What If we generated an Auto Odds Line
 

Using each horse's last five actual odds, we generate a predictive odds line to compare to actual odds. Then using a range, say +/- 25% we see who is falling inside or outside that range. Naturally there could be ways to make some adjustments based on finish last race or two, and class drops/rise, etc. I would make the average of the odds an exponential one, so that the last two carry more weight than previous three.

I know using a range, + or - 25%, is just a starting point. Using excel if you wanted, you could use a range within +/- 2 sigma limits or more. I know from following odds and betting for a long time, that often there will be only horse that will vary a lot from the ML. And often, there will be several.


The following chart(see link) shows three horses out of 8 that are outside those limits. Since the #1 is only slightly outside the 25% range, I'd only look at the #4, who is -61% below the predictive odds and #5 who is +45% above the predictive odds line. Is there some factor in their pp's that warrants #4 being bet much below where it usually is bet? Or does the #5 have the kind of value over its usual odds line? We've all seen a horse that's cold on the board, nothing's simple in this game.


https://docs.google.com/document/d/1...it?usp=sharing

Auto Odds line-rev.
 

Revised the auto odds line chart to show current odds more like tote board. That change takes the #1 out of consideration from previous statement.

https://docs.google.com/document/d/1...it?usp=sharing

My analysis centers about normalization. Recently, in my multiple re-reads of the book "Chaos Theory" I was curious about a term I read, renormalization. Look it up. It seems well beyond my pay grade. Would it have any application to horse racing analysis? Actually I may be doing it without knowing it🙂

I just wanted to congratulate you on discovering what many of us who follow money flow have realized for quite some time. While the handicappers attempt to develop their own subjective odds line for each entry based on PP data, others like us realize that the actual odds line has already been presented, and it’s completely objective.
How it’s interpreted is another story, but it looks like you’re headed in the right direction.
GL
Nick

Well, why didn't you just say so? Why all the beating around the bush?
Is there anything else you're not telling us? :rant:

Interesting.

Because I never brought it up!
You can thank FakeNameChanged for providing those comments.

In terms of following the money flow there’s an awful lot that could be disclosed, but because I’m not the developer of the program I use I’m not at liberty to discuss that information.

When you refer to money flow, are you trying to sperate "inside" more informed money from the money being bet by the general public?

.... if I may. And Nitro can correct me if I'm wrong. He doesn't seem to be making a distinction on whose money it is, just that the market is efficient. He hasn't expounded on it enough for me to know. But I think it's a given that late money is more efficient than earlier money.

It's like the Efficient Market Hypothesis in financial markets. A stock's value is exactly what the ticker tape say's because it reflects all known information about the stock. With any change in fact or expectation there's a thousand sharks waiting to pounce which keeps prices fair. But there must be some mispricing in the market or there is no opportunity to profit. It's one thing to buy stock and hold onto it as it increases in value over time as the overall market grows and another to be able to time the market and jump in and out as opportunities arise if anyone can really do that. EMH say's you can't.

In pari-mutuel markets there is a final end to volatility when the house quits taking bets and the tote board finally stops tallying.

Personally, I can find overlays by making a line. Really, all you need is a different opinion. This opportunity will always be there due to the randomness of the game. But ... and this is key ... I haven't made a billion dollars doing it, so I'm always listening and learning.

Late money more efficient? Perhaps, but some of my best returns occur when my pick goes into the gate as the favorite and comes out second choice.

When I refer to money flow I’m observing the periodic changes that the tote analysis provides during a typical betting cycle. The changes are the numerical values for each entry that the analysis provides. It also creates what’s called a Par value that’s based on All of monies combined being analyzed. These values are then visually compared at 3 pre-determined betting intervals prior to post time. The time between these intervals vary based on the size of the typical betting handle being examined. The entries with values closest to the Par value are of primary interest and will eventually establish our belief in which are the real contenders in the race.
From what I understand there’s no need to distinguish where the money originates from. The analysis works by monitoring its flow by evaluating its movement in terms of when, where, and how much is entering each of the betting pools at the desired time intervals. Be it early or late money the primary consideration is always the number value relationships between each entry and Par.
When I once explained how I now played the horse racing game to a close friend of mine, he suggested that it’s very similar to what he and his associates do when playing the stock market. They like to find companies to invest in where they’ve discovered that the principals have also a substantial vested interest. Apparently they believe that it’s a much better investment knowing that the owners themselves have such high regard for their company’s profit potential.

A good Example:
Post #23
http://www.paceadvantage.com/forum/n...eply&p=2603660

TOTE ANALYSIS for Race # 8 – (Entries w/Computed Values closest to PAR are always of interest)

I just don't think it's that hard to find the contenders with old school speed, pace, class and form elements of handicapping. Have you or whoever's system this is run a regression analysis around this Par?

That was in my original post ref Mr. Benter using a model. William Quirin offered a good discussion of regression analysis for his study in Winning at the Races appendix B p. 301.

Nitro, In trying to understand your example above, Wouldn't the :9: and :2: be closest to Par at 3MTP? Is your Par value, some combination of an average of ML+15 MTP(or 20)+ Prev. Race Ave.(my example)? I've seen others use (ML + 15MTP)/2 as their Par value. Okay, if you can't say exact formula, just interested in the combinations. Or do you recalculate the Par at 3M, 7M and 12M snapshots?
Since I like to follow money flow also, my plays usually preferred when the money flow is shaded away from some par or ML value. From the races I've analyzed, when the money flow follows a ML or Par very closely across the board, that's often exactly how the horses finish. It tells me that in a particular race there isn't any false favorites among the top three or four. My thinking may be wrong, but it always seemed logical.

I know the odds are generally efficient, but I'm still not seeing how tracking the money flow is "adding value" to the analysis. The idea is to find the inefficiencies.

For example, I've been playing horses for around 45 years. After all this time, I'm pretty good at estimating how a race is going to get bet within some range. Sometimes, I'll see a horse getting bet a lot heavier or a lot less than I expected. To me, that "could be" meaningful betting action because either I'm missing something important or someone knows something about this horse that's not available to the public (and might not be fully reflected on the board).

We see this with first time starters all the time because the public has so little information to work with, but IMO it's happening all the time. It's just harder to notice because there are fewer things insiders might know that the public won't know also.

Other than that, am looking for fundamental handicapping factors, ways of measuring them, and circumstances that the general public may not betting correctly.

I previously posted the results of my "automated" race flow metrics covering multiple years that showed flat bet profits on strong upgrades and severe underperformance on strong downgrades. I did that because the metrics are sufficiently complex and took so long to develop, I wasn't giving much away.

I've since developed some automated bias metrics that look very promising.

But it's a lot to ask people to give away insights and angles that can help produce profits and potentially destroy their own value by making them common knowledge. It's too damn hard to find anything worth anything in this game to just give it away if it's working.

These inefficiencies often don't last.

I once had a "place" inefficiency that was like a gravy train for a few years (and people laughed at me because I was betting to place) when NYRA had a 14% take on WPS (now higher), I was getting a 7% rebate (now lower), and they hadn't switched to "net pool pricing" for place and show playoffs yet (which hurt the edge a few percent). All those negatives plus more computer betting and that edge is long gone.

:ThmbUp::ThmbUp:
^^^^THIS^^^^

Exactly right! :ThmbUp: Why give knowledge away when it's still profitable? It's a lot smarter to do as the handicapping authors do. Give the knowledge away after it STOPS working. :)

I believe there’s an important distinction that has to made between what one might consider a real contender and an animal that just looks good on paper. I personally believe after seeing the results of hundreds of races when compared to their related tote analysis that the contenders can be objectively deduced. Having also used the old school handicapping methodologies in the past I can also say that any interpretation of those elements is purely subjective. I value the objective information primarily because it already has an integral significance since it’s based on money which after all is what this game is all about.

I’m not sure if a regressive analysis of the Par value alone would provide any consequence. I’ve been informed that the formulas being used in tote analysis are heavily influenced by the human psychology of gambling. How that discipline is converted to a mathematical formula is beyond my pay grade. I really don’t find it necessary to understand the underlying nut-and-bolts as to how the tote analysis functions when I can produce results like this:
http://www.paceadvantage.com/forum/s...d.php?t=160763
http://www.paceadvantage.com/forum/s...d.php?t=160454

Been saying this forever. I have no problem asking/answering questions to spawn new ideas or perspectives, but I'm not giving anything away that I KNOW is an edge. I was a losing player from 14-35 years of age. No immediate plans of giving "my stuff" away.

:lol:

Let's face it, even if you have an edge on this game, the amount time, work, and dedication involved in keeping it typically makes it a bad use of time unless you love the sport and challenge of trying to show a profit at the end of the year. To make any significant money you need the bankroll and psychological makeup to put large sums of money through the windows and deal with the ups, downs, and stresses of that long term. Not everyone is wired for all that. It's easier to write a book, sell information, or get paid to give away your picks and insights.

Right on the money as usual. :ThmbUp:

:ThmbUp::ThmbUp:

No matter how good the horseplayer is, his income remains very irregular and inconsistent...and this can fray the nerves of all but the most stoic players. I applaud those who leverage their gambling knowledge into a steady income. It's hard to overrate a regular paycheck.

Forget common knowledge, all it take is one betting team to latch onto the info and value is gone forever or as long as they are in business.

What about the rest of us? We got a right to earn.

There are definitely teams putting in the work, and some of the time they are all over certain angles where you may have appeared to find value in the morning line.

Have a look at WWW.HKRACING.COM. They have PDFs and a web version.

To your first point ... Benter found that there was an extreme bias between his strongest overlays and the outcome of the race, i.e. his horse nearly always lost!

In the end, Benter did the common sense thing ... he combined his own odds with the public's in another multinomial logit model, and used the odds from that model to base his betting on.

Dick Mitchell said the same thing - that his underlays outperformed his overlays.

I've also found that in the entire make-a-line-and-bet-into-the-tote process.

Benter made it work by... well, making it work. LOL

People from inside his former team, have told me that he started with a high weight for the public's opinion and, over time, was able to lower that weight substantially.

This approach, what I call, "smoothing," has fallen into disuse as the final tote odds have moved towards higher efficiency relative to gate odds.

The whales make it work by PREDICTING what those final odds will actually be. This naturally adds a new factor: "Who/How will the "really smart" money actually wager?"

My goal for the last 18 months or so has been to answer the question: "WHO WILL BE BET DOWN?"

IMHO, that is a game changer for the small handicapper. It is the great equalizer.

When you say "bet down" - bet down from what - the morning line?
From gate loading to final odds?

What if you your models odds-line rankings mirror the public? Your fav wins same % as public, 2nd fav same % as public etc. Any ideas on how this could be useful?

From gate loading to final odds?

Yes. First horse loads the gate.

What if you your models odds-line rankings mirror the public? Your fav wins same % as public, 2nd fav same % as public etc.

Those are, of course, two separate targets.

Since the goal is to outperform, I don't see how this would achieve that. However, know with a high confidence level what the odds would ultimately look like would solved half the problem: "What will the odds be?"

Would still leave you with "What are the horses probabilities?"


Instead, I've taken a different direction.

Thinking out loud here. So Wm. Scott tried using the top three choices only, because at that time, the top three in the wagering won ~67% of the time, and placed 90% of the time. Only one of his problems was that the top three at 3 MTP, 2 MTP, Post time, aren't always the top three in the ACTUAL betting. Today, it's probably closer to 69% to 70% by the top three betting odds.

How could that be useful, you ask? I tried Scott's idea, what was it, 35 years ago? and in 10 races you could not determine the top three consistently in maybe 3 or 4 races out of 10. Sometimes it was only 2, sometimes it was 5 out of 10.
So if we could focus our selection efforts on only the top three by pp's, money-flow, or whatever you choose, would that not get us closer to an even playing field? If you only focus your selection efforts to three choices, wouldn't our batting average improve? ROI, not sure?

So back to your question, If we could model an odds line that mirrors the final public odds, reliably, wouldn't selecting a winner or an ABC group for horizontals be improved? Not sure.

Let me think about it a while longer.

Nope. Same answer as last time. ;-)

First, William Scott's stuff simply never worked. In fact, it kind of worked better in reverse! (i.e. toss the plays and pick from the other 2). Doesn't really matter because that was the Stone Age compared to today.

Today the top 3 win 74%.

There is something special about that number but it is not what you'd imagine.


Since you asked...

In 2018 I found it so hard to win that I reinvented everything. Just when I got it working, got even and winner for the year so I could actually withdraw money, winter was upon us.

In early 2019 everything changed again.

That time, I decided to do something I'd never done before. That's when I really started trying to do the reverse of what everyone else was doing. Pretty much literally.

I'd always done some form of pace handicapping; building models (automatically in HSH) but without ever selecting a paceline. Haven't done that in maybe 20 years.

So, what I did was turn the tables. I used the pace handicapping to set the odds and bet into those odds with "how the public bets."

BTW, the biggest difference between those 2 approaches is that one (the pace) is what the public does and the other (projected tote) reflects what the whales do.

IOW, I wanted to EMULATE the whales betting into the HANDICAPPING.

(For those of you who chalk this up as the ravings of a lunatic, it worked.)

(Emphasis on the past tense was intentional.)

As I always do, I went public with this to my users in our annual seminar. (Not free, but very reasonable, cost considering all the work I put into it.)

For those who believe that nobody would sell anything that works... well, get this... It is how I have made my living for the past 30 years. I build stuff that works and find a way to get paid for it.

Then, winter 2019 came. As has been the case for a number of years, many of the systems go dormant from just before Thanksgiving to about the last week in January.

My belief is that trainer patterns change because the big boys simply take the holiday season off and let their assistants do most of the work. They also rest their horses.

They go back to work seriously just after New Years but it takes 3 weeks or so for things to get back to normal.

Trivia: Field sizes are actually larger in the winter and speed ratings do a better job of pushing winners to the top ranks. I wouldn't have bet on either of these being true if I'd not run the tests myself.

Just as The System was coming back (late Jan thru early March, 2020) Covid-19 shut everything down. What stayed running just made no sense!

The strategies were scenario-based. Those strategies were based upon the number of horses projected to be in the "1st tier" of the tote board projections. (Remember the tote projection was based upon the pace handicapping.)

Most of the strategies were literally reversed! Some scenarios called for betting overlay(s) and others called for underlay(s). A couple led to chaos races where we just bet the longest odds horses.

I've spent the last several months working all of this out. Started winning again in July, but it was still hit or miss.

I think I've got it dialed in now, but we'll see for sure in February. Meanwhile, no seminar for the guys this year. However, I am recording the current system that really seems to be working. It will be released to the guys who purchased last year's seminar as a free upgrade this week.

Got a little long-winded. It seems to be my nature. LOL

I understand that the final odds are fairly efficient and that the collective mind of all horse players (which includes sophisticated computer models, better informed insiders, and a lot of very smart individuals) is probably going to outperform the odds line any individual can make.

That said, this game is still about finding things that are NOT being reflected on the board.

Two of the most important scores of my life came a few decades ago when I knew about 50% of what I know now. The thing is, the one thing I did know was that trainer "x" didn't try with first time starters. However, 2nd time out, especially if he added blinkers, the horse would often make a huge jump up and win at a big price.

I wasn't building odds lines, watching money flows, or anything else. I had a piece of information that I was close to 100% sure wasn't being reflected on the odds board. The handful of bombs I caught on the biggest bets of my life up to that point FAR OUSTRIPPED the losses on everything else I was playing at that time.

I'm not sure that trainer patterns like that exist anymore with all the databases and Formulator screening tools that exist now, but I think the key for the average guy like me is not in greater complexity building odds lines. It's in finding those rare situations where you are almost certain the public and money flows are wrong.

Probably the most important trainer patterns in the game today are short term hot and cold streaks. Some very talented players have been able to recognize these trends early on in a meet. Doing this can swing odds in your favor. Being able to throw out known trainers with good patterns when they are cold is a talent and not easy to do. Some players pick this up by watching replays and are able to determine a certain trainers horses are not running as expected. Then they can tell when it seems to turn around and adjust accordingly. Some of these players just would throw out the trainer in their pick bets unless there was some sort of extenuating circumstance. Usually a very small field or a trip you can't ignore. With so many players betting multiple tracks, this talent seems harder than ever.

I don't have a problem with that approach other than I've always found it difficult to separate random hot and cold streaks from legitimate issues that could impact a trainers results.

For example:

In the spring when a lot of horses are coming off a layoff, a trainer might have all his horses ready off the layoff or they all might all be a race or a few works short. You can pick up on something like that.

Maybe a bug is going around in a particular barn and a bunch of horses missed some training and are running below par, but it's hard to know that.

Maybe a trainer freshened up a bunch horses for an assault on a particular meet like Saratoga.

Maybe a trainer has a new vet (ahem).

But a lot of the time I think it's just random noise. The horses may be running fine. They just aren't good enough.

Focus on 3 choices? Sure, but it doesn't matter if you focus on the field of 12 or 3 - you start with the field and handicap your way down to say 3. Schwartz always liked idea of half field plus 1 are the contenders - makes no diff - you still gotta make a choice of which one to bet or how many to bet.

You only make money when you differ from the public at a high enough strike rate and at at a price to overcome the take and breakage plus a positive return. So you gotta be at least 20% better. You need to find the races where Juan Q Publico is off by 20% or more on horses you like.

You pick 123 public picks 234. You figure 1 at 35%, public says 10%. That is what you want, obvious but every race isn't playable unless you have a computer based team.

Not sure why "streaks", be it a trainer or a gambler, as apart from the dynamic factors that are always determining outcomes, are anything more than the Gambler's Fallacy...

https://www.businessinsider.com/the-...20the%20future.

Is that the margin of difference you look for? Roughly a 250% difference? Or would 100% be enough?

Do you ever consider the reverse of that? Meaning, say the ML and your own line has a similar odds line of say, 8%(12/1), the public(whoever that is) says it's 20%(4-1). Does that have any interest to you?