Speed figure theories

[Valuist]

For those of you who make numbers, what do you feel are the most common mistakes made in making speed figures? IMO, some of the top ones are:

1. Not splitting up the routes and sprints. I've learned over the years that you cannot accurately match a 2 turn 1 1/16 mile race to a 6 furlong race. There are numerous reasons for this, among them wind and track maintenance.

2. Using speed charts instead of projection when there is only one race at a given distance for a race day. This is often a problem at tracks like Belmont or Arlington, which may run races at 7-8 different distances on the day. I believe these races that are the only ones run in a day at a track are often the ones most likely to be incorrect.

3. Do the gaps at the finish match the number? I recently saw a horse entered who won its two previous starts and the same horses ran second and third in both races, separated by roughly a length or so in each instance. Yet the Beyer number for the horses varied by 8 pts in each race. And the Brisnet number varied similarly. Did all three horses go down by about 8 pts? The speed charts may dictate that, but common sense would say otherwise.

4. Changing track conditions may dictate that an entire card needs to be projected. This is pretty self explanatory.

5. One of the toughest numbers to make is a projected number for a 3YO in summer or fall, who hasn't run since its 2YO year. 2YOs tend to run far slower numbers, especially routing, so the best thing to do, IMO, is project based on the 2nd and 3rd place finishers.

Any other ideas?

[cj]

Biggest error I see is not factoring extreme pace races, slow or fast, as a reason for a slow speed figure.

[Brian Flewwelling]

Valuist and CJ: I cannot add to your list, but i do recognize the need for Projections for single races at distance/surface. I only use charts and get some weird numbers for Turf because of both single races and wide variations in the Pace scenario.

What i would like to know, is how do you do the Projected speed thing... how many horses do you include? ... what determines your 'expected' speed ... last, best last 3, or ???

Can you or others give a few hints?

Fleww

[Valuist]

When I project, I'll look at the horse's last race number but also look at the horse's top back number at a similar distance. The winner and runnerup are the first two to look at. Do their last race numbers make sense if they were to run 1-2? If not, that's when I start to look at the back number. If that still doesn't come up with a viable number, its time to look at the 3rd and 4th place finishers. Most of the time, you can piece together a number distribution based on the top 2. But, I will point out, I never project numbers for 2YOs and rarely do for 3YO Mdns or Mdn claimers. If you have other races at the same distance you just figure based on those figures. If the 2YO races are the only ones at a given distance, then you may have to dig the dreaded speed chart out. But if I'm still unsure of the number I give, I'll put a question mark in the charts, to show that the number could be incorrect.

[sjk]

My numbers are about 85% based on projections. My candidate for a big mistake is to fail to recalculate variants on a regular basis projecting back from subsequent races. The information from races after the day in question is equally as valid as the information from the races before.

Of course this is most easily done if you use a computer to do the variants.

[RXB]

Originally posted by Valuist
But if I'm still unsure of the number I give, I'll put a question mark in the charts, to show that the number could be incorrect.

Bingo. It's not just the number that you assign, it's the confidence that you have in the reliability of the data from which you're working that matters just as much.

I'd also say that it would be a good idea to apply the different side of the same coin-- when the number is based on particularly sound, reliable data, you could mark the chart to indicate exceptional confidence in your assessment.

Hastings Park is one of the easiest tracks to make figures for because the majority of races are run at one distance (6.5 furlongs), and most route races are run at 8.5 furlongs. And from May to September, during the good old west coast summer climate when rainfall amounts are less than at most other tracks, it's a comparative breeze to make reliable speed and pace figures. Whereas, for the reasons noted earlier in this thread, some other tracks create more challenging conditions.

[delayjf]

I'd say the toughest nut to crack is distance equalization. How does a 7 furlong sprint compare to a 5 1/2 furlong sprint. What will a horses' figure be attempting to stretchout for the first time.

[kenwoodallpromos]

Some speed cappers may neglect to consider the non-obvious individual horse's excuses for a lower figure.

[Brian Flewwelling]

Making Speed Figs and Handicapping with them is NOT the same thing.

Speed figs are a representation of the speed a horse ran the race, and take track "speed" into account. They may take other factors related to the day and the track into account.

What they don't do is predict the speed the horse will run the next time out... that is the job of the handicapper. Excusing a slow speed doesn't make the speed higher:D

the pace at which the race ran is just that, the pace, not the speed.

Valuist: you are saying that Projections are a Human thing, not easily programmed!
Is that the concensus, or do some folks have an algorithm for projecting the speed of a race? and would they like to share it?

Fleww

[JackS]

I'm not very trusting of the TV and don't think others should be either. Unless the handicapper is making his own and is confident of the numbers he is producing I'd suggest ignoring the printed varient except when it is at an extreme up or down.
If possible, the handicapper should attempt to use a line or lines that appear reasonable and maybe within 3-4 points of average.
If forced to use a line with one of these super fast or slow TV's, a best guess adjustment should probably be attempted. My feelings anyway.

[kenwoodallpromos]

I will amend my answer to what the 'capper does after figures are made.
What do you say is the biggest mistake while actually making figures?

[hurrikane]

IMO
it is not complicated to program in a speed fig calculator. Where the problem comes in is when an anomoly occurs. These will generally return an erroneous number. If the creator does not verify the integrity of every number then you cannot have confidence in the number. And, if you are not the one that created the algorithm to produce the speed number then it will be difficult to see or understand these anomolies.

I'm thinking this is what happens when people see strange looking Beyer numbers.

The bottom line IMO. If you want to handicap using speed as your dominate factor then, you must create your own number, either manually or programmatically and verify it as well.

[Valuist]

Brian-

Yes, I do my calculations by hand, not computer.

[Brian Flewwelling]

Originally posted by kenwoodallpromos
I will amend my answer to what the 'capper does after figures are made.
What do you say is the biggest mistake while actually making figures?

Like yourself, i don't know enough about making speed figures to tell this crowd what they are doing wrong...

My hope is that one or more of the experts will give me a hint or two about including projections into my making of the DTV. I have made two attempts at making figures of my own, and both were generally useful, but i had to watch for the 'silly' numbers that show up occasionally... probably due to a (large) variant based on too few races.

I think the par-charts do well when there are 3 or more races at a dist/surface for the day. But when there is only one, like many turf races and sometimes dirt-routes, the variant comes back as the difference in raw speed and the Par for that class. If it was a fast bunch of runners, that gives speeds that are too low, or vice versa. A properly projected "Par" for that race would be better.

fleww

PS: damn this thread, now i am thinking about this and that puts me in the middle of another project.:(

[cj]

I have written a program to extract the variants from assigned Beyer Speed Figures, and mark any that are not in line with the rest of the day with a question mark. The question mark has to be checked out and removed or errors will occur when the program runs. Its a good way to avoid over-reliance on the program's output.

I don't need to spend my days making variants for all the tracks, and I know the Beyers are pretty good overall. Its those oddball races I want to check out myself. Mix that in with pretty decent pace numbers, and I think I have a nice edge over someone using the Beyers or BRIS or TSN numbers.

[sjk]

Brian,

I started to write a synopsis of how I calculate daily variants in case it would be of use to you and had a feeling of deja vu. Looking back I see that you asked a similar question to which I briefly responded in a thread called par times about a year ago. If further detail would help you let me know.

I do not split sprints and routes except on rare occasions which means there are fewer occasions where I base a variant on a single race. When that happens the variant ends up being based on projection rather than on the par. Indeed if the track fails to time a race, I have no problem with making up a time and making a variant based on it.

Beyond making the daily variant, I make some race specific adjustments to the figures for the horses when they run again. This would incorporate elements related to unusual pace as cj has mentioned and data outliers as you have mentioned.

I find computer generated figures quite satisfactory and if you take the time to set up a program to do your own calculations I would hope they could work for you.

[alysheba88]

95% of the mistakes I see is in the interpretation of them. People get way way too caught up in the numbers and searching for the "perfect number". Fruitless waste of time

[JackS]

Aly- I agree but numbers have to be generated to more approxamate what is actually happening on the track.
No-ones numbers will be perfect but in this game, close can produce huge benefits at times.

[alysheba88]

Jack, speed figures were the first tool to give me a real perspective on the game. Enabled me to be a winner even when I knew little else. This was back in the late 80's when they werent as well circulated. So will always have a soft spot for them.

I just think people go too far with them now. They try to make them something they arent. Some want to quantify other things into the number, like pace or whatever. Trying to fill the pot up with all kinds of ingredients in search of the perfect number. It doesnt exist.

Speed figures should give you an indicator of how fast horses ran in prior races. Thats it. For me anyway.

Just one tool to use. They are not magic predictors of future behavior and no amount of adjusting and reajusting them will make it so.

[JackS]

Aly- I agree absolutly with your entire response.
Trying to tweek speed figures to that last 1/100th or even 1/10th is probably a waste of time.
HC's should develop a set of SF's and live with them for awhile.
Barring any drastic changes to a track surface, no changes in SF's will do anyone any good.
Individual numbers are all going to be slightly different and very difficult to quantify which persons numbers are actually superior to anyone elses numbers.
Results could span an entire meet or even carry-over into future meets.
SF's- not the all to end all. Many other factors to consider.

[Valuist]

I've noticed that Beyer figure makers will change numbers later on, if they feel the original was a mistake. Definitely a good thing, and you can figure that the farther back you go in a horse's pps, the more likely that the number is accurate.