Boxing the Trifecta

[Ray2000]

This might be the Mother of all Red-Boarding but I found the results to be surprising.
Unlike the Exacta where the Straight wager always out-performs the box or wheels on top 4 picks,
......The Straight Trifecta does not, at least for my picks.

The table shows the results of betting the Robot's top 4 picks in various combinations.
Horse A is Top choice, then B-C-D... Races where any choice was scratched are not used.

The data covers 127,322 Races, .....All North American tracks from Jan 2009
The 2$ cost is $254,644 and is always the same
(i.e you buy 1/12th of a 24$ box4 tkt or a single straight $2 tkt)



Ticket Return ROI Hits Strike Rate
AB/AB/CD $278,665 1.094 8,777 6.9%
AB/AB/C $277,019 1.088 4,795 3.8%
A/B/CD $275,117 1.080 4,969 3.9%
AB/ABC/CD $270,636 1.063 11,489 9.0%
A/BC/BCD $269,553 1.059 8,861 7.0%
A/BC/BC $268,280 1.054 4,966 3.9%
ABC_3Box $267,452 1.050 11,447 9.0%
A/B/C $267,022 1.049 2,721 2.1%
A/BCD/BCD $261,405 1.027 11,901 9.3%
BCD/A/BCD $258,298 1.014 8,148 6.4%
ABCD_4Box $252,244 0.991 31,273 24.6%
AB/ABC/BCD $236,683 0.929 13,735 10.8%


Given the higher return and 3x the Strike Rate, AB/AB/CD is the much better bet.

I checked the ranking by eliminating all hits > $1000 and reached the same conclusion.
You need an enormous number of races when dealing with Tris :)



.

[precisionk]

This is an interesting read. Unless I am a FIRM believer in a horse, I will most likely box my trifectas. This is especially true when I am at Running Aces live. I will torpedo odds of my top horse like no other and look for value in my trifecta's on 2nd and 3rd choice.

[DRIVEWAY]

A very interesting analysis. Surprisingly my trifecta standard is not referenced.

Just curious how AB/ABC/ABCD worked out.

Thanks again for your analysis.

[Ray2000]

Mr T, DRIVEWAY
Thx for the replies.
When I did the AB/ABC/ABCD I noticed I had made a mistake on one of the other combos.
AB/ABC/BCD is 7 tickets not 8, that guy got shorted on the return :)

Here is the update.



Ticket Return ROI Hits Strike Rate
AB/AB/CD $278,665 1.094 8,777 6.9%
AB/AB/C $277,019 1.088 4,795 3.8%
A/B/CD $275,117 1.080 4,969 3.9%
AB/ABC/CD $270,636 1.063 11,489 9.0%
AB/ABC/BCD $270,495 1.062 13,735 10.8%.....corrected
A/BC/BCD $269,553 1.059 8,861 7.0%
AB/ABC/ABCD $269,280 1.057 15,357 12.1%.....new
A/BC/BC $268,280 1.054 4,966 3.9%
ABC_3Box $267,452 1.050 11,447 9.0%
A/B/C $267,022 1.049 2,721 2.1%
A/BCD/BCD $261,405 1.027 11,901 9.3%
BCD/A/BCD $258,298 1.014 8,148 6.4%
ABCD_4Box $252,244 0.991 31,273 24.6%

[CHeCK EyE]

As always Ray. Personally I like the A/BC/BCD and I was pleased to see it closer to the middle than the bottom of the list!

[imofe]

Ray

Any chance you could post the exacta chart?

[mrroyboy]

Ray do you have any stats on horses bet down below their morning line odds. Just curious if they pay more or less?

[Ray2000]

imofe

Using the same races and same picks for Exacta betting gives the following table of results.


Ticket Return ROI Hits Strike rate
A/B $259,778 1.020 10,287 8.1%
A/BC $258,522 1.015 18,182 14.3%
AB/AB $258,390 1.015 18,244 14.3%
AB/ABC $254,765 1.000 31,374 24.6%
A/BCD $252,437 0.991 24,268 19.1%
Box3 $246,256 0.967 41,026 32.2%
BCD/A $231,230 0.908 16,735 13.1%
Box4 $230,298 0.904 62,920 49.4%



When I starting looking at these races, I went with Box 3 and Box4 then limited the queries to Pick1 (A) finishing no worse than 2nd. If you have some other combo Exacta ticket you would like to see, involving top 4 picks, let me know.

mrroyboy

Unfortunately no, that would take some detailed mining of the charts, not just the results and I haven't done that, might not even be possible with the charts I have.

I was thinking of using randomly generated pick numbers as a comparison but even that gets complicated trying to stay away from A/A/A tickets and other scratch/fieldsize problems.



and oh yea, for "Mr. Pic3" out there (mel, I know you're reading this :)), the number of bet opportunities and scratch handling makes this a BEAR to try and compare..




.

[RaceTrackDaddy]

That is the way I play my superfecta wagers...AB/AB/CD/CD. I never thought of reducing it for Trifecta wagering..be the same cost. I would have two horses on top of the ticket instead of my single pick on top: A/B,C,D/B.C,D,E.

Think I will try going with the AB/AB/CD method,,,if it works out for me in the Supers, it should have better results in the tris.

thanks for posting this thread Ray..
rtd

[imofe]

Are these for all field sizes or do you limit it to full fields?

[Ray2000]

imofe

The computer rejects races with more than 10 or less than 7 program listed starters. I tossed out any race with a scratch on any of the top 4 picks but scratches on the others in the race could make for some 6 horse fields.

BTW
There is one metric which is fascinating. For each race, the computer sums the Morning Line Rank of its Pick1 & Pick2.
(If Pic1 is 4th in ML and Pick2 is 2nd that = MLR12=6)

If you sort all 127000+ races on those added Ranks and bet $2 on each race in sequence you get a profit balance shown in the graph. There is clearly a break point at bet no. 65000. That is MLR12 of 5. All races to the right of that point have MLR12 > 5.

As much as I complain about the incompetence of the Morning lines it has the biggest effect. I think that's because a lot of trifecta players are newbies and tend to rely on ML picks. If the computer agrees with the morning line, it's an underlay.

[imofe]

This is very interesting Ray. Would you then say that it is better to play at tracks where the ML is not that accurate?

For example, if your top four ranks are usually pretty close to the ML at Track A, the public can hurt your results.

If at Track B your top four ranks are somewhat different than the ML, your chances at profit are increased because the public will be on other horses.
This is assuming that you pick a close % between Tracks A and B.

[Ray2000]

I'll have to think about that one :confused: , in general it makes sense but my "bad" tracks are Chester, Freehold, Pokie, Woodbine and Windsor. Chester has the worst ML (IMO). And the computer's best tracks are Georgia, Balmoral, Dover, Harrington, Hazel and Tioga. And the ML guy at Tioga is pretty sharp.

Hard to figure...


.

[marksinger]

I like AC/ AD/ BC/ BD .
Anyone ever tried these?

[Ray2000]

mark

Do you mean Exactas AB/CD ?

[CHeCK EyE]

But never played! Two monster favorites based on odds and play them first and third with everyone else for second, hoping someone can squeeze in between, haha. I've said it before but never bet it. Not even sure if the prices would be worth it with big favorites like that.

[grant miller]

does this take into account that certian numbers are plaed more than others? 1-2-3 tri box, 1-2-3-4 dime super boxes seems alot of people call them out at my track (buff & batvia) great info ray:ThmbUp: :ThmbUp: :ThmbUp:

[Ray2000]

Grant

I haven't done any breakdown on average payoffs but my guess would be the total Return would contain more of the frequent, lower priced, '1-2-3's and fewer hi-priced '10-9-8's.

As for the picks used in this data set, the computer uses Post Bias as one of its rating factors so it also has a built-in preference for the '1-4-5's .


.

[SchagFactorToWin]

I use A/BC/BCDEFG (10 tix) because I have found that it results in a larger avg. payout. I tested all of the combos Ray posted, using my spot picks only (sample size=2000, Yonkers and Monti). My original combos had a better return but a few of the other combos were close, more so at Monti.

[Stillriledup]

Very interesting Ray, i have a question if you dont mind.

is the A, B, C and D the top 4 betting choices in order at off time in the win pool, or, are ABCD your top 4 picks regardless of actual price, or, is it something else?

Thanks!

[Ray2000]

Stillriledup

Those picks are my computer program selections, AKA RaysRobot, for the top 4 over the past 4 years for all tracks where internet wagering is available.

The "Robot" doesn't do races for Maidens, low class (<68), coupled entries, distances other than 1 or 1-1/16 mile, starters less than 7 or more than 10, or races with too many break lines.

I could break it down by track but when dealing with trifecta results, you need a large sample of races to be meaningful. As a rule of thumb, I use a minimum of 1000 for win analysis, 4000 for Exacta and 10,000 for trifectas.


.

[Ray2000]

Bumping this thread on Trifecta combination play to add the corresponding ROIs for playing the favorites. (Stillriledup's reply got me interested to check it out :))

In this case Pick1 (the "A" horse) is the crowd favorite, B is the crowd's 2nd choice, C is the 3rd, D is 4th favorite.

The data is several years old but the trend should hold true.
40,030 Races, all bet $2 as explained above.


Hits Return ROI Strike Type
2555 $65,518 -18% 6% A/B/CD
4448 $65,149 -19% 11% A/BC/BCD
5342 $64,959 -19% 13% AB/ABC/CD
4151 $64,226 -20% 10% AB/AB/CD
6405 $64,041 -20% 16% AB/ABC/BCD
7002 $63,566 -21% 17% AB/ABC/ABCD
5823 $63,487 -21% 15% A/BCD/BCD
2310 $63,396 -21% 6% AB/AB/C
3166 $62,598 -22% 8% BCD/A/BCD
1404 $61,909 -23% 4% A/B/C
12697 $61,394 -23% 32% ABCD_4Box
4980 $60,360 -25% 12% ABC_3Box
2467 $60,219 -25% 6% A/BC/BC

see below*
F_sum t/e # of races with F_sum
6 3.0 4980
7 3.4 3568
8 3.9 5358
9 4.4 5270
10 5.0 4927
11 5.7 4118
12 6.1 3710
13 6.9 2464
14 7.4 1863
15 8.3 1381
16 9.1 941
17 9.8 550
18 10.2 415
19 10.4 228
20 10.6 118
>20 10.7 139



*These numbers are just a curiosity which I found weird because of the perfect increment.

F_sum is the sum of the Odds Ranks of the 3 horses comprising the Trifecta.

6 has to be 1st fave, 2nd fave, 3rd fave in any order
7 has to be 1st fave, 2nd fave, 4th fave in any order
8 must be 1,2,5 ... or ... 1,3,4 any order
9 can be 126, 135, 234
and so forth...


t/e is the trifecta to exacta price ratio


The point being, How much more should the Trifecta pay when compared to the Exacta?

If the top 3 faves come in you should expect to get 3 times the Exacta.
If 3 longshots hit, then 10 times the Exacta is a reasonable price.

As you may have guessed, I think I got screwed on a tri-hit yesterday...:)


.

[traynor]

I don't want to give away the candy store, but you are headed in the right direction. However, I cannot imagine betting more than a quarter of a million dollars over a period of years to make less than 1% profit. That slim a margin is often ephemeral at best.

If you ignore morning line odds, and only consider your rankings, it should be fairly easy to establish ranges of odds values with associated results. Specifically, you may find (as many others have found before you) that the majority of races that comprise the strike rate are not worth betting. Rather than eliminating the high end of the range, look for ways to eliminate the low end--those frequent "successful" races that, when considered as a range, are non-profit.

Studies of the AB/AB/CD tris in which A is less than a specific value, and/or B is less than a specific value can lead to some interesting conclusions. When you categorize results by odds ranges, you are likely to find a large clump of races that eat up what would otherwise be a healthy profit. Eliminate those races, and while your strike rate may drop, your profits may increase dramatically.

[Ray2000]

....look for ways to eliminate the low end--those frequent "successful" races that, when considered as a range, are non-profit.



Solid logic as usual Traynor, I like reading your posts here and in the T-Bred Forums.

Eliminating the over used horses in the Exotic combinations, is what it's all about.

And now-a-days Trifecta wagering lends itself to getting excellent rebates from ADWs
which mitigates the absurd take-out vigorish and pits you against less discouraging foes.

GL

.

[marksinger]

mark

Do you mean Exactas AB/CD ?
Yes Ray, that's what I meant. :)