Extreme Pace Scenaios

[Tom]

Gus (Thaskalos) wrote in another thread about pace....

If you don't want to make your own figures...then I would suggest buying the DRF instead...and using the half-mile race time as the primary
component of the calculation

This is something I do to help me read charts and get an idea of what races might have been faster or slower than normal.

The idea uses Randy Giles' pace projection.
I'll use Aqueduct's inner dirt track 6.0 furlongs as an example, as I have lots of races to use and they will switching to the inner soon, where 6 is the dominant sprint distance.

Convert the final times to decimals, so that 1.12 becomes 72.0.
divide the final time by the pace time (4 furlongs). So 46.3 72.8 give you .636. Now, do this for all the races you have to use and calculate the average and the standard deviation. This is the race times, no beaten lengths.

For this distance, the average is .649 and the St Dev is .008. Half of that is .004. These statistics come from 3447 races over since 2005.

My assumption here is that races with a pace fraction within + or - 1 standard deviation will be average, and those outside +1 or -1 will be fast or slow.
To verify this, I looked at the beaten lengths to see if there was an impact on them by the race shape. Eventually, I added +1/2 and - 1/2 as significant borders for pace shapes.

Here is the story on Aqu Inner 6.0 iDirt races -

[Tom]

I now have race shapes of average, fast, very fast, slow, or very slow.
The results for very fast and very slow are shown below.
As you can see, the number for all races and only average races are very close, almost no differences.

When the early pace is very fast, the number of wire to wire winners decreases, and the number of closer's, 4 or more lengths behind, skyrockets to 25%.

When the early pace is very slow, the number of wire to wire winners increases, and the number of closer's plummets to 2%!

Note too, that when the pace is fast, horse running less than 2 lengths behind are also impacted negatively, with their worst performance of all race shapes.
I think this situation was commented on by CJ in an article he recently wrote about pace for Hana's magazine?

Here is the breakdown for the race shapes, using the pace of race, leader's raw times.

[Tom]

How can this help?

On 1/30/13 and again on 3/10/14, and 3/24/14 one race on the day was noticeably faster than the others, and closer's won all three of them.

On 3/10/14, all the races were slow, so nothing extreme about most of them.
the one race was less slow than the rest, a +5, so I would look at that one a little closer.

This give me an idea of how the races were running on any given day. A lot of days are nothing special, and some are just random, but sometimes, a race or two will stick out.

[Dave Schwartz]

Interesting work, Tom.

[Some_One]

What was the expected win percentages for each of the splits? Any indication the public could pick up on this.

[classhandicapper]

Tom,

Years ago I did something along these lines with the pace figures I was making for NY, but what you are doing is a few steps past that. This is great work.

I think it was around 1990 I was armed with an IBM AT clone, Dbase 3+ , and about a year of data.

I manually created a database with the standard info of date, race, class, distance, track condition, field size, fractions, my pace figures, my final time figure, and position and lengths ahead/behind of winner (maybe a few other fields). Then I ran the same kinds of queries using my pace figures to get at "very fast", "fast", "average" etc... and see where the winners were coming from at each distance. That's how I learned that field size was an independent variable in pace and that horses that were racing off an extremely fast pace were also being impacted negatively.

The one nifty thing I had was the ability to do "first 1/4 slow - 2nd quarter fast" and various combinations like that. I commented on some of that earlier in the thread.

What you are doing is a great learning tool and you appear to be doing a great job of it.

[Exotic1]

Thanks Tom - for the time and product.

Much appreciated.

[Fingal]

Originally Posted by Dave Schwartz

Interesting work, Tom.

Yes, just the concept alone is something I'm going to go over a few old race cards with & see what comes up.

[MJC922]

Originally Posted by Tom

Gus (Thaskalos) wrote in another thread about pace....



This is something I do to help me read charts and get an idea of what races might have been faster or slower than normal.

The idea uses Randy Giles' pace projection.
I'll use Aqueduct's inner dirt track 6.0 furlongs as an example, as I have lots of races to use and they will switching to the inner soon, where 6 is the dominant sprint distance.

Convert the final times to decimals, so that 1.12 becomes 72.0.
divide the final time by the pace time (4 furlongs). So 46.3 72.8 give you .636. Now, do this for all the races you have to use and calculate the average and the standard deviation. This is the race times, no beaten lengths.

For this distance, the average is .649 and the St Dev is .008. Half of that is .004. These statistics come from 3447 races over since 2005.

My assumption here is that races with a pace fraction within + or - 1 standard deviation will be average, and those outside +1 or -1 will be fast or slow.
To verify this, I looked at the beaten lengths to see if there was an impact on them by the race shape. Eventually, I added +1/2 and - 1/2 as significant borders for pace shapes.

Here is the story on Aqu Inner 6.0 iDirt races -

I know Randy from online way back, great guy. That ratio however isn't linear. It doesn't mean it can't be somewhat useful to ballpark however the further away from the average time it gets the greater amount of error is being introduced -- and when tenths matter not a lot of error is acceptable IMO.

[Tom]

Originally Posted by MJC922

I know Randy from online way back, great guy. That ratio however isn't linear. It doesn't mean it can't be somewhat useful to ballpark however the further away from the average time it gets the greater amount of error is being introduced -- and when tenths matter not a lot of error is acceptable IMO.

If I take the data and group it, here is what I get.
The difference between 69.0 and 74.0 is .010, while the +/- 1/2 StDev is .008, so it is close.

I took that data and did a scatter diagram and got an R=1.0 value for the trend line, so I will try revising the multiplier by final time and compare it to the same data set and see what shakes out. So far, I am getting some very consistent variants.

Thanks for the suggestion.

[DeltaLover]

Originally Posted by Tom

I now have race shapes of average, fast, very fast, slow, or very slow.
The results for very fast and very slow are shown below.
As you can see, the number for all races and only average races are very close, almost no differences.

When the early pace is very fast, the number of wire to wire winners decreases, and the number of closer's, 4 or more lengths behind, skyrockets to 25%.

When the early pace is very slow, the number of wire to wire winners increases, and the number of closer's plummets to 2%!

Note too, that when the pace is fast, horse running less than 2 lengths behind are also impacted negatively, with their worst performance of all race shapes.
I think this situation was commented on by CJ in an article he recently wrote about pace for Hana's magazine?

Here is the breakdown for the race shapes, using the pace of race, leader's raw times.

Excellent approach! I am doing something very similar for long time....

[cj]

This is pretty much exactly what red and blue highlighting of pace figures and fractions is in TimeformUS PPs.

[Tom]

Of course I steal ideas form the best!

[MJC922]

Originally Posted by Tom

If I take the data and group it, here is what I get.
The difference between 69.0 and 74.0 is .010, while the +/- 1/2 StDev is .008, so it is close.

I took that data and did a scatter diagram and got an R=1.0 value for the trend line, so I will try revising the multiplier by final time and compare it to the same data set and see what shakes out. So far, I am getting some very consistent variants.

Thanks for the suggestion.

No problem, revisiting my notes from 08... subtract your final time from the avg one to get a delta first and then apply the static multiplier at each call. So it's a linear relationship but should be calculated in a different way i.e. using a delta. FWIW another note that was made, for the first call baseline to be 95% accurate to within a fifth of a second at a track / dist / surface would require a 64 race sample. So for a race like the Ky Derby if someone tells you they know the fractional to final relationships within a fifth, it's not likely to be true without going back that many years, this would have to be a ballpark estimate as there aren't nearly enough races run there at 10f.

Good luck, now that I'm out of the time crunching game I figure I can give away some findings from long years of research.

[Robert Goren]

I have found out the hard way, two things. Betting a closer coming out of fast paced race is bad idea. Betting a front runner coming out of a fast paced race is not a much better idea despite the logic that says they should be contenders in a slower paced race. I don't have charts to back up those statements, just torn up tickets. In general horses coming out of fast paced races should be avoided. So should horses coming out of a slow paced race, because of a slow paced race is a sign of lack of quality horses in the race. IMO