Making your own speed figs? It's multi-dimensional now or bust!

[Capper Al]

It's always nice to see someone asking about how they should go about making their own speed figures. Usually they are starting their journey of building their own handicapping system. Even if it doesn't pan out, it's an enjoyable venture for those so inclined. But don't think you'll beat Beyer or BRIS speed figures head to head. You don't have the resources to do so. And even if your figures are 1% better, it probably will be hard to break even. Nowadays, one needs an arsenal of many types of speed figures, plus many class figs, and many pace figs, and connections figs. A single fig rarely gets it. Why? I guessing here, but if you look at a graph of speed figs, you'll find it looks more like a scatter graph. What BRIS might have as the fastest, Beyer might show as the slowest. Usually one finds that the top speed horses are in the top half for both, but not in the exact same order. It's amazing that they generally all fall in with a strike rate of about 25 to 30 percent at the end of a test period. So why multiple figures? Does averaging soothe the beast? I think so. With needing a higher degree of certainty because of the low payouts on top speed horses it might help. When your horse doesn't win, it might be easier to stomach if you tell your friends that it had both the best Beyer and BRIS fig. How else could you have picked the winner?

[thaskalos]

Originally Posted by Capper Al

It's always nice to see someone asking about how they should go about making their own speed figures. Usually they are starting their journey of building their own handicapping system. Even if it doesn't pan out, it's an enjoyable venture for those so inclined. But don't think you'll beat Beyer or BRIS speed figures head to head. You don't have the resources to do so. And even if your figures are 1% better, it probably will be hard to break even. Nowadays, one needs an arsenal of many types of speed figures, plus many class figs, and many pace figs, and connections figs. A single fig rarely gets it. Why? I guessing here, but if you look at a graph of speed figs, you'll find it looks more like a scatter graph. What BRIS might have as the fastest, Beyer might show as the slowest. Usually one finds that the top speed horses are in the top half for both, but not in the exact same order. It's amazing that they generally all fall in with a strike rate of about 25 to 30 percent at the end of a test period. So why multiple figures? Does averaging soothe the beast? I think so. With needing a higher degree of certainty because of the low payouts on top speed horses it might help. When your horse doesn't win, it might be easier to stomach if you tell your friends that it had both the best Beyer and BRIS fig. How else could you have picked the winner?

I disagree with you here, Al. It's VERY worthwhile to make your own figures...and you CAN beat the Beyer and the BRIS figures...simply because the Beyer figures exclude pace, and the BRIS pace figures are laughable.

And, no...you don't need an "arsenal of many types of speed figures", nor "many pace figures" to get the job done; one of each will do nicely. The trick is in the INTERPRETATION of these figures.

[classhandicapper]

As someone that has spent a lot of time looking at multiple sets of figures in the hope of gaining extra confidence in some races and perhaps figuring out who had it right before the fact in other races, IMO that strategy has both upsides and downsides.

It certainly helps point out some controversial races and days. So you may make fewer bets off bad figures. But it can sometimes be paralyzing when you aren't sure what really happened or which figure is correct.

[GaryG]

I have made my own Quirin-style figs for years. I not only have confidence in them but I know what went into them and that makes it possible to fine tune them after the fact. I look at the speed figure in conjunction with the pace figure rather than trying to modify one with the other. I also make a race shape figure that uses both bias and pace. This is similar to what the Plod Boys do at Racing Flow. As someone said long ago, "I don't intend to lose my money on somebody else's typographical error."

[DeltaLover]

Originally Posted by Capper Al

It's always nice to see someone asking about how they should go about making their own speed figures. Usually they are starting their journey of building their own handicapping system. Even if it doesn't pan out, it's an enjoyable venture for those so inclined. But don't think you'll beat Beyer or BRIS speed figures head to head. You don't have the resources to do so. And even if your figures are 1% better, it probably will be hard to break even. Nowadays, one needs an arsenal of many types of speed figures, plus many class figs, and many pace figs, and connections figs. A single fig rarely gets it. Why? I guessing here, but if you look at a graph of speed figs, you'll find it looks more like a scatter graph. What BRIS might have as the fastest, Beyer might show as the slowest. Usually one finds that the top speed horses are in the top half for both, but not in the exact same order. It's amazing that they generally all fall in with a strike rate of about 25 to 30 percent at the end of a test period. So why multiple figures? Does averaging soothe the beast? I think so. With needing a higher degree of certainty because of the low payouts on top speed horses it might help. When your horse doesn't win, it might be easier to stomach if you tell your friends that it had both the best Beyer and BRIS fig. How else could you have picked the winner?

Nice post..

I completely agree that we need several other metrics appart from pure speed, pace or performance figs to measure aspects of the race that are not so obvious to the average horse bettors.

As far as your horse does not winning the race, I have to say that this has to nothing to do with the validity of your figures and metrics though

[clemkadiddle]

BRIS or Beyer are based on the same flawed theory of parallel time. I use them...but only to reverse-engineer the final time of the race where the 100-point score was awarded and compare that to standard parallel time charts to calculate the variant...which I do as a ratio...so it can be used when parsing the segments of the race into pace calculations.

Parallel time is flawed because it assumes the base 10 logarithm of the average feet per second differs by .0064 for each furlong. For each additional turn, add .0040.

The biggest mistake that practically everyone (but me) makes is to add 1/5 of a second for each length to the time at the fraction at the point of call. Many don't realize that the points of call don't necessarily correspond to the fractional times and the lengths need to be pro-rated to synchronize them. So too is the fact that handicappers try to make their calculations from point to point in the race, but fail to see that this method is lost in the averages from the preceding points in the race. The best approach is to cut the race into segments and analyze the speeding up and slowing down of the horse...even then, one fails to see that the average feet per second within each segment are just that...averages. They don't visualize just how fast the horse was going at the beginning of the segment and again at the end.

The only method that has any value is one that can truly estimate the amount of "work" a horse performs during the course of the race. As a horse runs faster, he expends energy exponentially; as he begins to slow down, he is still expending energy but the mode changes into more of a coasting.

After about 8 years of research I do believe I have that model. After all, if Beyer, BRIS, Brohammer, Sartin...or any of these approaches that are available were really worth anything...we would all be rich. That's why I didn't bother with this stuff.

Me: Computer programmer/analyst for over 25 years. Been a race fan for almost 40. It was only about 8 years ago that I really thought that I would take a shot at this in my spare time...and it has been paying off since last June. (However, I really need to spend more time and fine tune some things but basically I am staying ahead of the game.)

HINT: A horse really doesn't start spending racing energy until he crosses the 44 FPS threshold...the "2 minute lick".

[thaskalos]

Originally Posted by clemkadiddle

BRIS or Beyer are based on the same flawed theory of parallel time. I use them...but only to reverse-engineer the final time of the race where the 100-point score was awarded and compare that to standard parallel time charts to calculate the variant...which I do as a ratio...so it can be used when parsing the segments of the race into pace calculations.

Parallel time is flawed because it assumes the base 10 logarithm of the average feet per second differs by .0064 for each furlong. For each additional turn, add .0040.

The biggest mistake that practically everyone (but me) makes is to add 1/5 of a second for each length to the time at the fraction at the point of call. Many don't realize that the points of call don't necessarily correspond to the fractional times and the lengths need to be pro-rated to synchronize them. So too is the fact that handicappers try to make their calculations from point to point in the race, but fail to see that this method is lost in the averages from the preceding points in the race. The best approach is to cut the race into segments and analyze the speeding up and slowing down of the horse...even then, one fails to see that the average feet per second within each segment are just that...averages. They don't visualize just how fast the horse was going at the beginning of the segment and again at the end.

The only method that has any value is one that can truly estimate the amount of "work" a horse performs during the course of the race. As a horse runs faster, he expends energy exponentially; as he begins to slow down, he is still expending energy but the mode changes into more of a coasting.

After about 8 years of research I do believe I have that model. After all, if Beyer, BRIS, Brohammer, Sartin...or any of these approaches that are available were really worth anything...we would all be rich. That's why I didn't bother with this stuff.

Me: Computer programmer/analyst for over 25 years. Been a race fan for almost 40. It was only about 8 years ago that I really thought that I would take a shot at this in my spare time...and it has been paying off since last June. (However, I really need to spend more time and fine tune some things but basically I am staying ahead of the game.)

HINT: A horse really doesn't start spending racing energy until he crosses the 44 FPS threshold...the "2 minute lick".

Is this a joke? Is there anyone out there who still adds a fifth of a second for a beaten length?

[Capper Al]

Originally Posted by classhandicapper

As someone that has spent a lot of time looking at multiple sets of figures in the hope of gaining extra confidence in some races and perhaps figuring out who had it right before the fact in other races, IMO that strategy has both upsides and downsides.

It certainly helps point out some controversial races and days. So you may make fewer bets off bad figures. But it can sometimes be paralyzing when you aren't sure what really happened or which figure is correct.

I hear you. One of the biggest problems with multiple numbers is paralyze from analyze as it is commonly referred to. Ainslie once told a joke of a man that drowned crossing a river when someone on the other side yelled, the average depth is only 4 feet. Averaging numbers can be dangerous also. I know no other solution to this than to run a few hundred races and see what numbers or number combinations works best for you. And then it's won't be the Holy Grail of racing either, just another number.

[Capper Al]

Originally Posted by GaryG

I have made my own Quirin-style figs for years. I not only have confidence in them but I know what went into them and that makes it possible to fine tune them after the fact. I look at the speed figure in conjunction with the pace figure rather than trying to modify one with the other. I also make a race shape figure that uses both bias and pace. This is similar to what the Plod Boys do at Racing Flow. As someone said long ago, "I don't intend to lose my money on somebody else's typographical error."

I didn't know of Plod Boys at Racing Flow, but running and tweaking Quirin-figs in an age of Beyer style parallel times would be an interesting study. They Quirin-figs might not have a higher win rate, but might have a better ROI since the crowd isn't using them anymore. Might be another number to add to my collection.

[Capper Al]

Originally Posted by DeltaLover

Nice post..

I completely agree that we need several other metrics appart from pure speed, pace or performance figs to measure aspects of the race that are not so obvious to the average horse bettors.

As far as your horse does not winning the race, I have to say that this has to nothing to do with the validity of your figures and metrics though

I use multiple metrics also, but must say my longest winning streak for this weekend worrier was 13 weekends using Gile style pace figs. Then all of a sudden, the numbers died- fortunately before I bought a Corvette. Nothing else that I ever did came close to this single factor play.

[Capper Al]

Originally Posted by clemkadiddle

BRIS or Beyer are based on the same flawed theory of parallel time. I use them...but only to reverse-engineer the final time of the race where the 100-point score was awarded and compare that to standard parallel time charts to calculate the variant...which I do as a ratio...so it can be used when parsing the segments of the race into pace calculations.

Parallel time is flawed because it assumes the base 10 logarithm of the average feet per second differs by .0064 for each furlong. For each additional turn, add .0040.

The biggest mistake that practically everyone (but me) makes is to add 1/5 of a second for each length to the time at the fraction at the point of call. Many don't realize that the points of call don't necessarily correspond to the fractional times and the lengths need to be pro-rated to synchronize them. So too is the fact that handicappers try to make their calculations from point to point in the race, but fail to see that this method is lost in the averages from the preceding points in the race. The best approach is to cut the race into segments and analyze the speeding up and slowing down of the horse...even then, one fails to see that the average feet per second within each segment are just that...averages. They don't visualize just how fast the horse was going at the beginning of the segment and again at the end.

The only method that has any value is one that can truly estimate the amount of "work" a horse performs during the course of the race. As a horse runs faster, he expends energy exponentially; as he begins to slow down, he is still expending energy but the mode changes into more of a coasting.

After about 8 years of research I do believe I have that model. After all, if Beyer, BRIS, Brohammer, Sartin...or any of these approaches that are available were really worth anything...we would all be rich. That's why I didn't bother with this stuff.

Me: Computer programmer/analyst for over 25 years. Been a race fan for almost 40. It was only about 8 years ago that I really thought that I would take a shot at this in my spare time...and it has been paying off since last June. (However, I really need to spend more time and fine tune some things but basically I am staying ahead of the game.)

HINT: A horse really doesn't start spending racing energy until he crosses the 44 FPS threshold...the "2 minute lick".

Nice post, my fellow computer programmer/analyst.

Parallel time isn't flawed. Parallel time still is the most published speed and the best single handicapping factor for a win percentage. I'm not disagreeing with making your figs. One might just get a higher ROI with numbers that the crowd doesn't use. My recent figs are feet per second (FTS) based also.

[Capper Al]

Originally Posted by thaskalos

Is this a joke? Is there anyone out there who still adds a fifth of a second for a beaten length?

I do! Why? A length isn't 1/5 of a second. Agree. Because after playing with parallel time it dawn on me that the final unit of displacement doesn't matter as long as it is a CONSTANT! The variability of time should be rolled into the base figure that is adjusted to a focal point of something usually like 100. Once adjusted the whole idea of parallel time is to sync these different speeds to a universal point system. It won't be a universal time if the lengths off are variable. Whatever a displacement a capper uses be it 1/5 point per length or 10 per length, it's a measure off the universal time. Lengths are universal. At that point the unit of measure has been converted from seconds to points. Why not use 1/5 per length? It is more user friendly than any other fraction.

[fmolf]

Originally Posted by Capper Al

I do! Why? A length isn't 1/5 of a second. Agree. Because after playing with parallel time it dawn on me that the final unit of displacement doesn't matter as long as it is a CONSTANT! The variability of time should be rolled into the base figure that is adjusted to a focal point of something usually like 100. Once adjusted the whole idea of parallel time is to sync these different speeds to a universal point system. It won't be a universal time if the lengths off are variable. Whatever a displacement a capper uses be it 1/5 point per length or 10 per length, it's a measure off the universal time. Lengths are universal. At that point the unit of measure has been converted from seconds to points. Why not use 1/5 per length? It is more user friendly than any other fraction.

Very well said.....In this way every horse in every race is judged against the same standard. Simple!...I am a firm believer in K.I.S.S methods.

[Capper Al]

Originally Posted by fmolf

Very well said.....In this way every horse in every race is judged against the same standard. Simple!...I am a firm believer in K.I.S.S methods.

I just checked my code on this. My constant is 1 point per length, not 1/5. Old age, I can't remember anything anymore. But note,it doesn't vary on the back end of the equation.

[thaskalos]

Originally Posted by Capper Al

I do! Why? A length isn't 1/5 of a second. Agree. Because after playing with parallel time it dawn on me that the final unit of displacement doesn't matter as long as it is a CONSTANT! The variability of time should be rolled into the base figure that is adjusted to a focal point of something usually like 100. Once adjusted the whole idea of parallel time is to sync these different speeds to a universal point system. It won't be a universal time if the lengths off are variable. Whatever a displacement a capper uses be it 1/5 point per length or 10 per length, it's a measure off the universal time. Lengths are universal. At that point the unit of measure has been converted from seconds to points. Why not use 1/5 per length? It is more user friendly than any other fraction.

Clem made the astounding assertion that everyone but him adds a fifth of a second to the horse's time for every length that the horse is behind the leader at the various points of call. That simply isn't true. A great many handicappers have realized that a length isn't worth a fifth of a second...and the pace handicappers have largely abandoned this "convenient" adjustment measure...because inaccurate adjustment measures DO make a difference...even when the measurement is a CONSTANT, and is applied to ALL the horses equally.

In this game...the pace handicapper is comparing speed horses to closers. If a closer gains 5 lengths during the stretch run of a given race, and this closer is credited a full second for his stretch gain...then the stretch runner has gotten an adjustment advantage that he doesn't deserve, relative to the speed horse that he is trying to catch. When a pace handicapper is comparing speed horses to closers...then a fifth of a second "mistake" during a fraction of the race makes a difference.

I have no real argument to make against the notion that simplicity is worth a small measure of inaccuracy...nor can I really argue with the guy who states that precision is an impossibility in this game, since even the employed method for reported lengths behind in this game is fraught with inaccuracies. The horseplayer makes his own adjustment decisions...and he lives with them. But Clem's post seemed to me to be decidedly smug and egotistical. EVERYONE does "this", but Clem does "that"...because he is a 25-year programmer, and a 40-year horseplayer. That's an odd way to talk...especially when you don't really know what everybody else is doing.