Predicting Speed Rating from Past Ratings

[steveb]

Originally Posted by TrifectaMike

DL, even if there is a measurable difference, what would you do with it?

Mike

once you have the average differences, then you have a great measuring stick.
you then can compare expected with actual, and much more
i have logit model with about 60 factors, and easily the 6 most important ones i have, are all derived from knowing these differences.
that is for australia though, but doubt where would make much difference.

[raybo]

Originally Posted by steveb

once you have the average differences, then you have a great measuring stick.
you then can compare expected with actual, and much more
i have logit model with about 60 factors, and easily the 6 most important ones i have, are all derived from knowing these differences.
that is for australia though, but doubt where would make much difference.

Not wanting to sound, or appear, pompous or defensive, but do you really know the way classes in the US are carded? Do you really know the differences between the way classes are carded in your country and in other non-US countries, compared to the way they are carded in the US?

[steveb]

Originally Posted by raybo

Not wanting to sound, or appear, pompous or defensive, but do you really know the way classes in the US are carded? Do you really know the differences between the way classes are carded in your country and in other non-US countries, compared to the way they are carded in the US?

if one studies racing, then I guess where it is held is no big deal.
i am pretty sure i would have no problems coming to terms with how it may be done in the states, if that is what i wanted.
australia has 6 states and several territories, and each of those places has many different standards of racing.
if i can do that, i don't imagine it would be too much harder in the states.
queensland may have 6 or more meetings of wildly varying standard on any saturday.
for all i know australia may be more variable than the states?

places like hk, singapore, japan and korea and others are all relatively easy because they don't have the same difficulties that there is in australia.

anyway, you can appear as pompous or as defensive as you like, because it will not make any difference either way.
i am pretty sure i know where i stand in this caper, especially where time is concerned.
hope i did not sound pompous there!

[raybo]

Originally Posted by steveb

if one studies racing, then I guess where it is held is no big deal.
i am pretty sure i would have no problems coming to terms with how it may be done in the states, if that is what i wanted.
australia has 6 states and several territories, and each of those places has many different standards of racing.
if i can do that, i don't imagine it would be too much harder in the states.
queensland may have 6 or more meetings of wildly varying standard on any saturday.
for all i know australia may be more variable than the states?

places like hk, singapore, japan and korea and others are all relatively easy because they don't have the same difficulties that there is in australia.

anyway, you can appear as pompous or as defensive as you like, because it will not make any difference either way.
i am pretty sure i know where i stand in this caper, especially where time is concerned.
hope i did not sound pompous there!

Well stated, and pretty close to what I thought. Thank you for your admission of not being familiar with the "class" system here in the US, sometimes (many times?) that degree of honesty is rare here!

[sjk]

Originally Posted by DeltaLover

I cannot detect any measurable difference when it comes to so close claiming price gaps. Reading some of the postings though, I can see that the opposite view seems to prevail among many and I would like to know if those who believe so, have some data to backup their opinion or it is just something they base on their intuition?

I calculate the difference at around 3.5 points. More than 10k races at each level.

[sjk]

[QUOTE=TrifectaMike]DL, even if there is a measurable difference, what would you do with it?

Mike[/QUOTE

Having a set of pars is one of the ingredients needed to make figures; at least it is for me.

[Tom]

Originally Posted by sjk

I calculate the difference at around 3.5 points. More than 10k races at each level.

You have 10,000 4,000 and 10,000 5,000 races - open races????

[TrifectaMike]

Originally Posted by steveb

once you have the average differences, then you have a great measuring stick.
you then can compare expected with actual, and much more
i have logit model with about 60 factors, and easily the 6 most important ones i have, are all derived from knowing these differences.
that is for australia though, but doubt where would make much difference.

Here , in the states, as useless as balls on a chick. That "info" and a stolen metro card "may" get you on a train.

Mike

[sjk]

Originally Posted by Tom

You have 10,000 4,000 and 10,000 5,000 races - open races????

I tried to select all races other than lifetime conditions. There are lots of subconditions (R4000NW16M, etc) but the mix should be similar at the two prices.

[TrifectaMike]

Originally Posted by clemkadiddle

10 squared or said differently "10 to the 2nd power" is 100.

The base 10 logarithm of 10 = 2 (=1). That's how logarithms work.

In getting back to my explanation, the "100 point score" for a 6 furlong race on a "standard parallel time chart" occurs at 70.32 seconds (1:10.32) which is 56.31 FPS.

The logarithm of 56.31 is 1.7506.

Subtract .0064 from 1.7506 to obtain 1.7442. Calculate 10 to the power of 1.7442 to obtain 55.49 FPS. This translates to 83.26 seconds (1:23.26) at 7 furlongs.

The theory of parallel time would state that a 6 furlong race in 70.32 seconds is equivalent to a 7 furlong race in 83.26 seconds.

If you want to see the equivalent at a 1-turn mile, subtract .0064 from 1.7442 to obtain 1.7378. Raise 10 to the 1.7378 power to obtain 54.68 FPS. This translates to 8 furlongs in 96.57 seconds (1:36.57).

To equate races around additional turns, subtract .0040 from each logarithm for each ADDITIONAL turn. For instance, a mile ran on a bull ring would negotiate 3 turns so subtract .0080.

This is my gift to you guys. The actual formula for the "work coefficient" is still going to be my secret.

I agree a power law is better as you suggested, if you want to work with point estimates.

Mike

[DeltaLover]

Originally Posted by sjk

I calculate the difference at around 3.5 points. More than 10k races at each level.

Not a sufficient answer IMHO! Some things that might clarify what you are saying here are the following:

What exactly do you mean by points?

How you compare? You just use averages or you do something else?

How clear are the conditions you are using? For example how you handle state breds, male vs females, age restrictions? How do you handle weight changes?

Do you only use winner's time or you add in the mix runner ups (or more starters)?

Can you provide some data proving your methodology?

Again my NULL hypothesis, is that there is NO measurable difference between the 4K and 5K claiming races, please explain how you can prove it wrong?

[sjk]

By winning money from people who believe in your null hypothesis

[Tom]

That would do'er!

[classhandicapper]

Originally Posted by TrifectaMike

DL, even if there is a measurable difference, what would you do with it?

Mike

This issue is style of handicapping.

Assume all else is equal.

If one horse ran a 60 final time number and another ran a 59 final time number, the pure final time handicapper will say the 60 ran a better race.

If one horse won a 4K race and another won a 5K race the class handicapper will say the latter ran the better race.

If the horse that ran a 59 did it in a 5K race and the horse that ran the 60 did it in a 4K race, a hybrid handicapper might say the 59 is better because it was earned against slightly better competition.

If the horse that ran against 4k ran a 63 and the horse that ran against 5k ran a 57, the hyrbrid handicapper might prefer the the 4k 63 because even though it was accomplished against cheaper, the larger gap in speed overwhelms the small gap in quality of competition.

Naturally, odds come into these things. But the thinking changes based on the philosophy of the handicapper. The philosophy on how to handicap varies because there are errors in final time figures and inefficiencies in the classing system. Neither method is perfect, but both have easy to identify strengths and weaknesses that I could elaborate on if you'd like. You pick your poison or try to combine them (which is a different brew of poison).

[classhandicapper]

Originally Posted by JohnGalt1

Which is why I now make and use Performance Class Ratings. Finish position, amount of horses faced and class adjusted.

I like to think it is objective as I adjust the final rating, but much of it is my subjective decisions.

From reading all your comments on class handicapping, it's obvious you "get it".