Pace and Speed in Harness

[Ray2000]

There have been many studies / discussions on T-Bred handicapping to determine the effect of pace on speed (some are going on right now in the General Forum) and they probably apply to harness, but I would be interested in harness player's opinion on the subject.

Consider the following 3 sets of fractional times,
(Ignoring racing wide, racing with cover, battling, number of turns, etc)

Horse A) 30 30 30 30 = 120 Final Time
Horse B) 31 31 31 27 = 120
Horse C) 29 31 31 29 = 120

1.
Is there agreement that horse "B" (single brush 31-to-27) has put out the most total "horse-power"?
Followed by horse "C" (2 brushes to 29), and lastly horse "A" (no brushes)?

2.
If so (or if not), what would be a good way to express a speed rating based on those fractions/final times?


I have been using Final Time + 4thQT* and have tried Final Time + bestQT in the past, but there are other possibilities. i.e.
Final Time + (4thQ * 2)
Final Time + 4thQ + 1stQ
Final Time + Best 2 Qs

BTW
Tom Ainslie** used Final Time + 1st Half Time (assigning arbitrary points from a table)

My current numbers for the above horses would be 150, 147, 149, (lowest is best, Horse B)
Ainslie is 240, 230, 240 (highest is best, Horse A or C)

I'd be interested in any replies and would be willing to compare any promising formulas by running a large sample set of races. More on that later.

Thanks for any input.



*all times adjusted for Track, wides, cover, DTV
** Complete Guide to Harness Racing p. 179

[melman]

Ray--It's funny how we differ yet many times come up with the same horses.

I would use final time plus middle half. I love to use the middle half (2nd and 3rd q times) as I feel many people overlook them.

That would give me "Horse A" then. :)

I do overlook a horse however who in the 1st q raced last or next to last then make a "middle move" into oblivion.

IMHO Stanley Dancer really changed how the driving of harness horses is done. You used to see the what is still done in much of France and Sweden with there trotters. And that is go slow early then have a mad scramble in the last q to see who gets up. Dancer went fast early and said "catch me if you can". And as poster Pandy has pointed out the new bikes also made speed more important. Nowadays even with a speed duel it's tough to "come from the clouds".

[Sinner369]

I always read that if Final Times are the same.......the horse with the faster
pace to the 3/4M..........ran the better race.

That applies to both thoroughbreds and harness.

[Tom]

mel, I use the three halves - got that from a book way back when - the 70's?
Can't remember his name, Arron Bernstein?

I add the faster of the first or third half to the middle half for a rating.

[melman]

Tom==Hope your doing well with that rating. It makes a lot of sense to me. I think the one item that gets overbet quite a bit is the fast last q. Best of trips.

[harness2008]

The way that I make my figures would give Horse B a big edge, followed by Horse C, then Horse A.

Since all 3 horses ran the same identical final time, the best horse of the 3 would be the one that rang up the highest oxygen debt. However the key is finding the right calculation that can compute this.

While I enjoyed the works of Tom Ainslie and his many great handicapping books, I believe he completely missed the boat by including the harness pace chart and the expectation of cranking out decent figs for harness horses with it. He should have stuck with the T-breds.

It must be understood that the Ainslie pace calculation chart will not work for harness racing. Because of the nature of harness racing, Ainslies chart would grossly under estimate a stretch runner that was hampered by a slow early pace. Any gain such as this in a fast last half would definately be a positive for a harness horse but because Ainslies chart leans towards faster times to the half,(which for a T-bred would be good) a horse such as this one would be unjustly downgraded.

The simple reason for this is that quarter mile times in harness races are run differently than T-breds. Generally speaking the T-breds quarter times will continue to decelerate as the race lengthens while this is not so in harness racing.

I've found that a fast fraction regardless of which fraction it was in harness racing is a good sign. I don't steadfastly look at any individual quarter and include it in a calculation, I look at all quarters and assign a value to each quarter based on whether or not that fraction is above or below the race par. So therefore the faster fractions run up a higher oxygen debt and need to weighted higher than the slower fractions of a race to reflect this.

What I want to do is give credit for any frontrunners that ran into a fast early pace while giving extra credit to the closing horses that were stuck behind a slow paced race and came flying late into a fast last half or quarter. Each would be considered an above average performance.

What I essentially do is give a %value to each individual quarter based on how fast or how slow each fraction is in relation to the final time. This differs of course with every track.

Here's the math for better or worse.

For example Horse A has equivalent quarters for each of the 4 fractions of the race. Each quarter then is basically worth 25% of the total time. If you multiplied each quarter mile time by 25%, then add up all 4 quarters times 4, you get the final time which is 2:00. Each quarter had the same 25% value because all were run in the same time.

Now look at what happens when a quarter or quarters are run faster than par.

Horse B has 3 quarters which are slower than the average and the last quarter is faster than average. Each of the first 3 quarters must be worth less than 25% since they were slower than par, with the last fraction obviously worth more than 25%. Depending on the track, and this calculation is just being used as an illustration, you may for example find that each 1 second increment below or above the race average may be worth 5%. This %number is being used for example purposes only.

So for Horse B since each of the first 3 fractions are slow by 1 second, each fraction is worth a value of 20% each, while the last fraction is 3 seconds above the par so that is worth 40%. Doing the math yields an adjusted final time of 157.6

So as you can see by this calculation, whenever there is a faster than average quarter in a race. it raises the oxygen debt of a horse and is in effect worth more percentagewise than a slower than average fraction.


Now for Horse C using this same 1 second = 5% formula, (and again this % is just being used as an example) the first and last fraction would be worth 30% each while the 2nd and 3rd fraction each worth 20%. Sparing you the math this equates to an adjusted final time of 159.2


I use a method somewhat different than the one described above which includes the horses actual fractional times in conjunction with the pace of the race but I wrote what I wrote above so you can get a general idea of what I have done for many years. My God, the calculator buttons I have pushed in my lifetime.

The bottom line with these figs for me is that I upgrade a horses final time whenever he runs into the teeth of the faster fractions of the race, whether it be in the early pace, late pace, or middle pace or any fraction for that matter. After years of trying to come up with some mathematical calculation that would do this justice, this is what I have used with a degree of success.

So to answer your question, if these 3 horses met in a race all other things being equal, Horse B in my opinion would win in a romp. If anyone disagrees, please fire away.

[Hanover1]

I fell asleep reading all the theory here. the old adage of: Study long, study wrong" hold water. Thats why I look at why a horse might be chalk, and figure out who has a shot to beat them, using no complicated formula or computations. Horses are very poor at math.........

[Ray2000]

All:

Thanks for the replies so far, some diverse approaches here leading me to do some refinement and testing.

Mel:

quote "...many times come up with the same horses"
is because we both recognize speed rating is only a small part of overall handicapping.

Harness2008:
Exactly the type of reply I was looking for, thanks. I would add (as a tie-breaker) that a pp line

27 29 29 29 ...is inferior to... 29 29 29 27

because maximum effort in the 4thQ is during the highest oxygen debt (lactic acid build up).

That's what led me to using my current speed ratings, Final Time + 4thQT but I'm starting to believe this is not the best way.


Hanover1:
The game has room for different types of players. For 20 years I was a steady track patron and used the printed program/magic-marker/human contacts to do my handicapping, but now I'm stuck in a corn field in Clarion, Pa and all I have to go on is a database of numbers. I was curious to see if those numbers alone could be sufficient to produce enough profit to pay for the hobby. The answer is yes, but in my case, requires making a lot of small bets on a large number of races (Some at teeny-tiny tracks). Searching for those races, requires a good number crunching program, which I'm always trying to improve.

Good Luck with your style of 'capping.

[Tom]

Wouldn't you prefer a 27 29 29 29 on a half mile track and 29 29 29 27 on a mile?

[Ray2000]

I see your point, Tom, but I was still applying my 'ignore list' of variables.

(Ignoring racing wide, racing with cover, battling, number of turns, etc)

a lot of adjustments go into my final Speed rating for a real pp line

[harness2008]

In Steve Chaplin's book, he explained that by taking the final time and dividing by the 4 fractions, the average 1/4 mile time could be calculated for a race. It was in this way that he established his shape ratings.

Do not believe it, nothing could be farther from the truth. I did use that math as an example in my earlier post only so our forum could grasp an understanding of what I was doing with each fraction.

However doing it Chaplin's way is completely wrong. You can't just take the final time, divide by 4 and come up with the average fraction and use those to see if the pace of the race was either fast or slow at different stages.

Each racetrack is unique in its characteristics and there are different par times for different track final times.

What I did was to take the average of fractional times in all races at a particular track to come up with my figures. Thru linear regression analysis I could determine based on a particular final time, what would be the normal fractional time equivalents for any final time in question.

There were times at some tracks where a 28 second opening quarter would be the norm for a time run in 2:00. Chaplin would have had this opening quarter as lightning fast and would upgrade a horse on the lead since he would have used the average of 30 seconds. Utter nonsense.

Before one can make any judgements as to whether or not a particular fraction in a race would be determined as fast or slow, one needs to examine hundreds of races at a particular racetrack and determine the fractional equivalents of all final times.

[Charlie]

Harness2008 ... Andy Beyer in his well written book "Beyer On Speed" used the same type of analysis. This can be found on page 146. For route races, he simply determined the average 6-furlong time for various final time ranges. Then he used this data to determine how fast or slow the pace was from the average. A very similar approach to yours, but one that I really haven't tried yet. I just may start accumulating the data to use at Yonkers. Thanks to all for the input. This is a very interesting topic.

[Hanover1]

All:

Thanks for the replies so far, some diverse approaches here leading me to do some refinement and testing.

Mel:

quote "...many times come up with the same horses"
is because we both recognize speed rating is only a small part of overall handicapping.

Harness2008:
Exactly the type of reply I was looking for, thanks. I would add (as a tie-breaker) that a pp line

27 29 29 29 ...is inferior to... 29 29 29 27

because maximum effort in the 4thQ is during the highest oxygen debt (lactic acid build up).

That's what led me to using my current speed ratings, Final Time + 4thQT but I'm starting to believe this is not the best way.


Hanover1:
The game has room for different types of players. For 20 years I was a steady track patron and used the printed program/magic-marker/human contacts to do my handicapping, but now I'm stuck in a corn field in Clarion, Pa and all I have to go on is a database of numbers. I was curious to see if those numbers alone could be sufficient to produce enough profit to pay for the hobby. The answer is yes, but in my case, requires making a lot of small bets on a large number of races (Some at teeny-tiny tracks). Searching for those races, requires a good number crunching program, which I'm always trying to improve.

Good Luck with your style of 'capping.
I have always admired you guys that had the time and inclination to make a profit at the game-kudos to you. I have always had the excuse of being a horseman, wich means study the horses, and leave the math to the owners...lol. I do look at workouts, but its usually live, and I rely on that, along with the word that gets around the backside on works, and intentions. I was never beyond a few bucks on one of my charges if I felt we had a shot, but the tough job is yours.

[harness2008]

Charlie, you are correct about the Beyer book. Analysis such as this is necessary in the T-bred game because of the deceleration of horses as the distances lengthen. You can't just take for example the time of a 6 furlong race, divide by 3 and have the average fraction. It wouldn't work because of the way horses decelerate.

However for some reason it became apparent to certain individuals writing books that an easier way in harness racing is to just take the final time, divide by 4 and Poof!!, you have the average 1/4 mile speed.

I did not write this to disparage Steve Chaplin and his harness book, but it's apparent to anyone who has had an interest in this sport that his calculations were incorrect. However I will say that reading his book even with his inaccuracies has led me in the right direction of finding a viable solution to the problem. So in a wierd sense, I'm glad to have purchased it and read it back in the day.

[am1947]

"Thru linear regression analysis "
Could you please explain this a bit more with an example?
Thank You
AM

[harness2008]

For people that are bored with any type of math analysis, you may want to just skip this read. For those that are not, here is the way that I compute par times for harness racing. These are par times for a specific final time run, not for any specific class of race run. There is a difference.

I have thought of different ways to compute this over the years and came up with something that works for me. Years ago I thought I would just mark down each final time on a page in a notebook and then write down each corresponding quarter time of the race. The problem with this method is that you need to have a ton of races for each specific final time and then you need to have some way of smoothing the figures if the averages don't work out right. Also, what if down the road final time is computed to the hundreds of a second like the T-breds? There was no way it could be done this way and I needed to find a better way to compute this.

I mentioned that one would have to chart hundreds of races at a harness track before you can compute typical fractional times that relate to individual track final times. Obviously the bigger the database that you have, the better the figures should be. I also wanted to utilize some sort of equation where I could plug the final time into and come up with each specific 1/4 mile par for the final time. I did not want to refer to a myriad of charts for this. A linear regression equation works just fine.

Here's are the times for a racing card run at Vernon Downs October 3rd, 2009. This is just 1 racing card and is used as an example only so I can explain the math behind it. Also to ease the calculations these times are in whole seconds with 10ths of seconds.

Race 1 - 27.6 / 58.4 / 88.4 / 116.2
Race 2 - 26.6 / 56.2 / 85.4 / 114.6
Race 3 - 29.6 / 59.2 / 88.2 / 118.0
Race 4 - 27.8 / 56.8 / 85.0 / 113.2
Race 5 - 27.6 / 57.6 / 86.0 / 114.6
Race 6 - 26.8 / 56.2 / 85.6 / 114.6
Race 7 - 27.0 / 57.0 / 85.2 / 113.0
Race 8 - 28.6 / 58.8 / 87.8 / 116.6
Race 9 - 27.2 / 57.0 / 86.2 / 113.2

Breaking down each qrt. mile segment we get,

Race 1 - 27.6 / 30.8 / 30.0 / 27.8 116.2
Race 2 - 26.6 / 29.6 / 29.2 / 29.2 114.6
Race 3 - 29.6 / 29.6 / 29.0 / 29.8 118.0
Race 4 - 27.8 / 29.0 / 28.2 / 28.2 113.2
Race 5 - 27.6 / 30.0 / 28.4 / 28.6 114.6
Race 6 - 26.8 / 29.4 / 29.4 / 29.0 114.6
Race 7 - 27.0 / 30.0 / 28.2 / 27.8 113.0
Race 8 - 28.6 / 30.2 / 29.0 / 28.8 116.6
Race 9 - 27.2 / 29.8 / 29.2 / 27.0 113.2

Now what I do is add each of the columns separately, divide by the total number of races, then the key to the method, add 1 standard deviation up and 1 standard deviation down from the average.

With a regression analysis, mathematically you need to have more than just 1 comparison. If I just took the average of each column, the only comparison that I would have would be to the average final time of the set and to no other final times. So mathematically this is the reason that the decision was made to add and subtract 1 standard deviation from the average.

The first qrt mile average is 27.644 with a standard deviation of .948
The average final time is 114.889 with a standard deviation of 1.727

Doing the math, I can now compare a final time with a first qrt. time

113.162 = 26.696
114.889 = 27.644
116.616 = 28.592

By plugging these numbers into a handheld calculator that has the regression function, I am able to compute a linear regression equation for the first qrt. mile segment.

The formula for the first qrt. mile equivalent is -

Final Time X .548928778 - 35.42187799


I can now input the final time of any race into the equation and arrive at a typical first qrt. mile time for any specific final time inputted.

Now what needs to be done is to compute these figures the same way for the 2nd, 3rd, and 4th fraction. So you will have 4 regression equations, each one specific to each qrt. of a race. If done correctly when all computations are complete, all 4 fractional times derived from the formulas will add up to the actual final time inputted to begin with.

In this way based on hundreds of charts of actual races run, I have accurate pace guidelines for any final time that is run at a given racetrack. Much better than Chaplin's taking the the final time and dividing by 4 method.

Also when you have computed all 4 fractional times for a specific final time, you now have the typical 1st half, 2nd half or middle half par times at your disposal by adding the appropriate fractions.

[Sea Biscuit]

For people that are bored with any type of math analysis, you may want to just skip this read. For those that are not, here is the way that I compute par times for harness racing. These are par times for a specific final time run, not for any specific class of race run. There is a difference.

I have thought of different ways to compute this over the years and came up with something that works for me. Years ago I thought I would just mark down each final time on a page in a notebook and then write down each corresponding quarter time of the race. The problem with this method is that you need to have a ton of races for each specific final time and then you need to have some way of smoothing the figures if the averages don't work out right. Also, what if down the road final time is computed to the hundreds of a second like the T-breds? There was no way it could be done this way and I needed to find a better way to compute this.

I mentioned that one would have to chart hundreds of races at a harness track before you can compute typical fractional times that relate to individual track final times. Obviously the bigger the database that you have, the better the figures should be. I also wanted to utilize some sort of equation where I could plug the final time into and come up with each specific 1/4 mile par for the final time. I did not want to refer to a myriad of charts for this. A linear regression equation works just fine.

Here's are the times for a racing card run at Vernon Downs October 3rd, 2009. This is just 1 racing card and is used as an example only so I can explain the math behind it. Also to ease the calculations these times are in whole seconds with 10ths of seconds.

Race 1 - 27.6 / 58.4 / 88.4 / 116.2
Race 2 - 26.6 / 56.2 / 85.4 / 114.6
Race 3 - 29.6 / 59.2 / 88.2 / 118.0
Race 4 - 27.8 / 56.8 / 85.0 / 113.2
Race 5 - 27.6 / 57.6 / 86.0 / 114.6
Race 6 - 26.8 / 56.2 / 85.6 / 114.6
Race 7 - 27.0 / 57.0 / 85.2 / 113.0
Race 8 - 28.6 / 58.8 / 87.8 / 116.6
Race 9 - 27.2 / 57.0 / 86.2 / 113.2

Breaking down each qrt. mile segment we get,

Race 1 - 27.6 / 30.8 / 30.0 / 27.8 116.2
Race 2 - 26.6 / 29.6 / 29.2 / 29.2 114.6
Race 3 - 29.6 / 29.6 / 29.0 / 29.8 118.0
Race 4 - 27.8 / 29.0 / 28.2 / 28.2 113.2
Race 5 - 27.6 / 30.0 / 28.4 / 28.6 114.6
Race 6 - 26.8 / 29.4 / 29.4 / 29.0 114.6
Race 7 - 27.0 / 30.0 / 28.2 / 27.8 113.0
Race 8 - 28.6 / 30.2 / 29.0 / 28.8 116.6
Race 9 - 27.2 / 29.8 / 29.2 / 27.0 113.2

Now what I do is add each of the columns separately, divide by the total number of races, then the key to the method, add 1 standard deviation up and 1 standard deviation down from the average.

With a regression analysis, mathematically you need to have more than just 1 comparison. If I just took the average of each column, the only comparison that I would have would be to the average final time of the set and to no other final times. So mathematically this is the reason that the decision was made to add and subtract 1 standard deviation from the average.

The first qrt mile average is 27.644 with a standard deviation of .948
The average final time is 114.889 with a standard deviation of 1.727

Doing the math, I can now compare a final time with a first qrt. time

113.162 = 26.696
114.889 = 27.644
116.616 = 28.592

By plugging these numbers into a handheld calculator that has the regression function, I am able to compute a linear regression equation for the first qrt. mile segment.

The formula for the first qrt. mile equivalent is -

Final Time X .548928778 - 35.42187799


I can now input the final time of any race into the equation and arrive at a typical first qrt. mile time for any specific final time inputted.

Now what needs to be done is to compute these figures the same way for the 2nd, 3rd, and 4th fraction. So you will have 4 regression equations, each one specific to each qrt. of a race. If done correctly when all computations are complete, all 4 fractional times derived from the formulas will add up to the actual final time inputted to begin with.

In this way based on hundreds of charts of actual races run, I have accurate pace guidelines for any final time that is run at a given racetrack. Much better than Chaplin's taking the the final time and dividing by 4 method.

Also when you have computed all 4 fractional times for a specific final time, you now have the typical 1st half, 2nd half or middle half par times at your disposal by adding the appropriate fractions.

You seem to be very good at maths.

My questions to you is two fold:

1.How long does it take you to do one program and putting all the values to each horse in a program.

2. Bottom line question. How successful are you in making money and beating this game.

Kudos to you my friend for bringing up this notion of adjusting final times by its pace numbers. You are probably the first ever that I have come across who has attempted to do that.

Sea Biscuit

[harness2008]

You seem to be very good at maths.

My questions to you is two fold:

1.How long does it take you to do one program and putting all the values to each horse in a program.

2. Bottom line question. How successful are you in making money and beating this game.

Kudos to you my friend for bringing up this notion of adjusting final times by its pace numbers. You are probably the first ever that I have come across who has attempted to do that.

Sea Biscuit


Sea Biscuit, thank you very much for the kind words. Hearing that from you means a lot.

I like to do the last 3 lines for a horse as long as its within the current form cycle. I've become quite adept at pushing calculator buttons in my lifetime but this is a method that does take some time to do accurately.

Not only must you calculate the interior fractions which I usually do in my head, you need to refer to the specific track chart where the horse ran for the appropriate equations and this must be done 4 times for each running line since of course there are 4 fractions.

All of this could be implemented within a computer program which would make the calculations instantaneous. However, I have zero experience in computer programming. A typical program of 10 races with 8 horses in each race, calculating 2 or 3 lines a horse, all done by hand would normally take me around 1 1/2 - 2 hours.

To me, its time well spent in an effort to uncover the true contenders in a race. My track friends think I'm nuts and they usually refer to public selectors or just peruse the program for a few minutes to come up with their choices. You know how it goes when you're into this game, you'd rather bet on your own losers than someone else winners.

The bottom line though is that I need to have accurate speed ratings, those I do not compute myself. I merely adjust the speed rating based on the pace figs. If the speed figs fall flat on their face, so do I.

However, I have kept records of my bets the past few years and have been successful at my home track here in upstate NY, Vernon Downs to the tune of a 22% ROI this year. However of course, I don't bet every race and there are times here at this track just like any other track where nothing seems to work. I use the numbers basically as a tool in determining how fast a horse is capable of running and more importantly if a horse is on the improve or may be tailing off. Just using raw final times does not accomplish this.

The new Trackmaster pace past performances sort of mirrors what I've attempted to do for years now. I am not sure yet if their pace numbers are truly accurate and a good tool to use but if they are would definately cut the workload for me. However, of course I would just rather do it myself.

[Ray2000]

My picks are getting clobbered at the tracks so I'll comment here.
Good explanation, Harness2008, thanks again


I'm also using linear regression but at a different stage of the handicapping process. After establishing a speed rating for each pp line for each entry I use the technique to project an expected Speed for each entry in today's race. I use 4-7 pp lines but no line older than 2 months. I don't use TrackMaster Speed numbers I prefer my own, (I don't believe TM recognizes the trailing post) but I'll use their scale as an example of linear regression (also called trend-line projections).

If the speeds for a horse's last 4 races are....93-94-95-96 then the math would say a 97 today. This example is overly simple but the method can draw a linear projection with any set of numbers. The method also produces the slope of the line which relates to improving or degrading form. Of course the animal may have reached form reversal, in which case you can throw the projection in the garbage, but you only need to be right more often than wrong.

[Sea Biscuit]

Harness2008:


Please check your in box for a message

Sea Biscuit

[am1947]

H2008
Thanks for supplying the excellent example.
Below I have tried to do the numbers out for Q2,Q3,Q4.

I think I followed well UNTIL the"handheld regression" part.
The formula for the first qtr mile equiv is
(Final Time) x 548928778 - .542187799

My question is where did the Bold numbers come from?
and the equation itself ?

The reason I am so interested is I use a std dev on the 1/2 mile
times of each race relative to the other races for a particular night.

Do you find that the quarter mile times are more relavent then the halves?

Thanks again for taking the time to share your interesting insights.
AM :)


Q1 Q2 Q3 Q4 Final
Avg. 27.644 29.822 28.956 28.467 114.889
------------------------------------------------------
StdDev 0.948 0.514 0.598 0.854 1.727
------------------------------------------------------

Q1 113.162 26.696 0.948 less
114.889 27.644
116.616 28.592 0.948 more

Q2 113.162 29.308 0.514 less
114.889 29.822
116.616 30.336 0.514 more

Q3 113.162 27.998 0.598 less
114.889 28.956
116.616 29.194 0.598 more

Q4 113.162 27.613 0.854 less
114.889 28.467
116.616 29.321 0.854 more

[harness2008]

Linear regression would produce an equation in the form y = a + bx

y, the predicted value represents 1/4 mile time

x, the predictor variable represents final time

The main purpose of a linear regression model is to determine what the proper values are of a and b

With a scientific calculator you would be inputting 3 sets of data points for the first 1/4 mile figures.

113.162, 26.696
114.889, 27.644
116.616, 28.592

Each data point contains two numbers, the first one would be the final time and the second one would be 1/4 mile time. The easiest way to show you how I get the actual equation is to go to this link which is a Linear Regression calculator.

http://people.hofstra.edu/Stefan_Waner/RealWorld/newgraph/regressionframes.html

Now input the data points that I listed above and click on the radio button just beneath it that says y = mx + b to get your answer. The equation is automatically done for you. The Regression coefficient in all cases should always be 1 or the initial values inputted are not correct. If it is less than 1, then there was an error somewhere when you were adding or subtracting the standard deviations from the average.

There is a way to do this exact calculation on a hand held scientific calculator (which is the way I learned it) but obviously this is so much easier.

[harness2008]

Linear regression would produce an equation in the form y = a + bx

y, the predicted value represents 1/4 mile time

x, the predictor variable represents final time

The main purpose of a linear regression model is to determine what the proper values are of a and b

With a scientific calculator you would be inputting 3 sets of data points for the first 1/4 mile figures.

113.162, 26.696
114.889, 27.644
116.616, 28.592

Each data point contains two numbers, the first one would be the final time and the second one would be 1/4 mile time. The easiest way to show you how I get the actual equation is to go to this link which is a Linear Regression calculator.

http://people.hofstra.edu/Stefan_Waner/RealWorld/newgraph/regressionframes.html

Now input the data points that I listed above and click on the radio button just beneath it that says y = mx + b to get your answer. The equation is automatically done for you. The Regression coefficient in all cases should always be 1 or the initial values inputted are not correct. If it is less than 1, then there was an error somewhere when you were adding or subtracting the standard deviations from the average.

There is a way to do this exact calculation on a hand held scientific calculator (which is the way I learned it) but obviously this is so much easier.


The problem that I see with using the half mile times in races (first half, last half, middle half) is that any moves by a horse would therefore have to sustained be in consecutive quarters by this method.

Horses generally have 2 moves in a race but not all of these moves are in consecutive quarters. By dealing with just the individual quarters, I am able to see where a horse actually moves within a race and to see if the move itself was done in a swift fraction.

A good quarter move anywhere on the track could get swallowed up if followed by a tepid fraction which could make it merely an average fraction overall utilizing half mile times. It's just my opinion but I prefer to use 1/4 mile times.

[am1947]

H2008

Thanks for the site and the detailed explanations.
Very interesting/different concept.
Thanks again for sharing ...

Regards
AM

[Hanover1]

Wouldn't you prefer a 27 29 29 29 on a half mile track and 29 29 29 27 on a mile?
I will have to check my tickets............But in the spirit of the thread I would prefer the 1/2 scenario-tells me horse left, made the top, put 'em to sleep, and jogged........

[Ray2000]

Thanks for the replies to this thread.

To test out some of the ideas discussed here, I collected data on 34412 Races, using 25 of the largest tracks in North America for Jan 1 to Oct 31, 2009. For each race I gave each horse a "Speed Rating" by adding his final time to a "Pace Time". The Pace Time was one of 7 different types as shown. The Speed Rating was an average of the most recent 3 good races (no qualifiers or pp lines w/break). All times were adjusted for Track Speed rating, racing wide in turns, and DTV. 1stQ times were adjusted for trailers but not for outside posts.

No other handicapping factors were used, strictly wagering on the horse with the best speed rating.

Results for betting $2 to Win ($68,824 total in) by using these 7 methods are:


FinalTime + BestQ...........$61,730 -10.3%
FinalTime + BestHalf........$61,354 -10.9%
FinalTime + LastHalf........$61,186 -11.1%
FinalTime + MidHalf.........$60,594 -12.0%
FinalTime + 4thQ............$59,887 -13.0%
FinalTime + 1stQ............$57,777 -16.1%
FinalTime + FirstHalf.......$57,668 -16.2%


I believe the results are significant enough to show I was using the WRONG pace method..FinalTime + 4thQ :bang: :bang:

[Stick]

Maybe the problem is in the average of the last three speed ratings. Except for horses that are dropping and horses that have an excuse because of a very poor trip or helpless spot because of post and pace, the last race is much more significant than the third race back. Horses that all of a sudden show a powerful quarter early or a fast close late in their last start are much more promising. Think about it this way. Say we have a horse that has had a good post and is racing at the same class in all of its last 3 starts. In 2 of the races the horse sat rail and made no significant move and finished in the middle of the pack. In the other start the horse raced 1st over into fast fractions and finished second. Would the horse have a better chance if the 1st over effort was the third race back or if it was its last race? I would strongly prefer the latter as long as those other 2 races don't have any excuses, and would rate this horse higher in today's race as opposed to an average of the last three.

[harness2008]

Kudos to you for your research. I am not at all surprised by your findings especially seeing your top 2 best pace factors + final time which showed the best ROI figures, albeit negative.

In essence, what is being done with the top pace figure ( best 1/4 + final time). is the fact that a fraction in the running line is being doubly weighted in the computation. Also the second best pace figure (best 1/2 + final time) is also being doubly weighted in the computation.

Just echoing what I stated previously in this thread and proven in the most recent posting, is that the faster fractions in some way need to be weighted higher than the slower fractions.

[Stick]

Harness

I agree. But I like the horse to be making a move in this fast quarter. 2 horses dead heat. The fractions are 28 55 125 154. Horse A is on the lead and decides to keep it when another horse tries to make a brush in the 27 second quarter. Horse B is 3rd on the rail and closes in the lane. Now horse A has a faster 1st q and horse B has a better last q. They both show a 27 second q, but horse A is doing all the work. If we just rate these horses off of final time and dominant quarter, it does not show how much better horse A has raced.

[harness2008]

Stick,

I may be missing something but here's my take on your hypothetical race. I don't necessarily agree with the way this is computed but Horse A does indeed come up with the better rating. Here is the math according to the method, final time + best half.

Horse A

28-0 28.0
55.0 27.0
85.0 30.0
114-0 29.0

114.0(final time) + 55.0(2 best fractions)


Horse B (assuming 3rd on rail, 2 lengths off the pace)

28-4 28.4
55-4 27.0
85-4 30.0
114-0 28.6

114.0(final time) + 55.4(2 best fractions)

[harness2008]

I do see your point on the dominant quarter and final time. That would not work in this scenario.

[Stick]

Ok My fault.

Fractions 28 55 125 154

Horse A 28 27 30 29

Horse B 29 27 30 28

Same situation as my previous post.

Now the horses are rated the same. See the problem?

[harness2008]

Ok My fault.

Fractions 28 55 125 154

Horse A 28 27 30 29

Horse B 29 27 30 28

Same situation as my previous post.

Now the horses are rated the same. See the problem?


Absolutely! I do see the problem with this and both horses handicapped this way would appear equal.

However what I am recommending is that the fractional quarters be weighted differently depending on how fast the pace segments of the race actually were.

In rudimentary form, the hierarchy of pace segments would be this according to Horse A being on the lead the entire race.

2nd qrt - fastest
1st qrt - fast
4th qrt - slow
3rd qrt - very slow

In this way the weights would be tilted above average in the first and second fractions while the weights would be lesser than average for the 3rd and 4th fraction.

Horse A - best fractions were in the fastest 2 pace segments of the race

Horse B - 1 pace segment in a fast quarter, the other in a slow quarter

Horse A would still get the nod in my opinion.

[Stick]

I understand. I thought both horses would get rated off of their best quarter or two quarters. This makes much more sense.

[harness2008]

I understand. I thought both horses would get rated off of their best quarter or two quarters. This makes much more sense.

You are right in the understanding that both horses be rated off of their final time in addition to their top 2 pace numbers.

However, I believe there is more to it than that. In my opinion it matters greatly whether those top fractions run by a horse were actually run in a faster portion of a race or a slower portion of a race. A 28 second final qrt run in an actual 28 second final qrt carries more weight than when it's run in an actual 30 second last qrt.