Odds Distribution

[traynor]

An interesting area of research that obviously has a great deal of application for those pursuing "value wagers."

Mutuel Prices (PACE)
The Meadows:
[LessThan6-0.76] [6to10-0.09] [10to16-0.07] [16to22-0.04] [Over22-0.04]

Monticello:
[LessThan6-0.90] [6to10-0.05] [10to16-Insufficient Data] [16to22-Insufficient Data] [Over22-Insufficient Data]

Without posting long laundry lists, it should be clear that if one is seeking high value wagers, the wagering tendencies (and handicapping tendencies of the bettors) vary considerably from track to track. In the examples above, 90% of actual win payoffs at Monticello were LESS than 2/1, while only 76% of actual win payoffs at The Meadows were in that range.


Values are extracted from current races (last 45 days). In this situation, "long term trends" (last year, last two years, whatever) are pretty much useless information when one attempts to use them to make decisions about wagering on individual races.

[LottaKash]

Without posting long laundry lists, .

Traynor, it is probably too much to ask, but I sure would like to see "that laundry list"... :jump:

[traynor]

Northfield (PACE)
[LessThan6-0.79] [6to10-0.13] [10to16-0.04] [16to22-0.02] [Over22-0.02]

Northfield (TROT)
[LessThan6-0.90] [6to10-Insufficient Data] [10to16-Insufficient Data] [16to22-Insufficient Data] [Over22-Insufficient Data]

One could reasonably conclude that lumping pace races and trot races together (in MANY instances) generates little more than semantic noise.

One could also reasonably conclude that conditional wagers at Northfield made on the basis of perceived value at zero minutes to post are more likely to be successful in pace races than in trot races. Specifically, last minute odds changes for winners are more likely to drop below 2/1 in trot races than in pace races (at Northfield, among a number of others).

[traynor]

Traynor, it is probably too much to ask, but I sure would like to see "that laundry list"... :jump:

I am extracting the information from current races. Separate files by track, gait, etc., rather than in a database of all tracks, all races, both gaits. New process, in conjunction with a new model I am building. When I have enough useful data from current races (for other tracks) to be meaningful I will post it.

[traynor]

Another way of looking at the data for Northfield would be to understand the significance of the [6to10-0.13] entry for pace races. Fully 13% of pace race winners are in the (highly desirable) 2/1 to 4/1 range. VERY few of the trot races at Northfield are in that range.

In general, the lower the final odds of winners, the more "predictable" the race outcome. One might pick more winners at Northfield by concentrating on trot races. One might make more money at Northfield by concentrating on pace races--especially in those races in which there is no "obvious favorite" (even money or less).

[traynor]

If that last post seems a bit obscure, an example might clarify it.

Current model output, specific model, all tracks:
AllTPace 0.81 201 L1 S
AllTPace 1.10 46 L2 S
AllTPace 2.29 21 L3 S

VERY tight contender selection criteria. "L1" means there was only one suitable selection in each of those races, "L2" means there were two, etc. In races with only one selection, it was often "discovered" by other bettors, and not profitable to bet. When there were TWO selections, the outcome was more in question. Betting BOTH (flat bet Dutch) was profitable, when there were THREE selections, even more so. This pattern is not an anomaly--it is typical of the models I use on a daily basis.

The more certainty one seeks in wagering, the less rewarding that wagering becomes.

[traynor]

I should probably explain that I am not much one for chasing rainbows. The analytical process for "ROI" in the models I make is automated and the input is cleaned, not "dirty." Meaning all mutuels have been corrected (rigorously) for outliers.

In the unlikely event anyone is actually interested in the process of correcting for outliers (for their own use), the process I have found the most useful is one of the simplest. The value at the 75th percentile is used for truncation--any mutuel greater than that amount is truncated to that value BEFORE any ROI is calculated. I used to use 1.5 times the mean of the interquartile range. The simpler way works better. Standard deviation is almost as likely to be deceptively tweaked by a few anomalies as using raw data (VERY deceptive).

[LottaKash]

The more certainty one seeks in wagering, the less rewarding that wagering becomes.

Are you certain of that ?... :D ..Thx T..

[Saratoga_Mike]

Specifically, last minute odds changes for winners are more likely to drop below 2/1 in trot races than in pace races (at Northfield, among a number of others).

...as players wait for the last possible moment to make sure there isn't a break before the gate opens? or is there another reason for this?

[traynor]

...as players wait for the last possible moment to make sure there isn't a break before the gate opens? or is there another reason for this?
Online bettors typically use "0 minutes to post" as a cutoff for conditional wagers. Same thing on-track--many bettors wait until the last possible moment to bet, (usually) hoping the final odds are close to the odds displayed at the time they wager. Because there is less ambiguity in the outcome of (many) trot races at Northfield (or more large last-minute wagers) an entry 2/1 near the start is more likely to drop below that figure in a trot race than the same scenario in a pace race. At other tracks the reverse is true.

[traynor]

Are you certain of that ?... :D ..Thx T..

Pretty much. The little old lady with the shopping bag of cash bet to show on Stagedoor Johnny is a classic example.

[Jess Hawsen Arown]

Sadly, the lack of handle in harness racing causes outrageous odds changes on the last flash.

The best indicator (but not always accurate) reflection of where the final odds will end up is found on the exacta board.

Multi-race probables can also work, but they are less reliable.

Since time immemorial, the best value in racing is a win bet on a horse clobbered in the last leg of the double.

[traynor]

Sadly, the lack of handle in harness racing causes outrageous odds changes on the last flash.

The best indicator (but not always accurate) reflection of where the final odds will end up is found on the exacta board.

Multi-race probables can also work, but they are less reliable.

Since time immemorial, the best value in racing is a win bet on a horse clobbered in the last leg of the double.

I agree that one can sometimes detect "less than obvious selections" that are (relatively) heavily bet on top or wired in the exacta pool. Used to work pretty well. Not so much now that major amounts can be wagered with a single mouse click at the last minute.

Some believe they are "tracking smart money." The "smart money" got that way by concealing what they are doing. Like many other areas of life, you can be fairly certain that if you see something, it is because somehow either wants you to see it, or doesn't care whether you see it or not. Especially things as easily concealed as exacta wagers.

[traynor]

Monticello today, favorite won 6 of 9. In the other three, the favorite placed.
Yesterday, favorite won 6 of 9, placed in 2, and finished worse than place in exactly 1 race--in which the 3rd and 2nd favorites finished 1-2. Not an easy track for "value bettors."

[HWIG]

The problem with looking at odds with live racing is that up until they are off, all you see is the money bet at the track. After the race is off, the outside money is put into the pools.
If you are using the charts to come up with live money, it cannot be accurate.
Years ago I liked to dutch races against the odds, it was a good wager. Today it is a poor wager, because you never know what the final odds will be when the pools close.

[traynor]

The problem with looking at odds with live racing is that up until they are off, all you see is the money bet at the track. After the race is off, the outside money is put into the pools.
If you are using the charts to come up with live money, it cannot be accurate.
Years ago I liked to dutch races against the odds, it was a good wager. Today it is a poor wager, because you never know what the final odds will be when the pools close.

I agree completely. That is why I find it so comical when some claim to be "tracking the smart money" with (or even more ludicrous, without) various home-grown Mickey Mouse computer apps.

My point about Monticello is that more than a few of those favorites are NOT so when the tote is locked. They only become so when it is too late to do anything about it. Additionally, those chasing imaginary paper ROI's with real money take a serious hit. On one hand, they pick the wrong horse to bet on and it loses to the (newly defined) favorite, or on the other hand they bet the right horse but the payoff is peanuts (and nowhere near the "expected return" indicate by their "research") because the late money was dumped on the almost certain winner.

There is actually a way out, that involves applying (similar) patterns in a large number of (similar) races. The results of any given race may be up, down, or "about even." If the model is accurate (and reasonably predictive) the ups will more than compensate for the downs and evens. That is the major reason I (usually) have so little interest in the outcome of individual races.

[traynor]

I should probably explain that I am not much one for chasing rainbows. The analytical process for "ROI" in the models I make is automated and the input is cleaned, not "dirty." Meaning all mutuels have been corrected (rigorously) for outliers.

In the unlikely event anyone is actually interested in the process of correcting for outliers (for their own use), the process I have found the most useful is one of the simplest. The value at the 75th percentile is used for truncation--any mutuel greater than that amount is truncated to that value BEFORE any ROI is calculated. I used to use 1.5 times the mean of the interquartile range. The simpler way works better. Standard deviation is almost as likely to be deceptively tweaked by a few anomalies as using raw data (VERY deceptive).

That may be about as clear as mud. In plainer English, create a list of doubles representing the mutuel prices. Sort and reverse (highest to lowest in revere order--or "sort descending"). Multiply the number of instances (".Count" in most civilized languages) by 0.75 and use the value of the double in that position as the "75th percentile." Iterate through the list again with a simple conditional--if the value of the double is less than the 75th percentile value, add it as is. If it is equal to or higher than the 75th percentile, add the value of the 75th percentile in its place ("truncate it").

If you are coding it for ROI calculations, total the values of the (truncated, corrected for outliers) mutuels, divide by the number of matches (the situations in which that bet actually would have been made) times 2 (cost of bet).

Unless you can routinely correct for outliers and still show a positive ROI, it is probably best to keep looking and pass on betting. Most "paper" ROIs (and most resulting from "regression analysis" of past results) are heavily dependent on anomalies in limited samples. Great for discussion and "bragging rights." Pretty much worthless for betting in the real world.

[LottaKash]

Great for discussion and "bragging rights." Pretty much worthless for betting in the real world.

Whew thx Traynor, for saying that...You lost me awhile back there...haha.. :confused:

[traynor]

Whew thx Traynor, for saying that...You lost me awhile back there...haha.. :confused:

In general, the lower the ratio of wins to matches, the more likely the "ROI" exists only in the specific clump of races used, and the less likely that scenario is to be predictive. In 1000 race sample, 300 matches, won 10 of the matches for an ROI 1+ whatever (or even 2 or 3 plus whatever). Pick3, Pick3, trifectas, exactas. Pretty much worthless for betting in the real world.

In general, the higher the ratio of wins to matches, the less likely the "ROI" exists only in the specific clump of races used, and the more likely that scenario is to be predictive. In 1000 race sample, 300 matches, won 100 of the matches for an ROI 1+ whatever (or even 2 or 3 plus whatever). MUCH more likely to be predictive (of the outcome of future races) with a 100/300 ratio than a 10/300 ratio--with same ROI, using the same base set of races.

It is relatively trivial to locate "positive ROI scenarios" with (readily available, free, well-designed, easy-to-learn) data mining apps (Anaconda3 for example). However, that is only the start of the process.

[Fox]

The problem with looking at odds with live racing is that up until they are off, all you see is the money bet at the track. After the race is off, the outside money is put into the pools.
.

I am positive that this is not completely accurate. I play at one if the biggest adws, and I am 100% certain that other than the normal delay in the tote updating cycle, my wagers go immediately into the pool. The 2 other adws I have used over the years have done the same. I am not saying that it is not possible that some adws do not immediately send there bets to the pools, but I am sure it is not as you said that all outside money is not reflected until after the off.

[traynor]

I am positive that this is not completely accurate. I play at one if the biggest adws, and I am 100% certain that other than the normal delay in the tote updating cycle, my wagers go immediately into the pool. The 2 other adws I have used over the years have done the same. I am not saying that it is not possible that some adws do not immediately send there bets to the pools, but I am sure it is not as you said that all outside money is not reflected until after the off.

It is not so much "all money" being added as it is "significant amounts" (including conditional wagers and most if not all wagers defined as "smart money") being dumped at the last minute.

"Tracking the smart money" by monitoring fluctuations in exotic pools and will pays is "interesting, but not especially useful." Including a number of (sometimes very pricey) apps that claim to offer the keys to the candy store by doing so. Hope springs eternal in the fantasies of software developers. Especially those who make more from developing software than from wagering on the output of that software.

[LottaKash]

It is not so much "all money" being added as it is "significant amounts" (including conditional wagers and most if not all wagers defined as "smart money") being dumped at the last minute.

"Tracking the smart money" by monitoring fluctuations in exotic pools and will pays is "interesting, but not especially useful."

Before the advent of ADWs and OTBs, I used to use an outdated now, ODDS MODEL, that was the invention of Prof Igor Kusyshyn, a noted Blackjack author (for his time) who held a Phd in Psychology, as well as Gambling, and who also wrote books on Harness Racing... I had some good success' with that Odds-Model once I got the hang of it...Once upon a time tho...

Some anecdotal thoughts about Kusyshyn...Back in the early 80's my brother and I attended a Harness Racing seminar in Toronto, and on the last nite's outing at Mohawk (when it was still a 5/8ths track,, we all set out to wager on the races that the Prof's system had tabbed...At the end of the nite, the Professor spied me counting up my hundreds, and said, "how did you acquire all that kash, the system went down tonite"...I said, "Oh, I don't use your system"... The power of what a few beers, hey.!?..I was fearless then..

Since tho, I am lost on the tote in many instances, so I just bet my plays no matter what now...Unless of course the odds on my play are too low, without the expectation of them getting any better...

Just some thoughts..

[traynor]

Before the advent of ADWs and OTBs, I used to use an outdated now, ODDS MODEL, that was the invention of Prof Igor Kusyshyn, a noted Blackjack author (for his time) who held a Phd in Psychology, as well as Gambling, and who also wrote books on Harness Racing... I had some good success' with that Odds-Model once I got the hang of it...Once upon a time tho...

Some anecdotal thoughts about Kusyshyn...Back in the early 80's my brother and I attended a Harness Racing seminar in Toronto, and on the last nite's outing at Mohawk (when it was still a 5/8ths track,, we all set out to wager on the races that the Prof's system had tabbed...At the end of the nite, the Professor spied me counting up my hundreds, and said, "how did you acquire all that kash, the system went down tonite"...I said, "Oh, I don't use your system"... The power of what a few beers, hey.!?..I was fearless then..

Since tho, I am lost on the tote in many instances, so I just bet my plays no matter what now...Unless of course the odds on my play are too low, without the expectation of them getting any better...

Just some thoughts..

Igor got me started in (relatively) serious wagering on harness races. I used his PPC in the early days (when he was teaching math and stats at York). I assume you know that Igor's nom de plume "Lance Humble" was a spin on that used by another blackjack writer of the time (and a similar double entendre) "Stanford Wong." I still get a smile from recalling Stanley Robert's realization that "Stanford Wong" was not Chinese.

Never used Igor's Odds-Model. Went a different direction. The NY/Toronto/Montreal circuit was great fun. Blue Bonnets was my favorite.

I used Igor's Hi-Opt II with success for awhile, before I switched to Revere's APC (and later Uston's APC when playing on and with various "investment groups"). A nice life, filled with strange and wondrous experiences, friends, events, insights, and more than a few adventures. The stay-at-homes filling their "lives" with gathering nuts for the winter they will never live to see are to be pitied.