Handicap The Late Money

[jay68802]

How would you go about doing this? You need the pool totals for a certain time, lets say when the first horse loads into the gate. You need the final pool total. Track the horses that had been bet at 3-1 or lower in the last dump.

What else?

[GMB@BP]

Originally Posted by jay68802

How would you go about doing this? You need the pool totals for a certain time, lets say when the first horse loads into the gate. You need the final pool total. Track the horses that had been bet at 3-1 or lower in the last dump.

What else?

how is it even possible when all the late money doesnt even hit till 2nd flash or so 30 sec to a min into the race?

[Dave Schwartz]

Originally Posted by jay68802

How would you go about doing this? You need the pool totals for a certain time, lets say when the first horse loads into the gate. You need the final pool total. Track the horses that had been bet at 3-1 or lower in the last dump.

What else?

It could certainly be done AFTER the race is over. The question is, "How could you use it BEFORE the race goes off?"

Peter Asch and Richard Quandt wrote a book back around 1986 where they studied tote movement for every race over an entire season at one track. (Of all places, it was Atlantic City. Go figure.)

Their purpose was to determine if the "late money" was "smarter" than the rest. Specifically, they defined "late money" as "money wagered in the last 5 minutes."

There hypothesis - namely that later = smarter - proved to be correct, but not enough to be profitable.

What they did (in today's equivalent) was take a snapshot of the tote at the first flash of 5 minutes to post and compute the precise odds on all the horses.

The repeated the process after the race was finished and built odds based upon the money that came down AFTER the 5-minute snapshot.

The "Whale Era" equivalent would be to take the snapshot as the 1st horse entered the gate so that (theoretically) you still had time to make a wager.

Of course, the rub is that the other information - the important part - isn't available until the horse is in the winner's circle.

Hope this helps.

Dave
PS: Were it me, the actual snapshot I would want would be the picture of the actual winner. Just sayin'.

[ccmd00d]

Hey Guys,

I've been building a script for fun to pull tote information every 30 secs from my ADW. As an example see the attached data on the Favorite (and winner) from Pompano Park R3 a few min ago. I like to play around with small pools because it is easier for me to see the money move. Now with my particular ADW, the final pool did not come in until a min or 2 after the race was off (indicated by the -1 in the MTP column.

In this case the Win pool for the #6 tripled. Sure seems like the late money was "smarter" in this case.

[JerryBoyle]

Originally Posted by GMB@BP

how is it even possible when all the late money doesnt even hit till 2nd flash or so 30 sec to a min into the race?

That's the whole reason we'd want to predict the late money. We can't wait until it has been bet, so we have to model what we think it'd be. From a modeling perspective it doesn't matter when it hits, you just need the prices at the time YOU would make your bet (e.g. first flash of 0 mtp), and you need the final price (the variable you're predicting at 0 mtp).

I've been toying around with trying to model the final price of each horse for a while now. As stated above, my inputs are the price at whatever time I bet + any other variables I think may be predictive of what will sway the public's perception of the horse at that point on, which is any handicapping factor one might use in his own fundamental model to make a prediction. And the variable I predict is the final probability. So far, I can do slightly better than the public's guess as of 0 mtp, but not enough to improve my roi.

The whole exercise has provoked a bit of a thought experiment for me: if you're modeling those who themselves are using models and features which are inputs to the models, can you do so without recreating a version of that model? Given how much money is bet by the CRW teams, and how smart that money is, the question stated another way is, can we model the output of the CRW teams' models without actually coming up with a version of their model? My guess is that there is probably some math proof or theorem that says you can't. This leads me to believe that if you actually had a way to successfully handicap the final price, regardless of how you got there, you'd have to be at least as good and as accurate as the CRW teams.

[jay68802]

Originally Posted by jay68802

How would you go about doing this? You need the pool totals for a certain time, lets say when the first horse loads into the gate. You need the final pool total. Track the horses that had been bet at 3-1 or lower in the last dump.

What else?

I get the feeling I was not clear enough on my idea. What I want to do is build a data set of the horses the late money falls on. Then build or create a program that will pick the horses most likely to get hit by the late money.

Why not try to use what they are doing for improving my handicapping?

Can it be done, and how?

[jay68802]

Originally Posted by JerryBoyle

That's the whole reason we'd want to predict the late money. We can't wait until it has been bet, so we have to model what we think it'd be. From a modeling perspective it doesn't matter when it hits, you just need the prices at the time YOU would make your bet (e.g. first flash of 0 mtp), and you need the final price (the variable you're predicting at 0 mtp).

I've been toying around with trying to model the final price of each horse for a while now. As stated above, my inputs are the price at whatever time I bet + any other variables I think may be predictive of what will sway the public's perception of the horse at that point on, which is any handicapping factor one might use in his own fundamental model to make a prediction. And the variable I predict is the final probability. So far, I can do slightly better than the public's guess as of 0 mtp, but not enough to improve my roi.

The whole exercise has provoked a bit of a thought experiment for me: if you're modeling those who themselves are using models and features which are inputs to the models, can you do so without recreating a version of that model? Given how much money is bet by the CRW teams, and how smart that money is, the question stated another way is, can we model the output of the CRW teams' models without actually coming up with a version of their model? My guess is that there is probably some math proof or theorem that says you can't. This leads me to believe that if you actually had a way to successfully handicap the final price, regardless of how you got there, you'd have to be at least as good and as accurate as the CRW teams.

My question to you is why try and predict the final price, without knowing what horse or horses they are going to bet?

[JerryBoyle]

Originally Posted by jay68802

I get the feeling I was not clear enough on my idea. What I want to do is build a data set of the horses the late money falls on. Then build or create a program that will pick the horses most likely to get hit by the late money.

Why not try to use what they are doing for improving my handicapping?

Can it be done, and how?

This idea is what I was referencing with the "thought experiment" comment. I don't know the answer, but intuitively, to me, if you're doing the above, you're in effect producing a replica of the models used by the large teams. If you're not predicting the price, and just trying to predict a yes/no answer to the question, "will late money fall on this horse", it might make the problem simpler, but I'm not sure and I lean towards doubting it. Perhaps it's easier to model, "is this a race that lends itself well to being handicapped by some model"?

With regards to your other question, doesn't being able to predict the final price assume that you DO know which horses they're going to bet?

[jay68802]

Originally Posted by JerryBoyle

With regards to your other question, doesn't being able to predict the final price assume that you DO know which horses they're going to bet?

I think we are on the same page here, I look at it as finding the horse as the handicapping part, predicting the price as a wagering decision. In other words, to me it is no use to predict the final price if I can't find the horse.

And if I can find the horse, predicting the price will be made easier. If you know that 40% of the pool is going to come in after the first horse loads, and that 50% of that money will be on 2 horses, you have a better chance of figuring the final price on those horses.

[davew]

Originally Posted by jay68802

I get the feeling I was not clear enough on my idea. What I want to do is build a data set of the horses the late money falls on. Then build or create a program that will pick the horses most likely to get hit by the late money.

Why not try to use what they are doing for improving my handicapping?

Can it be done, and how?

Some tracks have daily doubles on every race. The probable pays from the top couple favored in the previous race (doesn't matter who won) can be used to make a good estimate of final odds of race in question (second race in double).

Very late in the betting, the odds on most horses correct towards the daily double estimated probables.

Added to this are more people (relatively, since the betting population is declining) are watching warm-ups, horse condition, demeanor, and gate loading to adjust their betting/thinking on this race.


I wish you luck, but I feel many 10/1's that get late bet to 6/1, are actually 6/1's that have not been bet enough before the race.

When building a 'data set', you might find certain stables bet early / late / never on their horses which could help in determining live / dead horses in the race.

[Redboard]

Originally Posted by ccmd00d

Hey Guys,

I've been building a script for fun to pull tote information every 30 secs from my ADW. As an example see the attached data on the Favorite (and winner) from Pompano Park R3 a few min ago. I like to play around with small pools because it is easier for me to see the money move. Now with my particular ADW, the final pool did not come in until a min or 2 after the race was off (indicated by the -1 in the MTP column.

In this case the Win pool for the #6 tripled. Sure seems like the late money was "smarter" in this case.

Depends on whether there was an odds drop or not, but your chart doesn't show that. Since the horse ended up at 7/10, there probably was.

A chart showing the amount bet (to win only) on all of the starters over time would give a better illustration of whether the late money was "smart."

[formula_2002]

from another thread to which I posted today

here is a situation I just ran into which is not so infrequent.

race 5 at Epson, the 4/1 exacta paid $31.10, the $1 win bet returned $2.80

at 1 minuet too post the exacta pool was $3,740, the final pool was $11,071

using the 1 min to post exacta pool data I constructed a $1 payout of $4.04 on the winning combination, while the final pool data construct had a $1 payout of only $2.87 (which by the way as you can see, beat the straight pool $1 return.)

here is the 1 min to post construct indicating "your choice" as #4 over all.
the idea is, if your win bet choice is the #4, then you should bet the 4 over all, using the bet sizes indicated.Code:

			"UK Epsom
Race 
5

Post 
11:25

MTP 
1

$1 Exacta Pool
$3,704"	
	YOUR CHOICE	OVER	pgno	bet size
			1	13
			2	3
			3	21
			4	0
			5	27
			6	11
			7	20
			8	4
			9	0
			10	0
			11	0
			12	0
			13	0
			14	0
			15	0
			16	0
				100__________________

[AltonKelsey]

With a pool that tiny, the odds can shift wildly after you've bet.

Nothing wrong with betting the exactas if they offer better value than the win pool (used to happen more often), but this is a bad example.

[Tape Reader]

If we are talking about late money (money bet before the bell rings), IMO, it can often be anticipated by charting 20 MTP to PT.

Some of the changes that I look for are the following:

If a horse opens >ML it is a candidate.

If a number stays the same odds as other horses rise in price in a flash, it is a candidate.

If a horse moves down in odds while all the others stay the same, it is a candidate.

If a horse moves down in odds while all others stay the same or move higher, it is considered a “power bet” and should not be ignored.

Saratoga, IMO, is the best track for reading the tote.

[formula_2002]

tonight's dd r3

final ex pool $16,792 my $1 ex calculated return $6.71
1 min, $5471, my return $7.22
2 min, $4080, my return $6.91
3 min, $2735, my return $7.26
4 min $2116,, my return $7.13

actual $1 win bet returned $6.00
9/6 exacta $38.00

in this case, my return outperformed the win bet return at all listed intervals,
the best being at 3 min to post.

to get a better idea of my approach, see my recent picks in the "selection forum"

		"Delta Downs
Race 
3

Post 
8:04

MTP 
1

$1 Exacta Pool
$5,471"	
9	OVER	pgno	bet size
		1	7
		2	2
		3	19
		4	5
		5	16
		6	19
		7	16
		8	4
		9	0
		10	12
		11	0
		12	0
		13	0
		14	0
		15	0
		16	0
			100