PDA

View Full Version : cramer or massa study help


formula_2002
12-10-2004, 06:50 AM
Can some one post the explicit location where I can find the following study;

"researchers as Jim Cramer and Ken Massa have shown conclusively that 4-5 shots without early speed in dirt races will win less often, and lose more money, than 4-5 shots with early speed."

Thanks
Joe M

hurrikane
12-10-2004, 12:11 PM
joe,
you could try the htr2.com site and look through the newsletters. There are a lot of studies in there. You could also email Ken from that site and he would probably give you some direction.

formula_2002
12-10-2004, 01:54 PM
Originally posted by hurrikane
joe,
you could try the htr2.com site and look through the newsletters. There are a lot of studies in there. You could also email Ken from that site and he would probably give you some direction.

Thanks Hurrikane, I'll follow it up

Tom
12-10-2004, 11:12 PM
Formula,
Barry Medow printed a lot of Cramer's studies in his newsletters-if You have access to them, you might find it in there. I have many of them and will take a look for you-let you know Saturday afternoon if I find anything.

formula_2002
12-11-2004, 08:19 AM
Originally posted by Tom
Formula,
Barry Medow printed a lot of Cramer's studies in his newsletters-if You have access to them, you might find it in there. I have many of them and will take a look for you-let you know Saturday afternoon if I find anything.

Thanks Tom. I did go Meadow's and Massa's web site but was unable to pick out what I was looking for.
Massa did send me his lateset news letter which had to do with all "extra special fav.

Oddly, I had posted my latest series of picks on PA based along similar facts.
Best Bris speed,
MSW only
ML <=2-1
and the key was my own oddsline<=1

It turned out to be a good place bet, as did Massa's study.

RonTiller
12-12-2004, 11:39 AM
Hi,

I too don't know where the "conclusive" studies are that you refer to but I have the data at my fingertips to answer your question.

First, the statement "...4-5 shots without early speed in dirt races will win less often, and lose more money, than 4-5 shots with early speed." needs a little cleaning up. Since the odds are being held constant, at 4-5, any subset of horses that wins less often at 4-5 will by definition lose more money.

Second, "without early speed" is itself subject to all the vagueness, fuzziness and definitional infighting (both petty and subtantive) that handicappers in general so love to engage in. I'll just do an end run around all that and present the win percentages by several early speed measure rankings. I'm doing these particular measures because I have them at hand for all races since 1991. I have no fight with anybody who has their own favorite way of defining, measuring or evaluating "early speed".

Third, as a methodological point, I took all races where horses went off at final time odds between (and including) 0.7 to 0.9

1. All 4-5 horses since Jan 1, 1995 | Dirt Sprint | 7 horse fields (to normalize the win percentages) | Fastest 1/4 mile time in the last 10 races, ranked (1 = fastest horse to the 1/4, based on last 10 races, 2 = 2nd fastest to the 1/4, etc.

Results:
1st.....49% wins.....3472 starts
2nd....48% wins.....2287 starts
3rd.....47% wins.....1602 starts
4th.....44% wins.....1303 starts
5th.....42% wins.....922 starts
6th.....39% wins.....634 starts
7th.....41% wins.....421 starts

2. All 4-5 horses since Jan 1, 1995 | Dirt Sprint | 7 horse fields (to normalize the win percentages) | Fastest 1/2 mile time in the last 10 races, ranked (1 = fastest horse to the 1/2, based on last 10 races, 2 = 2nd fastest to the 1/2, etc.

Results:
1st.....50% wins.....3816 starts
2nd....48% wins.....2181 starts
3rd.....44% wins.....1398 starts
4th.....46% wins.....960 starts
5th.....40% wins.....648 starts
6th.....37% wins.....403 starts
7th.....41% wins.....237 starts

3. All 4-5 horses since Jan 1, 1995 | Dirt Sprint | 7 horse fields (to normalize the win percentages) | Fastest 1/4 mile time in the last race, ranked (1 = fastest horse to the 1/4, last race, 2 = 2nd fastest to the 1/4, etc.

Results:
1st.....50% wins.....3376 starts
2nd....48% wins.....2062 starts
3rd.....46% wins.....1403 starts
4th.....45% wins.....1045 starts
5th.....41% wins.....754 starts
6th.....44% wins.....533 starts
7th.....42% wins.....339 starts

4. All 4-5 horses since Jan 1, 1995 | Dirt Sprint | 7 horse fields (to normalize the win percentages) | Fastest 1/2 mile time in the last race, ranked (1 = fastest horse to the 1/2, last race, 2 = 2nd fastest to the 1/2, etc.

Results:
1st.....49% wins.....4167 starts
2nd....47% wins.....2058 starts
3rd.....47% wins.....1251 starts
4th.....44% wins.....775 starts
5th.....40% wins.....568 starts
6th.....42% wins.....389 starts
7th.....39% wins.....260 starts

Since the win% of 4-5 horses varies very little with field size (something I didn't expect, the following is all races, regardless of field size:

5. All 4-5 horses since Jan 1, 1995 | Dirt Sprint | All field sizes | Fastest 1/2 mile time in the last 10 races, ranked (1 = fastest horse to the 1/2, based on last 10 races, 2 = 2nd fastest to the 1/2, etc.

Results:
1st.....50% wins.....16,737 starts
2nd....47% wins.....9,357 starts
3rd.....45% wins.....5,939 starts
4th.....43% wins.....4.086 starts
5th.....40% wins.....2.683 starts
6th.....40% wins.....1,597 starts
7th.....37% wins.....872 starts
8th.....39% wins.....407 starts
9th.....35% wins.....165 starts
10th...31% wins.....77 starts

To reiterate, I'm not trying to convince people to use these "early speed" markers or that they are better or worse than anything else. I had the data available with no additional work. As long as the vague and fuzzy phrase "early speed" is at issue, I guess there will never be a "conclusive" demonstration of anything - you just defined it wrongly.

Sigh...........

Ron Tiller
HDW

formula_2002
12-12-2004, 11:51 AM
thanks, RonTiller.

I just printed it out and will digest it.

thanks
again

formula_2002
12-12-2004, 12:54 PM
Ron,
It may appear to some that the differences in the winning % for ranks 1 to 7 are due to the public’s error.
However, before I can make a judgment about that, I would have to see the expected win % adjusted for track take-out compared to the actual win %.
It would mean summing the 1/(odds+1) for all horses in each race and dividing that sum into the 1/(final odds+1) of the horses in question (>=.7 to 1 and <=.9 to one)

Your measures of “early speed” seems reasonable to me.

The statement "researchers as Jim Cramer and Ken Massa have shown conclusively that 4-5 shots without early speed in dirt races will win less often, and lose more money, than 4-5 shots with early speed." Is not mine.
I simply wanted to see the back-up for that statement before evaluating it.

Thanks again for your effort.
Would it be possible to post the “expected” win %’s

Joe M

hurrikane
12-13-2004, 07:55 AM
you da man Ron.

i'm wondering if early speed would not be better judged by who actually had it this race instead of who was thought to have had it. Doing the same study by horses position at the 1st and 2nd call in the race. This would tell you the effectiveness of early speed...no?

I'll run a few things myself but I aint got your data, that's for sure.

sjk
12-13-2004, 08:58 AM
A bet on the horse who has the lead at the first call has a huge positive return irrespective of odds. If you can find a way to get that bet down you need look no further.

Larry Hamilton
12-13-2004, 09:23 AM
You are correct, it will show a huge advantage, but NOTHING has the advantage like a horse who was leading at the top of the stretch.

In a 6 furlong race, saying that I will give this hoss a 2 length lead and cut 2 furlongs off the race (1st call), actually makes sense he would win most of the time. The later the call you use, the more likely the leader will win.

formula_2002
12-13-2004, 10:12 AM
Originally posted by sjk
A bet on the horse who has the lead at the first call has a huge positive return irrespective of odds. If you can find a way to get that bet down you need look no further.

A hugh positive return? I dont thinnnk so.
Only if it were true, I'd stilll be playing...

RonTiller
12-13-2004, 11:50 AM
FYI - I just looked at all the running lines since jan 1, 2002. Out of 177,711 horses who had the lead at the stretch call, 116,759 won, for 65.7%. I do take exception, though, to the statement "...NOTHING has the advantage like a horse who was leading at the top of the stretch.". Taking the same 3 year sample of races and applying the most sophisticated statistical techniques I am aware of, 99.9% of the horses on the lead at the wire won (alas, some horses were disqualified). Lets just cut out the middle man!

These kinds of queries about various factor's win percentages based on "...who actually had it this race instead of who was thought to have had it" seem to me to be pointless. In the final analysis, this is a game of prediction - if you can't predict it, so what? But if it opens up some avenue of research that ultimately leads to better predictions, I guess it's worthwhile.

Although the results I posted were for horses going off at post time odds of 4-5, the same staircase effect is in place for early speed no matter what odds range you look at (sorry, no matter what odds range at which you look). Horses going off at 20-1 show the same descending win percentages based on the various early speed factors as the 4-5 horses do. Likewise 10-1, 30-1, 5-1, etc. This should come as no surprise to anyone. For instance:

1. All 10-1 to 15-1 horses since Jan 1, 1995 | Dirt Sprint | All field sizes | Fastest 1/2 mile time in the last 10 races, ranked (1 = fastest horse to the 1/2, based on last 10 races, 2 = 2nd fastest to the 1/2, etc.

Results:
1st.....7.4% wins.....29,367 starts
2nd....6.8% wins.....36,987 starts
3rd.....6.5% wins.....42,006 starts
4th.....6.3% wins.....45,990 starts
5th.....6.0% wins.....47,673 starts
6th.....5.4% wins.....45,901 starts
7th.....5.6% wins.....37,822 starts
8th.....5.4% wins.....26,000 starts
9th.....5.2% wins.....15,639 starts
10th...5.1% wins.....9,035 starts

Notice that with the 4-5 horses, the fasteset 1/2 mile last 10 not only had the highest win% but it had by far the most instances. In fact, they drop off dramatically, each successively slower horse being bet less infrequently down to 4-5. I haven't plotted it out but it is definitely not linear. Recall that:

4-5 horses
1st.....50% wins.....16,737 starts
2nd....47% wins.....9,357 starts
3rd.....45% wins.....5,939 starts
4th.....43% wins.....4.086 starts
5th.....40% wins.....2.683 starts
6th.....40% wins.....1,597 starts
7th.....37% wins.....872 starts
8th.....39% wins.....407 starts
9th.....35% wins.....165 starts
10th...31% wins.....77 starts

With the 10-1 to 15-1 horses, although the fastest 1/2 mile horses win the most, the line moves in the opposite direction and the slope is not nearly as steep. In other words, for 4-5 horses, #1 ranked 1/2 mile has 16,737 horses bet to 4-5 and #6 rankes has only 1,597 horses bet to 4-5, a big difference. With the 10-1 to 15-1 horses, the #1 rankes 1/2 horse had 29,367 starts and the #6 ranked horse had 45,901 starts, a much smaller differential and a much less steep slope.

What does all this mean? Where's Barry Meadow when you need him?

Ron Tiller
HDW

formula_2002
12-13-2004, 11:58 AM
Originally posted by RonTiller
FYI -
What does all this mean? Where's Barry Meadow when you need him?

Ron Tiller
HDW

Wonderful study Ron.. Get in touch with Barry

The statement ,"researchers as Jim Cramer and Ken Massa have shown conclusively that 4-5 shots without early speed in dirt races will win less often, and lose more money, than 4-5 shots with early speed." was his in response to an e-mail I sent to him.

Joe M

Larry Hamilton
12-13-2004, 12:02 PM
no need to be insulting..a simple "butt out of my conversation" will do...

formula_2002
12-13-2004, 12:04 PM
Ron, perhaps you would commnet on the following (it's a copy of the e-mail I sent to Barry

Horse racing odds vs. Natural odds.

Based on my data;
Horses that run at 1-1 odds with an average track take-out of 15.6%, win 42.5 % of the time.
In groups of 10 races each, there were 93 of 420 groups where the number of wins exceeded the number of losses (22% winners).

Testing for natural odds of 42.5% with a random number generator for groups of 10 trials, 97 of 420 groups “won” for an average of 23% winners.

In this single test, it would appear that there is little or no separation (deviation) between natural odds and track odds.

I would propose, where there is little or no deviation between natural odds and track odds it would be impossible to achieve a profit at the track that would meet any meaningful significance test.

sjk
12-13-2004, 12:28 PM
A hugh positive return? I dont thinnnk so.

I get an ROI of +58% for all horses leading at the first call. What do you get?

formula_2002
12-13-2004, 01:21 PM
Originally posted by sjk
I get an ROI of +58% for all horses leading at the first call. What do you get?
Just go to any results chart. About 90% of those beast that were 1st at the 1st call and lost would have been mine..:rolleyes:

RonTiller
12-13-2004, 06:35 PM
Larry, I guess your comment "no need to be insulting..a simple "butt out of my conversation" will do..." was directed at me. Ouch. This is not anywhere near anything I meant. Perhaps you took a facetious comment on my part and interpreted it as a personal insult. Sorry dude. No insult intended whatsoever.

Regarding the "Horse racing odds vs. Natural odds", I'm not even sure I understand your point. Maybe this is due to my unsophisticated understanding of mathematical statistics and probability. I don't understand your distinction between natural and horse racing odds.

Your argument appears to be (cutting and pasting a bit):

1. If there is little or no deviation between natural odds and track odds it would be impossible to achieve a profit at the track that would meet any meaningful significance test.

2. There is in fact, as empirically verified by your study, little or no deviation between natural odds and track odds.

3. Therefore, it is impossible to achieve a profit at the track that would meet any meaningful significance test.

I take it that the point you believe you have proven is NOT
"it is impossible to achieve a profit at the track" but rather it is impossible to do so AND AT THE SAME TIME meet ANY meaningful significance test. I guess I'm just dense but I don't see the connection between your premise and conclusion. The impossibility of meeting ANY meaningful significance test seems to me at least to be a conclusion far exceeding any argument you have given, but I already admitted I don't really understand your argument so there it is. If I understand you correctly, are you saying no matter how big my sample size is for a a group of plays that shows a positive roi, it, as a matter of logic, cannot pass ANY meaningful significance test?

I guess the corollary to your point is "For there to be any meaningful significance test for achieving a profit at the track, natural odds must not equal track odds." So perhaps you could explain what such an inequality would look like. And not to be pedantic (well, maybe a little) but do you believe it is possible for there to be a meaningful significance test for achieving a profit at the track IF the natural and track odds are different?

Whew, its exhausting pontificating on things I apparently have no understanding of. Sorry to dissappoint you.

Ron Tiller
HDW

PS: I know you started a thread on this subject a while ago and I only blabber on because you started THIS thread and asked my opinion. Now you know better.

formula_2002
12-13-2004, 10:09 PM
Ron, first let me say why it’s important to express win % in terms of track-tack out.
Using your example #1
1. All 4-5 horses since Jan 1, 1995 | Dirt Sprint | 7 horse fields (to normalize the win percentages) | Fastest 1/4 mile time in the last 10 races, ranked (1 = fastest horse to the 1/4, based on last 10 races, 2 = 2nd fastest to the 1/4, etc.

Results:
1st.....49% wins.....3472 starts
2nd....48% wins.....2287 starts
3rd.....47% wins.....1602 starts
4th.....44% wins.....1303 starts
5th.....42% wins.....922 starts
6th.....39% wins.....634 starts
7th.....41% wins.....421 starts

Actual odds used were .7-1 and .9-1

At a 15% take-out track the .7-1 would be expected to win (1/ (1.7))/ (1.176) or 50%
At a 15% take-out track the .9-1 would be expected to win 45% of the time.
At an 18% take-out track the .7-1 would be expected to win (1/ (1.7))/ (1.22) or 48%
At an 18% take-out track the .9-1 would be expected to win 43% of the time.
This is only important if one is trying to determine how accurate the public is in marking the odds. This determination can be made by the expression, (actual win%/expected win %), where 1 would be perfect, <1 would indicate the public is over betting and >1 would indicate the public is under betting.

In order to break even, on the .7-1 the top ranked speed, as defined above, would have to win 58.8% of the time.
(1/ (.7+1))=58.8%

In order to break even, on the .9-1 the top ranked speed, as defined above, would have to win 52.6% of the time.
(1/ (.9+1))=52.6%

Compare these min win % to the results in example 1 and we see that no profit is achieved. Not that one was expected.

I think it was Barry’s purpose to indicate that 4/5 horse win at various rates, depending on “early speed”.

It appears that you may have demonstrated that.


More to your specific questions later.

Joe M

formula_2002
12-14-2004, 07:05 AM
Ron, your question;

“If I understand you correctly, are you saying no matter how big my sample size is for a group of plays that shows a positive roi, it, as a matter of logic, cannot pass ANY meaningful significance test?”

Not quite. I have an interactive “significant testing folder” on my web page.

Take the 49% winners in your example #1.

If the 49% winners represent .7-1 horses, you would have to win 60.5% of your bets to obtain 95% confidence that you would make 2.9% profit in 3472 plays.

If the 49% winners represent .9-1 horses, you would have to win 54.4% of your bets to obtain 95% confidence that you would make 3.4% profit in 3472 plays.

It has been my experience that you may want to perform the test on two different samples. That can amount to quite a substantial data base

Those are just examples. You can vary the inputs in the “significance table” to obtain values you may be comfortable with.

The key to using the table is to use a narrow odds range,

formula_2002
12-14-2004, 07:59 AM
Ron, your question

“I guess the corollary to your point is "For there to be any meaningful significance test for achieving a profit at the track, natural odds must not equal track odds." So perhaps you could explain what such an inequality would look like. “

Using your example #1 and assuming the results are for a 4/5 horse at a 15% take-out.

To break even the 4/5 horse must win 55.5% of the time.
We see that the top ranked horse actually won 49% of the time.

If the public is accurate in its odds making, the horse should have won 47% of the time. (1/1.8) / (1/ (.85)) = 47%

Your example would indicate that the public , with-in a fixed odds range, is under betting the top “speed” and over betting the remaining ranked “speed”.

That would be a stunning result. But I could not accept it with out being certain that only 4/5 horses were considered (which they were not, the example use .7 to 1 through and including .9-1), and that the actual win % were compared to the expected win % adjusted for the track take-out.

If the results held up to be “stunning”, then It will be a good argument that track odds do not correspond to natural odds.
In natural odds, a 4/5 probability is constant within an acceptable range.

The inequality is;
It may appear that the track odds of 4/5 are skewed beyond natural odds.
For track odds to be equal to natural odds, there can be no “ladder” effect as demonstrated in your example.
Skewed to make a profit? That’s another matter.

RonTiller
12-14-2004, 07:12 PM
First, I understand why it’s important to express win % in terms of track-take-out. This is pretty standard stuff: odds as the public's money determines them and odds actually offered to the public, with track take subtracted from the equation. No problem there.

Second, it sounds like you are back peddling a bit on the conclusion you initialy posted, namely, "... where there is little or no deviation between natural odds and track odds it would be impossible to achieve a profit at the track that would meet any meaningful significance test." Is it impossible or not? If it is impossible, is it just an empirical impossibility (it would require more samples than the total number of races that have been run since 1900) or is it a logical or mathematical impossibility (it is impossible to color a map with less than four colors where no two contiguous regions are colored the same, or it is impossible to beat a negative expectation game in the long run by varying the bet sizes).

Third, I still don't understand what you mean by "natural odds" (maybe I'm just incorrigible on this issue).

Fourth, the ladder effect, as you call it, is not due to my taking an odds range of .7 to .9 as equivalent to 4-5 odds. To call something "4-5 odds" is itself a range if one's discreet points are 4-5, 3-5, 2/5, 1/5, even, etc. Now it just happens that the odds stored in the Equibase tables are incremented in .05 increments, but those themselves are rounded. If you want to get technical, you can carry it out to however many places your calculator will go, to get the thinest of the thin slices of odds. But all this is unnecessary, as I just reran the queries for Exactly 4-5 odds (.8). This is actually the .775 to .825 range, because the odds are stored in rounded .05 increments.. Here are the results for one factor:

All exact 4-5 horses (as stored in the rounded odds field) since Jan 1, 1995 | Dirt Sprint | All horses | Fastest 1/2 mile time in the last 10 races, ranked (1 = fastest horse to the 1/2, based on last 10 races, 2 = 2nd fastest to the 1/2, etc.

Results:
1st.....51% wins.....5364 starts
2nd....46% wins.....3005 starts
3rd.....45% wins.....1853 starts
4th.....43% wins.....1306 starts
5th.....42% wins.....809 starts
6th.....40% wins.....455 starts
7th.....39% wins.....289 starts

Now this ladder effect goes with every odds range slice or every exact odds I test, high or low odds. Moreover, it is not just with these early speed factors. I'd guess there are dozens or hundreds of these ladder effect factors tested for each odds range or each discreet .05 rounded odds number. I've randomly checked a few and they are unmistakable. No, I'm way to busy to spend the day testing them so I can post them here. Alas, I've spent more time than I should have on this already.

I guess I have a few final observations:

First, I don't understand at all why this result is "stunning" as it seems to be exactly what I would guess it to be.

Second, you state "For track odds to be equal to natural odds, there can be no “ladder” effect as demonstrated in your example." As I mentioned, I don't understand what you mean by natural odds, so I don't understand what the ladder effect for a particular factor has to do with natural odds.

Third, if the ladder effect is in fact in place, and if, as you state "...then It will be a good argument that track odds do not correspond to natural odds" then does that mean that it is in fact possible to achieve a profit at the track that would meet a meaningful significance test.

Oh boy, I'm getting burned out on this. I guess I'll have to remain in my state of confusion on this issue (a state I am well acquainted with). However, my original intent was to answer your question on early speed and 4-5 horses. I hope I helped.

Ron Tiller
HDW

formula_2002
12-15-2004, 03:58 AM
Ron,
What I mean by natural odds is;

Flipping a coin has a 50/50 chance of heads or tails. Odds are 1-1
Rolling a 12 craps has a 1/36 chance. Odds 35-1
Natural odds games are those games where, in the long run, the number of expected wins equals the actual number of wins, producing a ratio of 1 (+- a very small fraction).

In my relatively small data base of 200,000 horses, that include over 80 factors (Bris/ All-Ways data file factors) I have sufficient data to indicate a top ranked factor wins more often then a second ranked factor, which wins more often then a third ranked factor , etc.
The AVERAGE odds of the factor correlate to ranking. The higher the ranking, the lower the odds.

You have demonstrated that in a fixed odds range, the win probability decreases as the ranking increases. To me that is stunning.

Regarding your comment
“Second, it sounds like you are back peddling a bit on the conclusion you initially posted, namely, "... where there is little or no deviation between natural odds and track odds it would be impossible to achieve a profit at the track that would meet any meaningful significance test." Is it impossible or not?”

No back peddling here. The statement still holds.

You appear to have demonstrated that within a specified odds range, there is a deviation between the natural odds and the track odds. For there to be no deviation, all ranked speed horses in the 4/5 range should win at the same rate.

Your top ranked won 51% for a dollar loss of only 8%
2nd rank, 46% for a dollar loss of 17%
3rd rank, 45% for a dollar loss of 19%
4th rank, 43% for a dollar loss of 23%
5th rank, 42% for a dollar loss of 24%
6th rank, 40% for a dollar loss of 28%
7th rank, 39% for a dollar loss of 30%

(The overall win% for the above is 47%. Just what one would expect for a 4/5 horse in a 15%+- take-out track.)

However, the deviation is not sufficient enough to over come the track take-out, but still, it is a deviation. The number of actual wins/ number of expected wins are much different then 1. For 1, the actual dollar loss should = the track take-out.


Thanks for your enlightenment. It’s the single most import thing I have seen presented here (exclusive of my own work of course)
;)

RonTiller
12-15-2004, 07:14 PM
I think I understand where our differences lie and where we may have a fundamental disagreement. You describe natural odds in terms of coin flips and dice rolls, the study of which spearheaded modern probability theory. Well and good.

The 50/50 probability of either H or T with the coin flip is not a 50/50 probability "in the fabric of the universe" (I am not comfortable with that phrase but I'll explain anyway). When I say that, I am contrasting it with the probability that a particular hydrogen atom will emit an alpha particle at time t or the probability that an electron will be at a certain location at time t . As I understand quantum mechanics, these probabilities are brute facts, so to speak. Given that there are no hidden variables anywhere in the system, the probabalistic nature of the alpha decay or the electron's location at time t is built into the fabric of the universe; no amount of data will allow a "better" prediction and the data we have is exhaustive and complete. it's just a bunch of probabilities with no further explanations possible.

Contrast that with:

1. The probability of H on this coin flip is .5.
2. The probability of rain tomorrow is .3.
3. The probability of there being other intelligent life in the universe is .004.
4. The probability of this horse winning race 3 at Calder is .3.
5. The probability of actress X winning the Oscar is .7.
6. The probability of the Lakers beating the Heat is .65.

I know, there are all sorts of debates concerning what 'probability' is and how many types of 'probability' there are and whether to be a Bayesian or not, ad nauseum.
I'm guessing that most people don't literally believe that the coin flip is a 50/50 shot in the same sense that an alpha particle decay is a 50/50 shot at time t. Rather, the coin flip is a paradigmatic case of extremely sensitive dependence on intial conditions (chaotic). We use coins and die for games of chance because the smallest minute variations in dozens of factors involved in their throwing produces in practice unpredictable variations in outcome.
In practice, the outcome of the flip or the throw is so sensitive to initial conditions that no matter how much information we gather about the relative humidity, wind conditions, thumb stregth, etc., we cannot improve our prediction. Yet I believe the whole sequence unfolds in a deterministic manner. The chaotic elements ensure a 50/50 distribution of outcomes. They also ensure, with our current level of technology, that nobody will, in the long run, do better than predict with 50% accuracy the outcome of the flips.

I don't agree that any horse, anywhere, at any time has natural odds of winning analagous to the coin flip or the dice roll natural odds. Neither do I believe that there are natural odds of it raining in Lexington Ky 5 days from now, analagous to the coin flip or the dice roll natural odds.

Setting 1-1 odds on a horse does not mean that the horse has a 50% chance of winning, just like the alpha particle has a 50% chance of being ejected from the nucleus. Nor can I make sense of the oft heard explanation "if the race were to be run 100 times, this horse would win 50 times." (what the heck does THAT mean?!) The only interpretation I find plausible is that the probabilities are relative to a set of information or data. Relative to the database model, historical data analysis or Ouija board proclamations, this horse has a 50% chance of winning. If I look at 1000 horses I had at 50% and 20 win, I did a bad job of assigning probabilities. Am I looking for THE natural probabilty of this horse winning, a natural probability analagous to the chaotic coin flipping system defined by the equiprobability of outcomes? I don't think so.

One horse may have an 80% chance of winning based on my complex analysis of a particular spot play with repeatable sample sets. Yet, relative to another data set, or angle, this horse shows 50% chance. The horse runs and wins. Which odds were correct? Well...I'm sure you have an answer and it may be a good one but my point is that this is not a case of ever closing in on the real, natural odds of the horse winning. It's simply making better predictions or making predictions that are, in practice, more useful. The weather forecast is not ever improving by closing in on the REAL NATURAL probability of rain on thursday. It's going to rain or not regardless. But with the super dual doppler radar, we have better predictions (at least that's what our local weatherman tells us nightly).

Having gotten off my soapbox, the ground is starting to feel a bit slippery now. You are obviously more knowledgable than I am on this stuff so this may all be a bunch of sound and fury, signifying nothing. Rats...

Ron Tiller
HDW

JustMissed
12-15-2004, 08:38 PM
I thought maybe Joe was comparing apples & oranges when he was trying to compare common events under controlled conditions like coin tosses or dice throws with uncommon uncontrolled events like horse races.


Glad someone much smarter than me exposed the error of the "if a horse ran a 100 races" example and the "natural odds" relevancy to horse racing.

JM

RXB
12-15-2004, 09:32 PM
The concept of basing horse race betting on fixed, single-fraction probability lines becomes obviously ludicrous to anyone who thinks about it. There is this thing called 'margin for error' and it had better be included in my probability estimates (and, therefore, minimum required odds) for outcomes in a given race. Add in the ever-increasing last-second fluctuations of the odds, and the idea of betting overlays off of a fixed 100% probability line becomes dubious.

There is so much nonsense tossed around even by so-called 'betting experts.' Stuff like: "find out your overall win %, and then never bet horses whose odds fall below the break-even point based on your overall win rate." Good grief.

formula_2002
12-16-2004, 01:38 PM
Ron, we agree on;
“The chaotic elements ensure a 50/50 distribution of outcomes. They also ensure, with our current level of technology, that nobody will, in the long run, do better than predict with 50% accuracy the outcome of the flips.”

We don’t agree on;
Setting 1-1 odds on a horse does not mean that the horse has a 50% chance of winning, just like the alpha particle has a 50% chance of being ejected from the nucleus.

Just look at your own data for 4/5 horses. ;


(The overall win% for the above is 47%. Just what one would expect for a 4/5 horse in a 15%+- take-out track.)

Although there is a win % difference between the top rank to the bottom ranked “speed” horse, the 4/5 group behave in the expected manner.
When , expected winners/ actual winners = 1(+- and very small amount), the odds are natural.

I find it interesting that there are variances within the group.

Perhaps not all electrons within an atom are equal, but the atom behaves in a singular manner. But let us not go there.

formula_2002
12-16-2004, 01:46 PM
Originally posted by JustMissed
Glad someone much smarter than me exposed the error of the "if a horse ran a 100 races" example and the "natural odds" relevancy to horse racing.

JM


Jm, test it for yourself but try a larger sample then 100 (but don't forget to adjust for track-take and entries. It means all the difference in the world)
It's not a test of the SAME horse, rather it is a test of all 1-1 shots.

If the actual winners / expeceted winners =1 (+- a small amount)
the odds are natural.

JustMissed
12-16-2004, 04:24 PM
I don't get it. What does track take have to do with anything?

If you pick the winner and it pays $6, you make $4.

If you pick the winner and it pays $12, you make $10.

You know it appears that when you do your test you get to thinking that the same horseplayer is going to play each and every one of those 20,000 races your are testing. Let me tell you this. A horseplayer, or at least a good horseplayer,does not play every race presented to him. He only plays those races that he thinks he can win and does not play the races he does not think he can win.

Your problem may be that in your test you are including all races as if someone would bet every race. That is not reality.

In your example of a 1-1 shot, if in one race that shot is an E8 among all P & S runners he is not the same horse as a 1-1 shot that is a late runner in a normal or slow-slow race. You can't compare apples to oranges.

JM

formula_2002
12-16-2004, 05:06 PM
Jm said
"I don't get it. What does track take have to do with anything?"

Now I'll really drive to nuts.;)

Would you belive that a 1-1 shot at a 15% take-out track has a better chance of winning then if then if it raced at a 18% take out track?

((1+odds+1)) / 1/(1-.15) = 44% chance of winning.

((1+odds+1)) / 1/(1-.18) = 41% chance of winning.

In a test of my data I got;
406 winners in 895 races for a 45.4 win % at a 15% track
541 winners in 1271 races for a 42.6% wins at a 18% track

RonTiller
12-16-2004, 05:24 PM
formula_2002,

It's clear that we have a very fundamental disagreement and it is equally clear that neither of us is going to convince the other of the error of his ways. I suppose we could go back and forth for weeks but I've got too much to do here and I'm weary of this topic; frankly, these types of analyses just don't get my adrenaline going the way they apparently get yours going. And our "stunned" zones are very far apart.

So, I'll let your last post be the last word on this exchange betwen us, since it is your thread to begin with.

Elvis has left the thread...

Ron Tiller
HDW

PS: You should update yourself to formula_2005, or, to ensure that you are always current, formula_currentyear.

JustMissed
12-16-2004, 05:44 PM
Joe, Maybe you have been drinking too much eggnog or maybe it is that "new math" I have heard about.

As far as I know, the bettors make their bets and the money goes into a pool. The track takes out a certain percentage to cover taxes, overhead, etc. and the balance of the pool is paid out to the holders of winning tickets.

As far as I know the track take has no effect upon the purse or how much the owner, trainer, jockey, jockey agent or valet make. Horses can't read so they don't care.

How the heck could 'take' have anything to do with the probability of a horse winning a race?

JM

GR1@HTR
12-16-2004, 05:48 PM
Track Take Out. I used to think it was best to chase tracks w/ the lowest take out...Still kinda do but Ironically my 2 best ROI tracks are the ones with the highest take out (CRC and PHA)..Go figure??!

formula_2002
12-16-2004, 05:54 PM
Fairwell my friend.It has been a very pleasant interchange of concepts, at least for me.

But I will leave you with the following;

One can show a long term profit at pari-mutual horse racing, if in any incremental odds range:

actual winners / expected winers >1+ track-take-out.:)

If actual winners / expected winers >1 and the exceeds the standard deviation of natural odds then odds are not natural.

Joe M

ps

"PS: You should update yourself to formula_2005, or, to ensure that you are always current, formula_currentyear"
Correct Laws of Physics never change!!

Buckeye
12-16-2004, 05:58 PM
"you got your technology, but you lost"

Wolfen.

formula_2002
12-16-2004, 05:59 PM
Originally posted by JustMissed

How the heck could 'take' have anything to do with the probability of a horse winning a race?

JM

I thought I already demonstrated that.
Perhaps one of the other data gathers on the board would post some results for us.

Buckeye
12-16-2004, 06:05 PM
"Perhaps one of the other data gathers on the board would post some results for us."

I could but I ain't gonna.

Besides, it doesn't matter so much what happened as what's going to happen, can you test that?

I can't.

formula_2002
12-16-2004, 06:49 PM
Originally posted by Buckeye
"Perhaps one of the other data gathers on the board would post some results for us."

I could but I ain't gonna.

Besides, it doesn't matter so much what happened as what's going to happen, can you test that?

I can't.

Oh that is so true. I wait each morning to see if the sun is going to rise again:cool:

Buckeye
12-16-2004, 07:13 PM
If you're a gambler and you get the right odds you'll take that bet.

Total darkness.

It's called "the future"

nomadpat
12-16-2004, 10:07 PM
GR1,

Do you think it has more to do with the skill of the competition rather than amount of takeout?

GR1@HTR
12-16-2004, 10:20 PM
No Mad,

Not sure, I know I do best at tracks where the horses have been running for a while. ie. I do very poor at meets where they first open. Since CRC and Philly run almost for the entire year, that seems to folllow the trend...However, I do think we should focus on tracks that we are best at or have an advantage at over the public.

formula_2002
12-17-2004, 04:35 AM
Part of the reason horses appear to win at different % rates within an odds range can be explained by track take-out and breakage.

in rounded odds range of 4/5 the win % for .75-1 through .999-1 (dime breakage) ,the win % could range from a high of 53.4% at at 14% take-out to a low of 44.8% at a 19% take-out track.

As I have previously stated, the win % rate analysis at various odd ranges must be demonstrated by using the track take-out (and breakage)

hurrikane
12-17-2004, 10:58 AM
Joe,

you just never cease to amaze me. You have a logic I have never experienced before.

simply amazing.

Equineer
12-17-2004, 12:21 PM
Originally posted by Formula_2002,
Jm said
"I don't get it. What does track take have to do with anything?"

Now I'll really drive to nuts.

Would you belive that a 1-1 shot at a 15% take-out track has a better chance of winning then if then if it raced at a 18% take out track?

((1+odds+1)) / 1/(1-.15) = 44% chance of winning.

((1+odds+1)) / 1/(1-.18) = 41% chance of winning.

In a test of my data I got;
406 winners in 895 races for a 45.4 win % at a 15% track
541 winners in 1271 races for a 42.6% wins at a 18% trackEven if the trip is short, you will drive us nuts unless you read your own posts and correct them once in awhile.

Your two formulas are straight from bedlam... pure gibberish... solve them yourself for 1-to-1 pari-mutuel odds. :)

However, your original assertion is worth a bit of consideration for the following reason:

When breakage methods are identical, for absolutely equal pari-mutuel odds, a higher percentage of the win pool wagering will be committed at the track with the lower takeout.

Thus, although two horses may be 1-1, the one at the 15% track is actually favored more than the one at the 18% track.

In the long run then, we might expect the 1-1 shot at the 15% track to win slightly more often than the 1-1 shot at the 18% track.

Using dime breakage:

At the 15% track, 41.45% of the public's money is needed to produce 1-1 odds.

At the 18% track, 40.00% of the public's money is needed to produce 1-1 odds.

So to the extent that the public knows anything, when many race are consider, 1-1 shots at the 15% track are slightly favored to win more often.

formula_2002
12-17-2004, 12:50 PM
E, OF COURES YOU ARE CORRECT

(1+odds) / (1/(1-.15)) = 42.5% chance of winning.

(1+odds) / (1/(1-.18)) = 41% chance of winning.

Now calculate the win % based on 15% take and dime breakage and 4/5 odds

[code]
I' sending you an excel file with some data.
Cant post it here because I can't get the columns to align

formula_2002
12-17-2004, 12:59 PM
E, I see I cant not send you an e-mail.
If you are interested in the back up to ;

"Part of the reason horses appear to win at different % rates within an odds range can be explained by track take-out and breakage.

in rounded odds range of 4/5 the win % for .75-1 through .999-1 (dime breakage) ,the win % could range from a high of 53.4% at at 14% take-out to a low of 44.8% at a 19% take-out track.

As I have previously stated, the win % rate analysis at various odd ranges must be demonstrated by using the track take-out (and breakage)"

send me you e-mail address and i'll send you the excel file.

formula_2002
12-17-2004, 02:43 PM
E, OF COURSE YOU ARE CORRECT

i'LL GET IT RIGHT YET

(1/(1+odds)) / (1/(1-.15)) = 42.5% chance of winning.

1/(1+odds)) / (1/(1-.18)) = 41% chance of winning.

Now calculate the win % based on 15% take and dime breakage and 4/5 odds


__________________

Equineer
12-17-2004, 03:27 PM
Originally posted by formula_2002
E, OF COURSE YOU ARE CORRECT

i'LL GET IT RIGHT YET

(1/(1+odds)) / (1/(1-.15)) = 42.5% chance of winning.

1/(1+odds)) / (1/(1-.18)) = 41% chance of winning.

Now calculate the win % based on 15% take and dime breakage and 4/5 odds


__________________ (1-.15)/(1+odds) = .425

(1-.18)/(1+odds) = .410

I think this algebraically is what you might prefer.

My previously posted results were a little different because I rearranged the general formula that incorporates estimating dime breakage.

formula_2002
12-17-2004, 03:43 PM
Originally posted by Equineer
(1-.15)/(1+odds) = .425

(1-.18)/(1+odds) = .410

I think this algebraically is what you might prefer.

My previously posted results were a little different because I rearranged the general formula that incorporates estimating dime breakage.

plus

"However, your original assertion is worth a bit of consideration for the following reason:...."


THIS IS A MAJOR BREAK THROUGH FOR ME.
some one appears to agree with me!!!

Equineer
12-17-2004, 03:58 PM
But to make clearer what I meant, my previous results (.4145 and .4000) represented an estimate of the average win pool percentages (incorporating possible breakage) for 1-1 odds.

Now, when pari-mutuel odds are already known (after breakage) the real possible ranges for win pool percentages for 1-1 odds are roughly .4050 to .4250 and .3905 to .4100.

So your bear in mind that your formula results always express the range maximums (.4250 and .4100).