Values as Ranges

[traynor]

A major difference between professional race analysis apps and most other race analysis apps is the consideration of values as existing in a range, rather than a specific. An example, taken from a generic win model for five-eighths tracks, is (condensed) below.

0.50 0.19 0.86 3.29 [LessThan6-0.89] [LessThan10-0.96] [LessThan16-1.00] [LessThan20-1.00] [Over20-0.00]

Values are Win%, Place%, ROI, (cleaned) mutuel, and percentages of winners BELOW specific ranges. (of course, the ranges can be set to any values you want, depending on the type of model you are building.)

From the example, it is fairly clear that finding a pattern match going off at 8/1 is pretty much a no-brainer--100% of entries that match that pattern that go off at odds exceeding 7/1 lose.

I have mentioned before that Katherine Jung's observation that most of the entries I bet at Fraser Downs that went off at odds of 3/1 or less won, and most of the entries that went off at odds more than 3/1 lost was the impetus for a LOT of study relating wagering patterns and final odds. While the primary purpose of that study was developing a "whale tracking" module, the generic principle can be usefully applied to most wagering. "Looking for value bets" is great, providing you have sufficient insight into the normal distribution of mutuels for your selections.

For some wagers at some tracks, seeing a selection that you think should be 2/1 or less go off at 5/1 is justification to make a larger wager. At others it is justification to either pass the race or look for other wagering opportunities. The trick is in doing enough in-depth, track-specific, pattern-specific analysis to know which is which.

I can't do it for you. If you want to win, you have to do it yourself.The effort is well compensated by the results, in either increased wins or decreased losses. Or both.

[traynor]

Of course, there is an upside in the modeling process outlined above. Also from win model for five-eighths tracks (condensed):

0.36 0.12 1.16 5.03 [LessThan6-0.59] [LessThan10-0.81] [LessThan16-0.90] [LessThan20-0.90] [Over20-0.10]

19% of winners pay $10.00 or more, and fully 10% pay $20.00 or more. While the win rate is lower than that of the first example, the return is substantially higher.

Obviously, this is a case in which "finding value" should be cause for both happiness and increased wagers. Providing one does not get too carried away with the happiness bit--this is only a horse race, after all.

[traynor]

Well, so what? Same model, folks. Lumped together, ROI is in the mid-90s, fodder for those rebate chasers who believe the only thing holding them back from fortune and glory is insufficient rebates. (And perhaps a half-mil bankroll and a cast of dozens to keep it all going.)

The models are the same. The method of locating the selections is the same, with one small difference. The first model is layered to those selections in post positions 1 through 5, the second model is layered to those selections in post positions 6 through 10. (I know about PP6 at 5/8th tracks. No commentary needed.) Simply put, the inner posts are overbet. The outer posts are underbet.

[traynor]

The information contained in views of results in ranges (rather than yes/no, plus/minus, profit/no profit) can readily be seen in an example of an exacta model for Northfield:

1.21 0.14 0.17 21.27 Nfld BOTH AP [LessThan40-0.18] [40 to 60-0.15] [60 to 80-0.07] [80 to 100-0.06] [Over100-0.01]

This is a highly desirable pattern, in that the range ratios (winning wagers compared to matches) are decent over an extended range of values, right up to 49/1. With only 1 in 100 winning exacta mutuels over $100 (at a cost of $200 to make all pattern match wagers in that range) it is not rocket science to be skeptical of including such in wagers.

In contrast, (as an example) exacta wagers in the 80 to 100 range cost the same $200 per 100 wagers, and returned $480 or more (6 wins at $80 plus).

It is easy to monitor such patterns, and can produce a lot of useful information. People (and their wagering patterns) are FAR more predictable than horses (and their potential performances). In that predictability is a lot of opportunity for profit, with little more needed than a bit of study.

[pandy]

When favorites were winning around 34%, a long time ago, I made most of my profit from posts 7 and 8 on half mile tracks. I believed these posts were underbet and loved it when I found a horse that I thought was ready for a big race. But now on some half and five eighth tracks it seems to me that many of the outside posts are overbet. There are horses at Yonkers that go off at anywhere from 6-1 to 15-1 from post 8 that should be more than double that. If a post is winning at 3%, there is very rarely a good bet from that post.

[precisionk]

Originally Posted by pandy

When favorites were winning around 34%, a long time ago, I made most of my profit from posts 7 and 8 on half mile tracks. I believed these posts were underbet and loved it when I found a horse that I thought was ready for a big race. But now on some half and five eighth tracks it seems to me that many of the outside posts are overbet. There are horses at Yonkers that go off at anywhere from 6-1 to 15-1 from post 8 that should be more than double that. If a post is winning at 3%, there is very rarely a good bet from that post.

I know they were doing 1 1/16 races at Yonkers a year or two ago. They still running those out east Pandy? Still waiting for them to fix that inside bias there.

[pandy]

Originally Posted by precisionk

I know they were doing 1 1/16 races at Yonkers a year or two ago. They still running those out east Pandy? Still waiting for them to fix that inside bias there.

Nope, they run mile and a quarter trot races that they simulcast to France, which are very good, not that speed favoring. No more 1 1/16 races. Saratoga runs some 1 1/16 races for their handicap races, which is a great idea.


Yonkers was supposed to shorten the stretch but for some reason they moved the finish line but never got it going.