By Tim Maas

Owner, Overlay Publications

In a previous article, I discussed the merits of using betting value (rather than winning probability alone) as a basis for placing wagers.  I also mentioned some possible approaches to detecting the presence of value.  I have great respect for people who can reliably "put a price on a horse" through visual observation, a qualitative "feel" based on experience, or through use of innovative techniques such as the equine biomechanics featured by casetherace.com.  However, the method of assigning fair odds to a race field that has worked most dependably for me, and that I would recommend as a skill that can be learned by anyone interested in value-based betting, is a numerical method based on established performance characteristics.

The quantitative measures that I have found most effective in this regard are “impact values”.  For those to whom this term is unfamiliar, impact values measure the power of handicapping factors by comparing the percentage of all horses exhibiting a particular performance characteristic in a representative sample of races to the percentage of all winners of the sample races possessing that same characteristic.  For example, if 25% of all horses in the sample had finished third or better in their last start, but 50% of sample race winners had done so, the impact value for the factor “Ran third or better in last start” would be 50% divided by 25%, or 2.00.  From this, you can see that if a factor is present in the same percentages of horses and winners, the impact value for that factor would be 1.00, and it would have no particular significance from a handicapping standpoint.  But, by the same reasoning, factors having a percentage of winners higher than the percentage of all horses would be positive indicators, with the beneficial effect increasing as the impact value rose, and factors producing a percentage of winners lower than the percentage of all horses would be negative, with the adverse effect on winning probability becoming more pronounced as the value moved downward from 1.00 toward the lowest possible value of zero.

Using impact values for probability-computation purposes is preferable to employing raw percentages, because impact values reflect the true relative strength of a factor, as well as the effect of the field size in which a horse is competing.  Knowing that a certain percentage of all race winners possesses a given positive attribute is useless from a handicapping standpoint if at least that percentage of all horses also possesses that characteristic.  In addition, a factor's winning percentage itself will vary, depending on whether the horse in question is racing against four other horses today, or against eleven.  Employing impact values corrects both these weaknesses.  Impact values accurately take field size into account when assigning winning probabilities.  They provide a statistical indicator that automatically rates factors by their actual handicapping significance.  They’re also mathematically easier to use, since they don’t necessarily decrease constantly in value when multiplied, as percentages do.  And multiplying impact values is the key to combining factors into a composite winning probability for each horse in a race.

In order for the multiplication of impact values to be valid and accurate, the factors used must be independent and non-repetitive.  For example, use of factors such as "in-the-money finishes in last ten starts" to measure consistency and "average earnings-per-start" as a class metric in the same odds-prediction model would result in inaccurate probabilities, since the performance of a horse in one category would overlap or have a direct relationship with its performance in the second category.  The two are inextricably joined.  By the same token, use of a factor such as "days since last in-the-money finish" in conjunction with "days since last race" would also skew the results, because each element is measuring recency of performance.  Using them both would give an inordinate, unjustified amount of handicapping weight to that area.

In seeking appropriate factors to use in handicapping, we are also looking for those that measure or assess fundamental aspects of handicapping, rather than peripheral or insignificant attributes.  We give preference to factors that can be used to rank entire fields from top to bottom, and that have a wide spread and smooth progression of values from high to low between those two extremes.  Finally, we are also trying to find elements where the performance of horses possessing them is so good (or bad) that it cannot be accounted for by random chance, but is instead the direct result of the influence of that factor working independently by itself on race outcomes.

Based on these criteria, the handicapping categories and factors that I have found to produce the most broadly-based and accurate results in constructing a fair-odds model have been class, speed, condition/form, early speed/pace, consistency, "connections", track contours, distance, and running surface.  The references that have been most useful to me in assigning numerical values and weights to these areas have been the statistical studies conducted and published by Michael Nunamaker.  These studies contain impact values pertaining to several measures for each handicapping area, broken down by distance category, sex, age, and running surface.

The class, speed, and consistency factors address underlying quality.  Condition/form, early speed/pace, track contours, distance, running surface, and connections influence whether the horse has been generally prepared to display that quality to best advantage, and will be able to do so under the unique circumstances of today's race.  And the odds obtained by multiplying each horse's impact values from these individual areas, and then comparing the result to the sum of the products for the entire field, indicate whether -- even with the best of preparation, circumstances, and intent -- the horse is worth a wager in light of the public's overall assessment of its chances.

Following is a specific example of impact values pertaining to measures in a suitable mix of categories that can be combined through multiplication to produce a fair-odds figure for each horse in a field.  The three values listed on each line correspond to the impact value for horses running in sprints (races of less than one mile) on dirt; routes (races of one mile or more) on dirt; and routes on turf, respectively.  (Turf sprints are not included.)  (The expression "good race" (a term coined by Dr. William Quirin in his book Winning at the Races) refers to a race in which the horse finished third or better, or (regardless of finishing position) finished within two lengths of the winner in a sprint, or within three lengths of the winner in a route.  "Speed points" refer to the early-speed measure of that name also devised by Quirin for that book (in which he also discussed how speed points are calculated).    Speed figures used are either the Beyer figures published in the Daily Racing Form, or those calculated by Bloodstock Research Information Services (BRIS).  "Top Five" jockeys are those ranking among the top five (or tied for one of the top five spots) in victories at the current race meeting, according to the jockey standings list published on the day of the race.  The "rear half" category in a field refers to the bottom three horses in a field of seven; the bottom four horses in a field of nine; and so on, always rounding down if the field size is not evenly divisible by two.)

Best Speed Rating in Last Thirty Days (Rank in Field):

First  2.08  2.07  1.83

Second  1.52  1.45  1.55

Third 1.18  1.09  1.39

Front Half of Field (but not in Top Three)  .97  1.06  1.04

Rear Half of Field  .59  .59  .60

Number of "Good Races" in Last Three Starts:

Three  1.54  1.53  1.45

Two  1.29  1.25  1.19

One  1.03  .93  .86

None  .67  .61  .64

Quirin Speed Point Total:

8  1.58  1.58  1.22

7  1.54  1.34  1.22

6  1.52  1.26  1.12

5  1.24  1.15  1.01

4  1.20  1.02  1.01

3  1.10    .90  1.01

2   .93     .89    .90

1   .91     .88    .90

0   .65     .65    .90

Two-Year Earnings (Rank in Field):

First  1.65  1.57  1.60

Second  1.35  1.35  1.39

Third  1.16  1.13  1.19

Front Half of Field (but not in Top Three)  1.05  1.01  .92

Rear Half of Field  .72  .74  .76

Jockey

Top Five at Meeting  1.50  1.37  1.37

Previous "Good Race" on Horse in Past Performances (but not in Top Five)  1.17  1.07  1.07

Neither of the Above  .58  .64  .64

see Betting Value Part 1: