Speed figures - expressing levels of confidence

[Saratoga_Mike]

For those who make their own speed figures, does anyone attach an expression of their level of confidence to the speed figure (even qualitative - e.g., high, medium, low)?

For example, let's assume a track runs six sprints on a given day and the figure maker calculates a variant of +4 for every race (would rarely happen in the real world, I concede). The figure maker's level of confidence in the speed figures for those races would be very high. I guess the level of confidence could even be expressed mathematically (maybe using a standard deviation on the variant?). A figure maker's levels of confidence would be low if the variants were widely dispersed or the sample size was low (e.g., only one turf race was run on a given day).

Given the sample sizes for any day are so small, I'm not sure a mathematical expression would be of much value, but it seems like applying a qualitative expression of confidence (high, medium, low) might be of some value - just not sure how to get there.

Here's where I see value in this approach: there are two speed figure standouts in a race, little separates them (similar speed and class). What might separate them is differences in the figure maker's level of confidence in the past couple of figures.

Thoughts?

[Dan Montilion]

When I have done figures, I include a rather cumbersome "use caution".

[cj]

We do this at TFUS. Not necessarily a level, but we make note of any figure that was either tough to make (lack of data), broken out from other races, or has horses getting figures that look suspiciously high or low based on past races. Horses can be noted for more than one of the reasons above as well, a double ?.

[dasch]

I dont add a number for the level of confidence but I do add an "mw" (matched well) when most of the horses match, or a "mew"(matched extremely well) for races where all/almost all match. It makes it much easier to settle on a number in difficult races with limited info.

[sjk]

I look at the standard deviation of the variant data but mainly to catch data errors (which are pretty rare these days) and to see if there might be a need to split the card (I almost never do).

When a horse runs back and I need to use the speed figure I adjust the figure that comes from the daily variant using additional information specific to the race he ran in.

For all of the horses that ran in that race I calculate the difference between the speed rating as determined by the variant and what I would have projected that horse to run based on his adjusted speed rating in the last previous race and next subsequent race. I move the speed figure part of the way in the direction indicated by these differences.

This means that speed figures are not static. That is not a problem for me; I update them every few weeks anyway to incorporate subsequent events.

[JohnGalt1]

Speed figures from a recent race is more representative IMO, than a figure from a horse's line after a lay off.

To differentiate I'll write down the pace figures from a race after a month or more layoff in parentheses, which labels them as advisory.

Then, even if the lay off horse's best race is faster than the horse with current races, I ask can it beat the it? How are the work outs? Is the trainer good after lay offs? etc.

Another example of me not fully trusting my figures is if a race has an extremely high or low variant. I may use another pace line to confirm the number.

[Saratoga_Mike]

We do this at TFUS. Not necessarily a level, but we make note of any figure that was either tough to make (lack of data), broken out from other races, or has horses getting figures that look suspiciously high or low based on past races. Horses can be noted for more than one of the reasons above as well, a double ?.

Is this info published somewhere for your customers?

[cj]

Is this info published somewhere for your customers?

Not yet, but I believe that is coming.

[classhandicapper]

I look at the standard deviation of the variant data

That's a pretty good idea.