Vito raised a couple of issues.
First, about \"pairs\" (really meaning horses running in a tight figure range). The reason Vito has the view that horses running in a tight range is a negative is that as he himself says, he doesn\'t know anything about how figures are made, and doesn\'t want to.
All speed figures are made using the past figures of the horses, whether they are crude ones (made from pars), or much more sophisticated ones incorporating ground loss and weight, made by the \"projection method\", which just means looking at the figure histories of the individual horses you are making figures for today. That\'s the whole basis for not only making but using figures-- if there isn\'t correlation between figures horses have run in the past and figures they will run in the future, what\'s the point of the exercise?
But if that premise IS correct it means, by definition, that THE MORE HORSES RUN BACK TO PREVIOUS FIGURES, THE MORE ACCURATE YOUR FIGURES ARE.
Now, different figure makers have different ideas when it comes to figuring out how \"fast\" the track is-- I cover that in Changing Track Speeds (Archives of this site). Ragozin as far as I know was the only figure maker who assumed that tracks stayed the same speed all day, unless there was rain, a freeze, or a thaw (he says so in his book, co-written with Friedman). From the 2014 Derby numbers they have just posted it looks like Jake believes that too, more on that later.
But on one thing we all agree-- the relationships between horses WITHIN a race are fixed, in figure terms. For Beyer the relationship between the winner and the others, and the others with each other, is fixed by beaten lengths. For those making performance figures it\'s by beaten lengths, weight differentials, and ground loss differentials. (NOTE-- the above does not apply to those building pace into their figures, they can give a horse more or less credit than someone else in the same race).
That paragraph above has a very important ramification-- if a horse has a lot of 4\'s, and another has a lot of 8\'s, a figure maker can\'t give them a 4 AND an 8 UNLESS THE RELATIONSHIP, AFTER LENGTHS, WEIGHT AND GROUND, WORKS OUT TO 4 POINTS. Otherwise you can give one a 4 or the other an 8, but not both. You can\'t pair up multiple horses in the same race unless the relationships AND figure histories support it.
Now, horses come out of different races and tracks-- it\'s not like the same field runs against each other every time. So, if a set of figures has lots of horses running in a tight range, it means one of two things-- either the figures for all the races they come out of are right, and the data base is tight, or the figure maker is breaking a cardinal rule, screwing with the relationships between horses in races. Now, why would someone do this? It would make the figures LESS accurate, making it harder for people using them to win (including the figure maker, who in my case also recommends horses for purchase, by the way). So logic leaves the first possibility.
As an extreme example of what you NEVER want to see, Jake recently posted their figures for the 2014 Derby. Not a single one of the 20 (TWENTY) horses ran within a point either way of their previous top. One ran better, 19 ran worse (stake level 3yos, no less). The past figures had NO correlation with the figures assigned for the Derby. And I would love to hear Jake, Len F or Eric explain how they got to that result. On the face of it, they clearly didn\'t use the Derby horses to make figures for the Derby horses.
tgjb your post was informative and insightful. I have no expertise but value your figures and find them to be most trustworthy.
I do however still believe that in a 20 horse derby field the great majority of horses will regress simply because of the traffic and additional distance. Very seldom can you pick a horse who at some point in the race is compromised by traffic. I\'m not speaking about having to go wide i\'m talking about stuck behind horses and can\'t run or shut off from a hole or knocked sideways ( vicars in trouble, intense holiday etc)
I regularly use thorograph and not ragozin simply because I like the additional data and manner of presentation, but I\'m in agreement with Jake that most if not all of the derby horses regressed.
Most and 19 out of 20 are two different things. We had 13 run more than a point off their tops. That\'s most.
Anybody who is remotely serious about betting shouldn\'t accept any figure at face value. It has to pass a sniff test. Most on this board have seen 1000s of races and can make their own judgments about performances.
I am not a shill for TGJB, as I have openly questioned multiple figures and the entire California circuit. That said, the Rags figures for the derby and for many of the preps leading up to it, are downright stupid. Chrome running a 7.5 in the derby. Still not back to his 2 year old top.
It is astonishing to me that the rags clientele doesn\'t see those problems and question them. I have to believe that people that paying 35 bucks for figures and betting real money have enough acumen to do some type of analysis.
Some of their figures are really indefensible.
New poster here. Love this topic so had to chime in. So you calculate the relative figures of each horse based on lengths beat, ground loss and weight. Then you slide this set of figures up and down to best fit the previous figures for the horses. And this is called the projection method.
You don\'t actually project a figure for the horse before the race and use that as a reference point to create the best fit. Am I correct? If you did it would be awesome to publish that projection in the page for each horse.
It would also be cool to create a projection and a variance for each horse pre-race. That variance being a function of the horses age, rest, trainer, previous consistency, etc.
I understand you guys are the experts, not me. But with this kind of information you could start to create a line off the projections and variants. Could even be a new product you sell once you prove it\'s worth.
Thanks for your time
Jeff
No , we don\'t actually project a figure. Don\'t know if you have looked at the Thoro-Pattern on each horse, but it lives in the general neighborhood you describe.
I do and that\'s very useful but it is independent of the horse, trainer and time since last race. It also wouldn\'t factor in surface change, distance change, equipment change, etc. With enough time I could come up with a program to do it for me but would need data in .csv format. With a projection and a variant for each horse you can simulate the outcome on computer a million times. You can even throw in additional factors related to pace in the simulation. Seems like a powerful way to extract more value off the TG figs. Okay, I\'ll go back in my cave now.
P.S. Would you ever sell your data in .csv format?
Back a few years ago an acquaintance of mine (friend of a friend) got pretty loaded in a bar in Michigan with a few others one night, and they decided to road trip to watch the annual rock, paper, scissors tournament. This is a true story bye the way. So after barhopping across a few states, the conversation morphed into the 3 of them entering. And one of them wound up winning the darn thing, no kidding.
Now there is also a guy out there, right now, who has mathematically modeled the game of rock paper scissors and feels there are a few basic profiles that all rock, paper, scissors players fall into. And he swears he can very accurately predict what another player is going to do next based on figuring out their profile and looking at the last 3 choices they have made. If you google it he may come up. But anyway, he claims like an 80% plus accuracy rate. And he may be right. Now i would say he probably has an edge against me and my buddy\'s buddy if we play him heads up. But he\'s never won the tournament and my buddy has.
The point i\'m trying to make is that it is important and valuable to have good data and be able to interpret that data correctly. But there will never be a computer program or a pattern that will be the answer. This is and always will be a game of percentages. These are flesh and blood animals and they don\'t always do what the pattern \"says\" they are supposed to do. That\'s why you have to really appreciate it when a good plan comes together. And over the long haul, you can lose the race but beat the races.
Great story.
So you don\'t believe a computer could accurately predict the probability of a horse winning a race better than a human? There are entire industries that project results with computers for profit. My theory is a Monte Carlo simulation could yield a probability which when wagered on when a significant edge exists would yield profits in the long term. So I think we are in agreement with the long term goal via probabilities to beat the races not the race
Miff please direct this gentleman to the Math/Computer section of the Bowery; 3rd burn barrel to the left please. REALLY!!!
I feel a Rocky post coming...
Rocky is a completely different animal.
Beaz,
Brought this subject up a year so ago.
I let it go because there was little interest.
My thinking is that using a database of performance figures, one could theoretically create an algorithm that would identify patterns. New tops, bounces, pairs etc..... whatever metric you wanted.
The handicapper does not have to identify the patterns. Many \'hidden\' patterns might emerge.
Download the entries for a race and the program will identify those with forward moving patterns....etc.
Analytics and Algorithms
Whether there is a \'volume\' wagering opportunity ...I do not know. Some young whizzes employed by SAI (they do a lot of high level government work) told me that it could be done....
Bob
No. I don\'t think a computer could pick winners better than a human. Just like i don\'t think a computer could predict what i am going to do next better than my wife. And she\'s been trying for 15 years.
Frank D,
GOD came up with the perfect algorithm to beat the game ,he\'s with the math guys et al hovering around the barrel of fire on the Bowery.
Mike
Jerry-- LOL
beaz-- If you\'re interested in creating your own probabilities (fair odds line), check out the link in this post for a story about some who have done it with spectacular success: Contingencies (https://www.thorograph.com/phorum/read.php?1,82525,82566#msg-82566).
If you\'re really ambitious, I\'d suggest reading the B.Benter paper in the book in the link in this post as well: Efficiency of Racetrack Betting Markets (https://www.thorograph.com/phorum/read.php?1,82525,82568#msg-82568).
Be forewarned though, Miff is partially right - for every B.Benter and A.Woods that have succeeded in beating the game with a better probability line than the public\'s, there are countless others who have tried to emulate them and failed miserably because they either miscalculated their edge or overbet their bankrolls. Benter and Woods spent 10 man-years and wrote over a million lines of code, and went through their entire $100K initial stake before they finally got their R-squared high enough to produce a consistent profit with their probability lines.
Best,
Rocky R
Thanks for the links
You could develop a program like this, but the end result would be a fair odds line of YOUR OPINION OF THE RACE. The odds line that the program spits out is a direct function of the data you input, and the data to input is not always clear cut. Too often (like on every single horse), you have to make a decision as to what \"top\" to use--recent or lifetime, turf or dirt-- or which probability distribution to use (%top, %pair, %off, %x). Those decisions reflect your opinion.
I have toyed with this (more like \"obsessed over this\") for about 20 years now. I speak from experience. For a short period, a friend and I bet using the same simulation program. 9 times out of 10 we would land on a different set of key horses. The reason for the differences always boiled down to how we \"read\" the horses in question. Do you use the Turf top on a Turf race? Always? Do you use the 90+ trainer thoro-pattern for a layoff horse? What if it\'s a late 2yo and you feel the horse is ripe to explode off the layoff but the thoro-pattern for 90+ (which includes ALL AGES) does not really indicate that?
At the end of the day, whoever read the race right (OPINION) ended up cashing.
My point is that even with a magic black box, there will always be a human element to the handicapping game.
Good luck to you.
Tommy and other respondents,
Based on the content of the responses, I am unsure if they were directed at my post, Beaz or both.
For clarification purposes. I suggested that by mining data one may be able to discover previously undiscovered patterns. After all, many sheet players constantly look for patterns (i.e pair of 2yo top 1st or 2nd out 3yo start) ... and many sheet players apply these patterns in their handicapping.
What if, via analytics and algos, one could identify these yet to be discovered patterns. These patterns might be explosive to the upside or disastrous to the downside (patterns that result in Big X\'s).
Has nothing to do with projecting a morning line etc.
One would apply this information to assist in their handicapping.
Regards
Bob
Sorry, Bob,
I do not post often. My response was more directed to the bigbeaz posts. I will try to take more care in the future.
tommy