PDA

View Full Version : Extremely rare occurrence could provide “THE” answer


rmania
01-18-2005, 09:53 AM
“THE” answer is to the question: How fast is a horse going to run today?

Looking at the past performance of any horse you’ll notice that few ever run back-to-back equal times. Horses usually improve or regress from one race to the next. What causes this? Could the effort displayed today be a DIRECT result of the horse’s previous race?

The extremely rare occurrence happened on Saturday, Jan 15th at Santa Anita. It involved two horses, which in their previous race ran neck-and-neck from start-to-finish. This in itself was rare. It became extremely rare when the two horses came back to face each other the next time out.
On that day I posted the following in the Selections forum:

“Race 9
I’m not predicting a winner. Rather, I’m predicting that #3 Attilas Storm and #5 Only in Reno will finish within @ 1 length of each other. Whether first and second, last and next to last, or anywhere in between.

Sounds like a silly prediction but it could be significant if it holds true.

Will explain later.....”

My prediction basically came true. The official distance separating the two at the wire was 2 lengths however the #3 horse was involved in an altercation (leading to a DQ) which probably cost him at least 1 length.

Bottom line... after running IDENTICAL races in their previous effort, both horses displayed the same amount of improvement (or regression). And of course this has always been the case in the past with different horses / different races. Why else would I have made an issue of it.

Now, think about this.... Let’s imagine these two horses going separate directions after their head-to-head duel. Would each display the same amount of improvement/regression? Probably... So by noting the performance of the first to run, you could basically determine (in advance) how fast the second one would run. And, in doing so, you’ve answered “THE” question.

hurrikane
01-18-2005, 10:23 AM
That is an interesting idea r-

but as you say this is a rare occurance.

how would you use this for any horse that did'nt race neck and neck? and how do you rate the horses 2 races back?
and does the 3 still fall into the same place because he would have been there but for the bad racing luck. and how do you quantify that?

it is an interesting perspective. now how to apply it in real life.
I'm guessing you've thought about this? :D

cj
01-18-2005, 10:40 AM
I like to look for what I call a bounce point. I look at the horse's lifetime PPs, then see what is the highest number the horse has ever improved upon next time out. You can find lots of horses bet off of big numbers who are very likely to regress next time out.

Here is an example, most distant to most recent:

65-68-54-81-50-73-69-72-59-80

The highest number this horse has ever improved upon is a 69. He would be an almost automatic toss as people will bet he is going to run 80 again. These types of horses are not as rare as you would think, either.

I disagree, however, with rmania's assertion. Let's say the horse above ran head and head around the track with a this horse:

90-92-95-65-80

No way I am going to project this horse to regress similar to the first horse.

rmania
01-18-2005, 10:50 AM
That is an interesting idea r-

but as you say this is a rare occurance.

how would you use this for any horse that did'nt race neck and neck? and how do you rate the horses 2 races back?
and does the 3 still fall into the same place because he would have been there but for the bad racing luck. and how do you quantify that?

it is an interesting perspective. now how to apply it in real life.
I'm guessing you've thought about this? :D
The rare occurrence just helps to provide support of this theory.

As for the way the rematch went down, I’m trying to emphasize that (after running identical races) both horses returned to displayed basically the same amount of improvement/regression.

Now, let’s suppose that some other horse, on some PREVIOUS date ran a race that was IDENTICAL to the head-to-head encounter. And let’s suppose that that same horse ran again. Could it be that the same amount of improvement or regression shown in that next race could be applied to the Jan 15th race? If so would this tell us, in advance, how well these two horses would perform?

hurrikane
01-18-2005, 10:50 AM
so your assertion would be to bet him if he ran a 69 or less again. Or are you saying look for him to improve below a 70 and decline above a 69?

with the usual caveats imposed(placed right, trn etc)

thoroughbred
01-18-2005, 11:01 AM
“THE” answer is to the question: How fast is a horse going to run today?

Looking at the past performance of any horse you’ll notice that few ever run back-to-back equal times. Horses usually improve or regress from one race to the next. What causes this? Could the effort displayed today be a DIRECT result of the horse’s previous race?

The extremely rare occurrence happened on Saturday, Jan 15th at Santa Anita. It involved two horses, which in their previous race ran neck-and-neck from start-to-finish. This in itself was rare. It became extremely rare when the two horses came back to face each other the next time out.
On that day I posted the following in the Selections forum:

“Race 9
I’m not predicting a winner. Rather, I’m predicting that #3 Attilas Storm and #5 Only in Reno will finish within @ 1 length of each other. Whether first and second, last and next to last, or anywhere in between.

Sounds like a silly prediction but it could be significant if it holds true.

Will explain later.....”

My prediction basically came true. The official distance separating the two at the wire was 2 lengths however the #3 horse was involved in an altercation (leading to a DQ) which probably cost him at least 1 length.

Bottom line... after running IDENTICAL races in their previous effort, both horses displayed the same amount of improvement (or regression). And of course this has always been the case in the past with different horses / different races. Why else would I have made an issue of it.

Now, think about this.... Let’s imagine these two horses going separate directions after their head-to-head duel. Would each display the same amount of improvement/regression? Probably... So by noting the performance of the first to run, you could basically determine (in advance) how fast the second one would run. And, in doing so, you’ve answered “THE” question.

Did you check to see if the difference in Trackk Variant between the two races might explain what appeared to be a similar change in the performance of the horses?

rmania
01-18-2005, 11:08 AM
Did you check to see if the difference in Trackk Variant between the two races might explain what appeared to be a similar change in the performance of the horses?
I would assume that any change in the Track Variant would have an equal effect on both horses.

The point is not that they both improved or both regressed. The point is that the amount was virtually equal.

thoroughbred
01-18-2005, 11:25 AM
I would assume that any change in the Track Variant would have an equal effect on both horses.

The point is not that they both improved or both regressed. The point is that the amount was virtually equal.

I see now. You are correct. Thanks.

Light
01-18-2005, 11:36 AM
I think this was the case between Affirmed and Alydar in their Triple Crown series.

pmd62ndst
01-18-2005, 01:20 PM
Isn't this the "key race" theory? If a couple horses from a particular race improve and do well in their next race, then watch other horses from that same race in their next race?

PMD

cj
01-18-2005, 05:00 PM
so your assertion would be to bet him if he ran a 69 or less again. Or are you saying look for him to improve below a 70 and decline above a 69?

with the usual caveats imposed(placed right, trn etc)

Bet him? Depends on who he's in against and if the improvement makes him a contender, and is he value on the board.

The second part is the key. I'll look for the improve below a 69, and a decline above the 70, which is the important part. Find out when a horse has run too fast for his own good. A top number in the last running line is always bet by the public, but knowing if it is likely to cause some regression can lead to nice profits.

I probably should have mentioned this doesn't apply to young, lightly raced horses.

rmania, no comment on my example match up?

delayjf
01-18-2005, 05:36 PM
Cj, rmania

Interesting points to ponder on both sides. Congrats you've made the delayjf version of "the best of PA". I'm sure you both are numb with awe.

CJ
do you see any bounce patterns with your pace figs?

JustRalph
01-18-2005, 05:48 PM
delay.... I was thinking "what a great thread" when I read your post. I agree, this is what makes this board so interesting.

CJ........do you recommend the use of "lifetime PP's" ? I assume this is so.........

rmania
01-18-2005, 10:58 PM
rmania, no comment on my example match up?
I would have to stick to my guns. If the two horses ran neck-and-neck from start-to-finish I would (based on what I've seen) pick them to run within a length of each other in their rematch. Of course I'm not saying that they would run neck-and-neck again, as this has never been the case, but they would be close at the wire.
One thing I should note is that all of the research I've done in support of this theory has been with sprints.

Tom
01-18-2005, 11:15 PM
http://www.netcapper.com/TrackTractsArchive/TT010330.htm

rmania
01-18-2005, 11:39 PM
Good article Tom. But then one would expect that from Gordon.

Towards the bottom there was some discussion about regression and it was suggested that horese regress when moved up in class.

My experience suggests that horses usually don't regress when moved up. If they fail it's because the competition is too tough. However, I've noticed that horses dropping in class tend to improve more than expected

Tom
01-18-2005, 11:42 PM
There is lots of good reading in the Track Tracts archives.

cj
01-19-2005, 05:42 AM
I would say yes Ralph, lifetime PPs can be an advantage, but you have to use them in context. I wouldn't look for a "bounce point" for a horse that happened four years ago in NW3 allowance races if the horse has been a claimer the last 3 year.


delay,

As for bounce patterns, generally, a pace figure alone won't cause a bounce, its the total performance. Improving pace figures do seem to point to improving horses. Generally a much higher than normal pace figure will cause the speed figure to tail off considerably, but the horses seem to bounce back next time.

Example, a horse usually runs 70-70 or 75-65 or even 80-60 pace speed combos. However, if the horse then has to run 90 pace, don't look for a 50 speed, it could really fall apart. There seems to be a breaking point, the horse could fall right off the map to 30, or even 0.

cj
01-19-2005, 07:45 AM
Here is a case in point from Santa Anita today. I wouldn't look at lifetime PPs, just the last 10 for this horse. He is not the stakes caliber performer he once was.

7 HOT MARKET 9-5 E

SA 12/29/2004 9.0| 97 87 97 87| 83| 93*
--------------------------------------------------------------------------------
SA 10/09/2004 8.5| 103 82 103 79| 67| 88
Dmr 08/21/2004 8.0| 89 86 89 86| 83| 86*
Dmr 07/31/2004 8.0| 96 94 96 94| 92| 94
Hol 06/13/2004 8.5| 95 93 90 81| 74| 82
Hol 05/16/2004 8.5| 99 95 99 95| 91| 95*
SA 04/11/2004 8.0| 90 91 89 86| 82| 85
SA 03/26/2004 8.5| 91 83 89 78| 70| 80
SA 02/21/2004 7.0| 97 110 94 89| 87| 90
SA 01/24/2004 9.0| 112 104 100 72| 61| 77


The horse has 4 figures of 90 or better in his last 10 races. The first one, a 90, cause a 10 point regression. The second one, a 95, caused a 13 point regression. The next was a 94, which was followed by an 8 point drop, even though the horse won the race. Now, last out was a 93. I'm betting he'll decline again, especially seeing as that he is the 9-5 ML favorite. Now, you just have to decide if anyone is capable of beating him if he drops to the mid 80s or below.

rmania
01-19-2005, 08:35 AM
Good analysis CJ...

I'm not sure how beneficial this approach is by itself. It seems there would always be too much of a range to consider. And assuming the same process is used for every horse in the race, each with it's own wide range to consider, you're now back to just making a guess.

One factor you didn't mention is that this horse was running, on average, about once a month from Jan to Oct. Then he gets almost a three month rest and comes back to score a 93 (I'm assuming these are speed ratings). Does the layoff suggest to you the he will regress more than usual today?

rmania
01-19-2005, 10:31 AM
Getting back to the premise of this thread...

I’m convinced, based upon a lengthy analysis and successful predictions in the past, that a horses effort today can, in almost every case, be attributed to it’s previous race (at least in sprints). Simply put, the WAY the horse ran last time out has a direct effect on how fast it will run today.

I’m also convinced that it is possible to actually predict just how fast horses are going to run (+ / - a length). I’m convinced because I’ve done it. ;)

hurrikane
01-19-2005, 08:36 PM
nice analysis cj.

do r....how does the layoff affect your decisions.

2 horses run. one come back in 14 days the other in 45.

how do you handle something like that. very common occurance.

rmania
01-19-2005, 10:42 PM
nice analysis cj.

do r....how does the layoff affect your decisions.

2 horses run. one come back in 14 days the other in 45.

how do you handle something like that. very common occurance.
I'd just go with what I know. The extended time off may result in a little unexpected regression but then again maybe not.

But listen to this......
This extremely rare occurence happened again today......
In the 6th race at Santa Anita the #4 and #7 horses ran together in their last race and ran virtually head-to-head from start-to-finish.
You might just want to check out how they finished in today's race. :eek:

rmania
01-20-2005, 11:22 AM
Here's a link to check out yesterday's results:

http://www.equibase.com/premium/eqbChartResultsDisplay.cfm?TRK=SA&CY=USA&DATE=01/19/2005&STYLE=EQB

Race 6, #4 & #7

After dueling head-to-head in their previous race they finished a nose apart in the rematch.

hurrikane
01-20-2005, 12:12 PM
this is interesting but how can you apply it for value.

if you use it as a key race as in the 7 ran this figure today to she 4 should race within 1 lg of this tomorrow. it doesn't give any relationship to the other horses in the field. Surely they will not all fill the bill.

also, curious if this would only apply to horses in contention.
for instance. You would not use the 4 and 7 again as in this race they were not in contention.

Fastracehorse
01-20-2005, 06:58 PM
Predicting form is fascinating - often ambiguous but neccessary for me.

Your example was great.

I have a fun form - regressor predictor - I call the negative jock switch.

If a horse runs well last start, but makes a jock switch, I look at the animal with tongue-in-cheek. Why the switch??

Today at the Gulf there was a switch to John Velazquez on a horse after he ran super last start. Most people might not notice or acknowledge the switch as negative - because Johnny is one of the best. Well the horse didn't hit the board as the fave - unfortunately my horse came 2nd :(

Side Note: Velazquez is Pletcher's main go to guy. Pletcher had two horses running in race 7. Velazquez is the distractor as Pletcher's other horse went w-to-w at 14-1 with DeCarlo.

fffastt

rmania
01-20-2005, 09:42 PM
this is interesting but how can you apply it for value.

if you use it as a key race as in the 7 ran this figure today to she 4 should race within 1 lg of this tomorrow. it doesn't give any relationship to the other horses in the field. Surely they will not all fill the bill.

also, curious if this would only apply to horses in contention.
for instance. You would not use the 4 and 7 again as in this race they were not in contention.
OK, we've seen twice within a 5 day period where two horses raced
head-to-head (HTH) in their previous race and then finished within a length
of each other in their next race. And, as mentioned earlier, my observation
over the years has shown that this has ALWAYS been the case.

So let's try expanding on this and see where it takes us.

If two horses were to run HTH and then wind up in separate races next time
out would they each display the same amount of regression/improvement?

If a "yes" answer would seem possible/logical then continue reading.

If, on the day of the HTH race, the two horses actually ran in different
races, at the same distance, and each recorded identical fractional times
from start-to-finish (STF), and then raced together in their next race,
would they finish within a length of each other thus displaying the same
amount of regression/improvement?

If a "yes" answer would seem possible/logical then continue reading.

If, as in the previous scenario, two horses recorded identical times in
different races on the same day, at the same distance, then raced in
different races the next time out, would they display the same amount of
regression/improvement?

If a "yes" answer would seem possible/logical then you are subscribing to
my theory that horses which run identical races (time wise from STF), are
likely to display the same amount of regression or improvement in their
next start.

Now, let's assume that there was someway that you could take a horse
running today and find a past race run by some other horse that was
identical. Then by applying whatever amount of regression or improvement
that other horse showed in it's next race, you were able to predict just
how fast today's horse would run. And, if you continued the process for
every horse in today's race then you would be able to predict how fast
every horse would run. If this were possible then picking the winner would
be pretty simple. Right??

hurrikane
01-20-2005, 10:31 PM
R-
I assume you are implementing this on the exact same surface /distance/horses age(older vs 2-3yo).

Interesting idea...have to chew on this for a bit

Overlay
01-21-2005, 12:15 AM
"After dueling head-to-head in their previous race they finished a nose apart in the rematch".

I can't comment on the duplication of this kind of neck-and-neck competition between the same two horses from one race to another, but this discussion falls in line with consistent findings I have seen about a horse's participation in a duel for the lead in the closing stages of its latest race being a positive sign for its performance in its next start (whether it comes back racing against the same horse it duelled with previously or not, and regardless of the length of elapsed time between the duel and the horse's next start, or the distances of the races involved). And, contrary to what one might expect, the positive effect on subsequent form gets stronger the longer the duel in the last race lasted. (For instance, a duel lasting through the last three calls of a horse's last race is more powerful that a duel over only the last two calls.)

delayjf
01-21-2005, 02:25 PM
Here is a case in point from Santa Anita today. I wouldn't look at lifetime PPs, just the last 10 for this horse. He is not the stakes caliber performer he once was

Code:
7 HOT MARKET 9-5 E

SA 12/29/2004 9.0| 97 87 97 87| 83| 93*
--------------------------------------------------------------------------------
SA 10/09/2004 8.5| 103 82 103 79| 67| 88
Dmr 08/21/2004 8.0| 89 86 89 86| 83| 86*
Dmr 07/31/2004 8.0| 96 94 96 94| 92| 94
Hol 06/13/2004 8.5| 95 93 90 81| 74| 82
Hol 05/16/2004 8.5| 99 95 99 95| 91| 95*
SA 04/11/2004 8.0| 90 91 89 86| 82| 85
SA 03/26/2004 8.5| 91 83 89 78| 70| 80
SA 02/21/2004 7.0| 97 110 94 89| 87| 90
SA 01/24/2004 9.0| 112 104 100 72| 61| 77

The horse has 4 figures of 90 or better in his last 10 races. The first one, a 90, cause a 10 point regression. The second one, a 95, caused a 13 point regression. The next was a 94, which was followed by an 8 point drop, even though the horse won the race. Now, last out was a 93. I'm betting he'll decline again, especially seeing as that he is the 9-5 ML favorite. Now, you just have to decide if anyone is capable of beating him if he drops to the mid 80s or below..



CJ,

I think this is one I would have mis-read. Looking at the pace figures for the above horse, I might have concluded that his bounce after the 94 / 95 speed ratings was caused by the high pace figures he earned in both races.
In his last race, his pace ratings were not nearly as high, indicating to me that he was not all out. I wouldn't have played him as the favorite, but I would have expected him to run about a 92-95 or so. I guess that's why I couldn't get into the Air Force.

rmania
01-22-2005, 11:03 AM
R-
I assume you are implementing this on the exact same surface /distance/horses age(older vs 2-3yo).

Interesting idea...have to chew on this for a bit
Let me say first that I typically try to avoid 2yo races.

Second, I would only try to apply this theory in sprints.

With regards to surface, I don’t believe that the surface type has much to do with the amount of improvement or regression a horse displays.

So, what I’m basically saying is that I would attempt to apply this theory to any (non 2yo) sprint race.

If you recall the title of this thread, I made reference to “THE” answer. That’s because there’s really only ONE question about this game that no one has ever been able to answer. And that is “How fast are these horses going to run?”.

There are many ideas (both published and unpublished) which suggest ways to determine if a horse will improve or regress. However, the actual amount is never part of the equation, as was the case with cj’s contribution to this thread. And, as we all witnessed, even these ideas (which merely suggest a direction) are not all that dependable.

rmania
01-22-2005, 11:56 AM
BTW, here is another example of the phenomenon (?) which prompted this thread. This example is from 1999 which should suggest that this is nothing new. Rather it’s something that has taken place all along.

These are the past performance lines of two horses that met in a Nov 13th race. Note that they ran virtually head-to-head from start-to-finish in that race.
http://www.paceadvantage.com/forum/attachment.php?attachmentid=496&stc=1

Now, here is the chart from the rematch.
http://www.paceadvantage.com/forum/attachment.php?attachmentid=497&stc=1

As you can see, in that rematch they both displayed the same amount of regression (within 1/5 of a second).

rmania
01-26-2005, 08:58 AM
In trying to find a recent example that substantiates my theory, I came across one that may work.

On Jan 5th at Santa Anita, Race 7, two horses coming out of DIFFERENT races, yet running identically time wise, finished 1st and 2nd in the Jan 5th race.

Unfortunately, I don’t have the chart so I don’t know how close they were at the wire. Maybe someone on the board has the chart and can post how close they were at the wire.

The two horses were #6 Tizakitty and #10 Mazella.

hurrikane
01-26-2005, 10:28 AM
R-

[
75000 Stakes Ng Track wet fast Time 22.05 45.22 109.56 116.26 Weather clou

-----------------------------------------------------------------------------
sc horse pp st 2f 4f 5.5f 6.5f odds
-----------------------------------------------------------------------------
6 Tizakitty 3 2 1-0 1-0 1-0 1-0 5.4
10 Mazella 6 1 2-2.5 3-1.5 2-2.5 2-1 8.8
9 Dream of Summer 5 6 5-4 4-4 3-5 3-1.5 0.4
3 Cryptos' Best 1 5 6-5 5-5.5 5-7.5 4-6.5 17.1
8 Dis Miss 4 3 3-2.5 2-1.5 4-5 5-8.5 18.1
5 Madringa 2 4 4-3.5 6-9.5 6-10.5 6-12 7.0
-----------------------------------------------------------------------------

6 Tizakitty.................. $12.80 $6.40 $2.10
10 Mazella.................... $8.00 $2.20
9 Dream of Summer............ $2.10



I'm having a hard time finding a way to apply this to test it.

are you saying every horse at SA in a sprint that runs the exact time these 2 did previously will improve/regress the same amount these 2 did?

rmania
01-26-2005, 10:58 AM
R-

[
75000 Stakes Ng Track wet fast Time 22.05 45.22 109.56 116.26 Weather clou

-----------------------------------------------------------------------------
sc horse pp st 2f 4f 5.5f 6.5f odds
-----------------------------------------------------------------------------
6 Tizakitty 3 2 1-0 1-0 1-0 1-0 5.4
10 Mazella 6 1 2-2.5 3-1.5 2-2.5 2-1 8.8
9 Dream of Summer 5 6 5-4 4-4 3-5 3-1.5 0.4
3 Cryptos' Best 1 5 6-5 5-5.5 5-7.5 4-6.5 17.1
8 Dis Miss 4 3 3-2.5 2-1.5 4-5 5-8.5 18.1
5 Madringa 2 4 4-3.5 6-9.5 6-10.5 6-12 7.0
-----------------------------------------------------------------------------

6 Tizakitty.................. $12.80 $6.40 $2.10
10 Mazella.................... $8.00 $2.20
9 Dream of Summer............ $2.10



I'm having a hard time finding a way to apply this to test it.

are you saying every horse at SA in a sprint that runs the exact time these 2 did previously will improve/regress the same amount these 2 did?
Thanks for the info hurrikane.

And as I suspected, the two horses finished within a length of each other, thus displaying the same amount of improvement/regression (+ or - a length).

I’m really not surprised that you, or anyone else, might be having a difficult time testing this theory, especially the part where horses from different races are used.

And just for the record, these two horses ran their “identical” races at Hollypark.

So what I’m basically saying is that every horse that ran the exact time these 2 did previously will improve/regress the same amount these 2 did.

hurrikane
01-26-2005, 08:16 PM
ok r.
might be able to test that. what are the parameter...within 1/5 of a second of each time?

what do want to use for the measure of regression/improvement speed? fps? actual times?

I don't know what if anything we'll get out of it but it's an interesting idea.

Tom
01-26-2005, 10:19 PM
You are assuming that all horse share the same form cycle?

rmania
01-27-2005, 08:12 AM
You are assuming that all horse share the same form cycle?
No....
These findings are actually contrary to the notion of a form cycle as they suggest that a horse’s effort today is directly attributed to it’s previous race only.

In EACH of my examples, two horses ran identical races (start-to-finish) then produced nearly identical finishing times in their next race. In NONE of my examples did the two horses run identical races again in that next start. So, in each example, after the second race, both horses were on a different path for their next race.

delayjf
01-27-2005, 06:17 PM
Just to clarify, by identical race, you mean that the two horses in question either dueled with each other the entire race or both pressed the same pace or both rallied from off the pace? Or are you saying that the qualifing horses must have dueled the entire race?

Now to complicate things, how does your theory handle distance changes. Say for example one horse from the duel in a sprint wins a route race in its next start. Can we expect the same level of performance from the second horse if it to is entered in a route?

rmania
01-28-2005, 08:32 AM
Just to clarify, by identical race, you mean that the two horses in question either dueled with each other the entire race or both pressed the same pace or both rallied from off the pace? Or are you saying that the qualifing horses must have dueled the entire race?

Now to complicate things, how does your theory handle distance changes. Say for example one horse from the duel in a sprint wins a route race in its next start. Can we expect the same level of performance from the second horse if it to is entered in a route?
When referring to two horses running identical races in the same race, the
premise is that they are side-by-side the entire trip. They can be on the
front end, mid-pack or at the rear (closing or not). In other words, their
"fractional times" for the race are identical.

If you scroll up to my most recent example you'll see that it identifies
two horses that had identical "fractional times" in different races, then
met in their next start and displayed an equal amount of
improvement/regression.

As for changes in distance, I wouldn't think that any of this would apply
to horses going a mile or longer. There are just too many factors going
long that could change the outcome.

Now here's something else to chew on....

In yesterday's 6th race at Santa Anita there were two horses which raced last at DIFFERENT DISTANCES yet, through a conversion process I use, ran (what came out as) identical fractional times and yep they both showed a similar amount of regression yesterday. One of the horses was the favorite and both were well beaten.

hurrikane
01-28-2005, 11:11 AM
interesting r-

curious were they shortening up or stretching out.

also I"m curious. are you using fractionals to the 1/5 of a second.

my data is at 100ths and I need to round up to get more hits as I'm sure you can imagine not many horses run exactly to the 100th of a second

rmania
01-30-2005, 10:31 AM
interesting r-

curious were they shortening up or stretching out.

also I"m curious. are you using fractionals to the 1/5 of a second.

my data is at 100ths and I need to round up to get more hits as I'm sure you can imagine not many horses run exactly to the 100th of a second
1) They were stretching out to 7fs.

2) I'm using 1/5s.

rmania
02-02-2005, 08:44 AM
So hurrikane, how's the hunt going?

I noticed that in a race on Sunday at Santa Anita, two horses (Going On Again & Ms Forty Second St) had come out of the same race. Though their fractional times were not "identical", they were basically within a length of each other the entire trip.

In Sunday's race (Race #1) they ran completely different races yet finished one length apart at the wire.

Just for the record, Going On Again worked between races whereas Ms Forty Second St had not posted an official workout since November.

BTW, in addition to using 1/5s I'm also using just the 3 major points of call (i.e., qtr, half, and finish) when determining "like" races.

JPinMaryland
02-02-2005, 04:12 PM
The lowest bounce pt. article is interesting, and the author does not claim to have the last word, however the article doesnt seem to address a factor that must exist in all this, that simple law of averages will tell you that if a horse ran at his highest OR even merely above his own average he is more likely to fall down.

Yes? Well assume just for argument sake, that horse performances from race to race was behaving randomly. Equal chances up or down for next race. THe patterns would like more like static electricity or white noise or something...

If you were to cull from that set only those data pts. that meet the criteria of Lowest Bounce Pt. these data pts. would no doubt be in the upper reaches of the spectrum. Maybe in the top 1/3 of data pts. or top 2/5 whatever, my pt is valid so long as this subset is in the top 1/2.


So if the horses were behaving randomly, and you picked out a subset of lowest bounce pt data pt, the law of probability would say that the very next data pt. is probably going to be lower...

RIght? Maybe 80% of the pts will be lower or maybe 53% but more than 1/2 will be a lower score.

Bill James did a similar study in baseball and called it the "plexiglass principle." He was studying teams that finished above average and did they really have a long term improvement or was it one season that was different. His analysis may be inadequate but the term plexiglass and the image it draws is illustrative. It is as if each team was plexiglass and if you flex them up they will no doubt bounce back. The tend to remain a flat sheet of plexiglas over time but they can move up and down.


Well my only pt is this. Any such stud of bounce pts. will have to account for how much of this effect is randomness, as if horses were randomly doing this and how much is greater than expected from randomness. I.e. one has to zero out for random fluctuations, and then substract all that and if there is still something left over that might be relevant.

THe article said 2/3 of the horses had dropped down in the next race. At least 51% of that would have occured randomly anytime you took a set of above average data pts.

The question then becomes: how high up in the scale was this study's set of data pts? I.e. how much of a zero out correction do we have to make to accont for random fluctuation behavior?

Well I dunno but think about this, the article found 2/3. One standard deviation in statistics usually covers about 2/3 of the data pts. Maybe the equine researcher is simply staring into the pile of data pts and it is staring right back at him...