Ranking entries on factors if a factor doesn't apply to every entry

[podonne]

Greetings,

I have a few factors that rank an entry in comparison to the other entries in the race. When the factors consider things like past performances, like best last beyer, some entries may not be able to be ranked, like first time starters.

How should I treat these rankings then? If you can only calculate a factor for half the entries, what does it mean to rank #1, or #4, in an eight horse field? If you can't calculate a ranking for every horse (100%), do I throw the factor out completely? 90%?

Thanks,
podonne

[InControlX]

I've had fair success with two adjustments to handle "partial field rankings", for what it's worth...

1. Adjust your handicapping numerical values for neutral = 0 (better = positive, worse = negative), then zero entries you can't rank. This at least puts the unknowns in the middle.

2. Assign a Confidence Ratio to each race by dividing the number of entrants for which you've determined a ranking value by the total number of entrants. Test this Confidence Ratio vs. your race winning percentage over as many samples as you can to yield a "go/no go" decision level for a particular race type. A good handicapping approach may be OK with a confidence ratio of 0.4 for a Maidens Race, but need 0.8 for open claimers.

ICX

[proximity]

podonne,

you could consider putting the horse's morning line ranking in for such factors.

proximity

[completebill]

It seems that you are really asking how to rate, or rank, unknowns. There are various types of unknowns of course, with the most prominent types being firsters And foreign shippers. The next ?? group would be 2d-time starters and long layoff horses. Some of these, obviously, can't be directly compared, but that doesn't mean you shouldn't use what you have.

Rate the known quantities. Next, compare your top-ranked horse to PAR for this race level, whether using Beyer par, pace pars, or a class/competitive rating. If your known element can't run to the race par, the unknowns require much more scrutiny and may well produce the winner.

Various Forum members here (including myself) have previously written about evaluating FTS's. There are even books on the subject. Briefly, it's trainer, "connections" (incuding owner, breeder, trainer, and jockey), works, and pedigree.

Much the same can be said about 2d-time starters. Many times the debut is a throwout. Evaluate as if a FTS. Did the horse show any "moves" in the debut? Any speed? Are there equipment changes today? Jock change? Improved works? Trainer's record with 1st vs. 2d-timers? Any significant trouble in the debut?

Foreign shippers: Look at class of competition and success in the country of origin. Evaluate trainer and works. Compare TimeForm ratings.

Layoff horses: Look at previous performances, trainer records, tr./ Jk combos, and works. Look for signifivcant positive or negative class moves.

Briefly, USE EVERYTHING YOU HAVE! It's a tough game, but it's a wonderful intellectual challenge, and, with hard work, it's beatable!

Good luck!!

[cj's dad]

Greetings,

I have a few factors that rank an entry in comparison to the other entries in the race. When the factors consider things like past performances, like best last beyer, some entries may not be able to be ranked, like first time starters.

How should I treat these rankings then? If you can only calculate a factor for half the entries, what does it mean to rank #1, or #4, in an eight horse field? If you can't calculate a ranking for every horse (100%), do I throw the factor out completely? 90%?

Thanks,
podonne

I am struggling to get to the point where any race with any 1st time starters is a watch only race. I live and die by PP's especially certain ones and the 1st timers are killing me.

[Overlay]

I subtract first-time starters from the field size when dealing with ranking variables that apply only to horses that have raced previously, and calculate the relative rankings of the remaining horses based on the reduced field size. I have a different set of criteria (not geared to field size) that I apply specifically to the first-timers in order to provide a basis for comparing them against the horses that have raced before. On factors that are confined to individual horses (rather than comparative in nature), I apply the factors to horses that exhibit them, and take no action with regard to horses that don't.

[completebill]

To C.J.'s Dad------Research, using the search function, both this forum and the computer handicapping forum.There have been some good posts regarding firsters.

There are books and, I believe, videos, on the subject. When I first posted that firsters have a higher win % for favs than all other races, most readers were initially skeptical. Those that did data base searches soon confirmed this. First-time starters are VERY handicappable, and you'll be severely limited in your betting opportunities until you leasrn the process.

Several of the handicapping programs discussed in these forums have some GREAT tools for analysis of FTS s-------workout speed, frequency, and VOLUME, trainer stats, pedigree, etc.

[podonne]

Appreciate the replys. I realize I used a FTS factor in my example, but I don't mean to specifically ask about how to evaluate first time starters. My concern is more generally, when a horse can't be ranked, for whatever reason, how do you deal with it. Other, different examples might be auction price, or trainer win pct this year (at the beginning of the year).

InControlX, your idea was very interesting, worth exploring. If I hear you right, you are essentially reducing the detailed rating (rank 1-8 in an 8 horse field) into a more general one, -1,0,1. Perhaps we can prove this, calculating a win pct for each category, then adding in the "unknowns" as zero and seeing if the probabilities hold true.

[podonne]

Again, I appreciate any replies that do not specifically address how to rank first time starters.

You are all correct that there are many, many, many, other places to discuss that topic.

[robert99]

Again, I appreciate any replies that do not specifically address how to rank first time starters.

You are all correct that there are many, many, many, other places to discuss that topic.

I would approach this long these lines - businesses make these type of decisions under uncertainty all the time:

If a factor is missing how important is that factor?

What is the likely range and average figure for that missing factor - if you assign the horse population average as best estimate what difference does that make to your rankings?

If you assign the trainer average for that missing factor what difference does it make to your rankings? Some trust is needed that the trainer has made some estimate to enter the horse in the right race.

How does the trainer score when he introduces change - class, distance, aids, jockey etc?

What extra price do you need to cover the uncertainty caused by the missing data?

At what point will you miss the race out?

[podonne]

I would approach this long these lines - businesses make these type of decisions under uncertainty all the time:

If a factor is missing how important is that factor?

What is the likely range and average figure for that missing factor - if you assign the horse population average as best estimate what difference does that make to your rankings?

If you assign the trainer average for that missing factor what difference does it make to your rankings? Some trust is needed that the trainer has made some estimate to enter the horse in the right race.

How does the trainer score when he introduces change - class, distance, aids, jockey etc?

What extra price do you need to cover the uncertainty caused by the missing data?

At what point will you miss the race out?

Interesting thoughts robert99. Perhaps it depends on how you are using the ranking then. If you're purpose is just to identify the first, and you are reasonably sure that an unknown is not going to be first, then it doesn't matter so much if data is missing. If you were trying to do a more complete analysis, then for any given factor, you would need a reliable way to predict what that value would be.