
Rabatash
Remarkable News
Silent Name

The cutoff would have been 105 and change. I accidentally left Kip Deville off the list considering his 106 in thee Makers Mark.

Method gets the exacta

[stu]

Remaining are:

Balance
Bear Now
Ginger Punch
Hystericalady
Indian Vale
Lady Joanne
Lear's Princess
My Typhoon
Nashoba's Key
Prop Me Up
Teammate
Unbridled Belle

Ginger Punch - Hystericalady gives the elimination method another exacta

[stu]

The RPR knock out Champs Elysees and Shamdinan thus leaving us five remaining in the BC Turf.

English Channel winner

[stu]

Remaining are:

Any Given Saturday
Curlin
George Washington
Lawyer Ron
Street Sense

After 8 races, the average +/- 2 stdev elimination method left:

3 superfectas
2 exactas
2 winners
1 huge flop of a race

I will definitely use the method to evaluate contenders next year.