MLO vs Final Odds?

[InFront]

I was wondering if anyone has done these studies through a large database.

1. What percentage of all races are won by horses going off lower than their morning line odds?

2. What percentage of horses go off lower than their MLO vs those higher than their MLO?

3. What is the ROI if you bet all those horses to win that went off lower than their MLO vs what the ROI for those that went off higher than their MLO?

Any help would be appreciated. Thanks.

[Dave Schwartz]

In this study, we focus on the difference between odds and morning line. Both Odds and ML are expressed as a "booking pct," with the former divided by the latter. As an example, if a horse was:

ML 4/1 = 25
Odds 2/1= 33

33 / 25 = 1.32

Thus, horses in the first group were bet up by over 10%, while the middle group was within 10% either way, and the last group was bet down by at least 10%.


The sample is from all races, all tracks in 2006. (No very minor tracks - what we call "circuit "0" tracks.)


2006.
431-OddsML1
------------------
WIN BETS
Field1 Field2 Starts Pays Pct $Net IV PIV
--------------------------------------------------------
0.89 220,143 12,172 5.5 $1.41 0.47 0.90
1.10 58,071 8,137 14.0 $1.60 1.14 0.97
above 127,454 28,827 22.6 $1.64 1.80 0.98

Total 405,668 49,136 12.1 $1.51 1.00 0.96



I have done much work with this stat and can tell you that it becomes much more meaningful when each race is normalized for current field size (i.e. after scratches) and mistakes in the line (i.e. the line does not add up correctly).


Regards,
Dave Schwartz

[Cangamble]

Shouldn't 4-1 be .20?

[cj]

Shouldn't 4-1 be .20?

Yes, but I don't really think it takes anything away from the meaning of the post, do you?

[InFront]

Thanks. So if I am reading this correctly almost twice as many horses go off well above their MLO than well below their MLO? Does that $1.41 mean those horses have a ROI of -30% vs the well bet group lose only -13%? Also can you tell what % of ALL races are won by horses with lower final odds vs their MLO? Someone thought that 75% of all races were won by a horse with lower final odds vs their MLO but then read it's actually only about 50% which may be possible since even those these overlay horses have a small win% as a group there are many more of them per race?

One more thing. Is there a way to study how accurate the MLO is vs final odds. Once again I read it is pretty accurate with horses under 10/1 but those with 10/1 or higher it is not accurate at all. Thanks again.

[Cangamble]

Yes, but I don't really think it takes anything away from the meaning of the post, do you?
Absolutely not. I was just being anal.

[bigchump]

I have a database of over 661,800 races run in the US and CAN.
I don't want to break your heart, but you'd best forget about the ML if you want to survive at the track.

The only good thing about the ML is getting big ML odds on the horse you like.

If you still want to know the stats on the ML, PM me. But, you won't like it...

[InFront]

When building different spot play angles and running queries sometimes I use the MLO but not to see what is picked but to try and get a better price on my winners overall. I have found by sometimes by demanding a higher MLO it helps the bottom line. But I was just curious about the MLO stats I asked and posted on which horses win.

[Dave Schwartz]

Cangamble,

LOL - Yes. My mistake. (Actually, I was just testing you.)


Dave

[Cangamble]

The ML favorite is just one handicappers opinion of who the public will bet the most. A lot of people throw these horses in exactors, picks, tris etc.
I really cringe when I have a program fav I need to use, I try to avoid them whenever possible because they are such natural underlays usually.

[ryesteve]

Thanks. So if I am reading this correctly almost twice as many horses go off well above their MLO than well below their MLO?
That's because ML makers generally don't like to make a horse over 20-1... so every longshot will end up falling into the "underbet" category.

[InFront]

The MLO favorite wins 30% of all races loses more than 20 cents on the dollar. But it's pretty amazing based on someone's opinion days before the races run that it wins almost as much as the post time favorite which is based on everyone's opinion. But yes cause of it's overbet so much they are generally bad bets. Now we just need a MLO formula that wins as much but the public way underbets it. Or find some kind of overall flaw in the MLO favorite where they are still overbet but lose more often.

[hdcper]

Infront,

Just thought I would post a little more info on your question. The Morning Line Odds Ratio equals Post Time Odds / Morning Line Odds, thus <= 1.00 are all horses going off at less than their morning line odds. Further, of the 24,897 winners in the database, 16,598 of them (66.6%) went off less than their morning line. These 16,598 winners were out of 76,103 horses that fit this category (38% of the 200,286 horses in the database). The other questions you asked can certainly be calculated using the output listed below.

Hope it helps,

Hdcper



Data Summary Win Place Show
Mutuel Totals 307052.00 301610.50 300087.80
Bet -400572.00-400572.00-400572.00
Gain -93520.00 -98961.50-100484.20

Wins 24897 49594 73118
Plays 200286 200286 200286
PCT .1243 .2476 .3651

ROI 0.7665 0.7529 0.7491
Avg Mut 12.33 6.08 4.10


By: Morning Line Odds Ratio

>=Min <Max Gain Bet Roi Wins Plays Pct Impact
-999.00 0.00 0.00 0.00 0.0000 0 0 .0000 0.0000
0.00 0.50 -4942.80 33376.00 0.8519 5426 16688 .3251 2.6156
0.50 1.00 -22290.60 118830.00 0.8124 11172 59415 .1880 1.5126
1.00 1.50 -21645.50 97158.00 0.7772 5080 48579 .1046 0.8412
1.50 2.00 -12467.00 58636.00 0.7874 1876 29318 .0640 0.5148
2.00 2.50 -10028.20 35360.00 0.7164 735 17680 .0416 0.3344
2.50 3.00 -7810.10 21650.00 0.6393 291 10825 .0269 0.2163
3.00 3.50 -5059.10 13640.00 0.6291 154 6820 .0226 0.1817
3.50 4.00 -2232.90 8326.00 0.7318 91 4163 .0219 0.1758
4.00 4.50 -2179.60 5142.00 0.5761 36 2571 .0140 0.1126
4.50 5.00 -1808.00 3186.00 0.4325 16 1593 .0100 0.0808
5.00 5.50 -1248.20 2024.00 0.3833 8 1012 .0079 0.0636
5.50 6.00 -710.80 1172.00 0.3935 4 586 .0068 0.0549
6.00 6.50 -287.20 726.00 0.6044 4 363 .0110 0.0886
6.50 7.00 -88.00 448.00 0.8036 3 224 .0134 0.1077
7.00 7.50 -324.00 324.00 0.0000 0 162 .0000 0.0000
7.50 8.00 -210.00 210.00 0.0000 0 105 .0000 0.0000
8.00 8.50 -116.00 116.00 0.0000 0 58 .0000 0.0000
8.50 9.00 -82.00 82.00 0.0000 0 41 .0000 0.0000
9.00 999999.00 10.00 166.00 1.0602 1 83 .0120 0.0969

[InFront]

Thanks for the detailed chart. With a good size sample as this we can safely now say that 66% of winners go off less than their MLO. I heard 75% and then read 50% so it looks like the real number is right in the middle. Not sure how we can use this info to find value but at least I now know.

[BMeadow]

"Further, of the 24,897 winners in the database, 16,598 of them (66.6%) went off less than their morning line. "

Some players take this to mean that "betdowns" (from the morning line) are a good play. Instead, the major reason that the stat above is true is this:

More than half of all races are won by the favorite or second choice. For various reasons, morning-line makers are conservative (e.g., the typical favorite is 5-2 in the morning line but goes off at 7-5). The overwhelming majority of favorites and second choices go off at below their morning line, which skews the stat above.

In the same way, if a mornng line-maker thinks a horse will go off at 60-1, he'll make him 20-1 or 30-1, thus further skewing these results.

In a survey published in Meadow's Racing Monthly involving more than 48,000 horses, Jim Bayle compared horses who went off at 5-1 through 8-1 (who could just as easily have been above the morning line as below it), dividing them into two groups: those whose odds were above their morning line, and those who were below their morning line. Statistically speaking, the two groups were virtually even in their win percentages and ROI.

All this seems to indicate that comparing a horse's odds compared with his morning line seems to be a waste of time.

[bigmack]

All this seems to indicate that comparing a horse's odds compared with his morning line seems to be a waste of time.
Amen. While some ML makers are quite adept at generating appropriate odds others are so off that ML's become less than meaningful.

[ryesteve]

In a survey published in Meadow's Racing Monthly involving more than 48,000 horses, Jim Bayle compared horses who went off at 5-1 through 8-1 (who could just as easily have been above the morning line as below it), dividing them into two groups: those whose odds were above their morning line, and those who were below their morning line. Statistically speaking, the two groups were virtually even in their win percentages and ROI.

All this seems to indicate that comparing a horse's odds compared with his morning line seems to be a waste of time.
The problem is that this analysis was done backwards. If one is hypothesizing that bet-down horses perform better, placing limits on odds is going to end up eliminating the horses you'd expect to do best. What you want to do is look at a narrow range of ML. If instead you look at horses whose ML was 5-1 thru 8-1, who similarly could've just as easily ended up going off at odds lower or higher, you see a significant difference in ROI and a huge difference in win%; .827 vs. .770 and 17.0% vs. 6.8%, over 185,000 horses. If you require a 20% shift in either direction, the differences are even more pronounced: .838 vs. .763 and 19.2% vs. 6.0% over 140,000 horses.

This suggests to me that there is some value in making an odds/ML comparison.

[RonTiller]

I don't get any flattering results when looking at MLO versus actual odds.

Below is all the horses with MLO = 5 versus the actual oods the horses went off at. This is using 345,684 horses with MLO = 5.

Odds W% ROI
All.....13.0%.....0.80
0 - 1.....45.4%.....0.83
1 - 2.....31.2%.....0.82
2 - 3.....22.8%.....0.81
3 - 4.....18.0%.....0.81
4 - 5.....14.6%.....0.81
5 - 6.....12.5%.....0.81
6 - 7.....10.7%.....0.81
7 - 8.....09.6%.....0.81
8 - 9.....08.5%.....0.81
9 -10.....07.3%.....0.77
10-11.....06.9%.....0.82
11-12.....06.6%.....0.83
12-13.....06.3%.....0.84
13-14.....05.4%.....0.77
14-15.....04.7%.....0.73
>15.....3.7%.....0.73

I get similar results looking at other MLO values so it seems pointless to print them. Maybe its just me but I don't see much here to hang one's handicapping hat on.

Ron Tiller
HDW

[Maji]

Ron,

How about the following twist?

What kind of ROI do you get when the current ML Odds is less than the actual odds the horse went off in its last race? If there are a number of choices in the race, then you choose the one with the lowest previous odd.

This assumes that you are dealing with horses with a minimum of one race.

Based on my limited observations, there were some very good prices that can be obtained if only this criteria is used for selection.

Thanks.

[DanG]

An often overlooked point is the many goals of the morning line maker and the people they answer to.

I had a long conversation some years ago with Brad Thomas and among other things we discussed the MLO creation.

We all know Brad and he could put out as sharp a ML as anyone in the country for Monmouth if predicted off-time odds were the sole criteria. Track management does not want the II-trip / wise guy / bias compromised / buried fig horse to be 7/5 MLO, even though the whales may inevitably push them there.

A major function of the MLO is to stimulate wagering…while maintaining some semblance of off-time odds…while not insulting owners / trainers who fill races with a reality based 65-1 slap in the face.

The line is much more “political” than we give it credit for and as Dave and others mentioned any serious database work should be normalized for scratches / fld size and surface changes to reflect reality.

As Ron Tiller has pointed out…Still my favorite MLO- stat is the Mountaineer >= 30-1 classic…

(Ron has data going back pre-civil war, but I’m only going back to mid 2005 here)

PS: The line maker is slipping…Only 3 winners last year and already 7 in 2007. :)

MLO SAMPLE WINS WIN% ROI MAX
>= 30-1 2213 11 0.5% $0.20 $182.00

[spilparc]

Amen. While some ML makers are quite adept at generating appropriate odds others are so off that ML's become less than meaningful.

I believe the morning line has great psychological impact. What if there were no morning line? What would happen then?

I would like to see a "no morning line" tried sometime.

[InFront]

I believe the morning line has great psychological impact. What if there were no morning line? What would happen then?

I would like to see a "no morning line" tried sometime.

I think for smart handicapping players no MLO listed would be a good thing. But for those players who do little or no handicapping or just use newspaper picks it may work against them. I don't like when I like a horse and then see that the MLO is one of the favorites. This usually means cause I like it for certain reasons the public will even bet it lower than it's already low MLO. If the public didn't have this MLO wonder what my average odds would be on such horses.

I started this subject days ago and it seems from all the posts especially the ones that contain large database test that in general the MLO and using it is useless for value. The only thing it can tell us which certain horses will be bet more than others especially when using the more acccurae MLO lower than 10-1.

[46zilzal]

Add up the equivalent probabilities listed in the morning line and you might be surprised at how far over 100% they can go.

Had a program handy and it added up to 125%.

[Maji]

Add up the equivalent probabilities listed in the morning line and you might be surprised at how far over 100% they can go.

Had a program handy and it added up to 125%.

Perhaps they don't deduct the track take when they calculate those ML odds?

[Overlay]

Perhaps they don't deduct the track take when they calculate those ML odds?

No, they do deduct the track take. That's why the combined odds of all the horses add up to more than 100%, the same as the odds posted on the tote board in each race. The deduction of the track take and breakage off the top of the pool before the odds are calculated and posted increases the percentage of each horse's part of the pool. For a simple example, take a two-horse race where each horse has had 50% of the pool bet on it before take and breakage, which would be natural odds of 1-1. Assuming an 18% take and dime breakage, that would mean that, instead of a $4.00 payoff, backers of the winning horse will receive only a $3.20 payoff for every $2.00 bet, which would be reflected on the tote board as odds of 3-5 for each horse. So the percentage represented by each horse's take-adjusted odds has now gone from 50% (1-1) to 62.5% (3-5), causing the combined percentage for both horses to add up to 125% instead of 100%. The principle works the same in races with larger field sizes.

[BMeadow]

The problem is that this analysis was done backwards. If one is hypothesizing that bet-down horses perform better, placing limits on odds is going to end up eliminating the horses you'd expect to do best. What you want to do is look at a narrow range of ML. If instead you look at horses whose ML was 5-1 thru 8-1, who similarly could've just as easily ended up going off at odds lower or higher, you see a significant difference in ROI and a huge difference in win%; .827 vs. .770 and 17.0% vs. 6.8%, over 185,000 horses. If you require a 20% shift in either direction, the differences are even more pronounced: .838 vs. .763 and 19.2% vs. 6.0% over 140,000 horses.

This suggests to me that there is some value in making an odds/ML comparison.


Disagree here (though I usually agree with your comments), because your analysis favors the lower-odds horses (e.g., a 6-1 ML who goes off at 3-1 OF COURSE is going to win more than a 6-1 ML horse who goes off at 9-1, with probably a better ROI since there remains a slight favorite-longshot bias).

My own take, having checked the morning line and final odds for every race in California run over the past 20 years: Nearly all the time, when a horse goes off significantly below its morning line, it's because the linemaker goofed. No matter how good he is, he is going to make mistakes.

In an article I did about this subject for Meadow's Racing Monthly, I gave some egregious examples of horses with high morning lines who had such obvious attributes (class drop, top trainer, leading jockey, a string of good Beyers, a suddenly improved race, etc.) and who went off at small odds that just lead you to scratch your head over how the guy could have goofed that badly. But we all make mistakes, including the morning-line guy.

Ron's post shows quite clearly that it's the FINAL ODDS that matter, not some employee's prediction of the rank order of the entrants.

[Jeff P]

posted by BMeadow: Ron's post shows quite clearly that it's the FINAL ODDS that matter, not some employee's prediction of the rank order of the entrants.Ahh... but doesn't the morning line drive the betting to some extent?... or rather doesn't the actual betting mimic the morning line... at least to some extent?


posted by InFront: I started this subject days ago and it seems from all the posts especially the ones that contain large database test that in general the MLO and using it is useless for value.I disagree with this... please understand I'm not picking on you... I'm disagreeing because I actually do make use of the morning line in my own live play.

If the morning line drives the betting to some degree then doesn't it follow it might be possible to use the morning line to predict whether or not final odds will be high enough on a given selection to offer value?

Understand that this prediction won't be perfect... But if such a prediction contained ENOUGH accuracy... then suddenly the morning line becomes VERY useful to me.

Most players use live odds at 2-5 minutes to post as a huge part of their play or pass decision making. Nothing wrong with that... or is there?

If that describes your own play... if you need to see live odds as post time nears before you bet... then you know there's a price to be paid in order to play horses: You end up spending hours on end each day watching the toteboard... essentially making you a "toteboard slave."

What if you could use the morning line instead? If you did, wouldn't you have a LOT more free time? What if you could accurately pre-bet (or pre-bet accurately enough) entire cards and still be consistently profitable? Wouldn't that free you from being a "toteboard slave?"

Imagine how nice it would it be to be able to play without the need to sit for hours on end watching the toteboard... maybe spend that time enjoying something else... like a day with your family... or round of golf? Again, how nice would that be?

In JCapper there are several odds line ratios designed for the purpose of aiding the player in play or pass decision making. One of my favorites is called MLOR3 or Morning Line Odds Ratio 3. MLOR3 is calculated by dividing ML Odds by the program's JPRMLProb Odds Line.

MLOR3 = ML / JPRMLProbOL

Using Workout Brilliance rank=1 horses with 1st time starters eliminated (the exact criteria I posted about in the Understanding Workouts Better http://www.paceadvantage.com/forum/showthread.php?t=38512 thread) as the basis of a spot play - with no other handicapping whatsoever - here is what my calendar year 2006 database shows for these horses broken out by MLOR3:


UDM Definition: WoBrill-1
Divisor: # UDM Def Divisor: 999
Surface Req: *ANY Surface*
Distance Req: *ANY Distance*

Running Style: ALL
WoBrill: MinRank= 1 MaxRank= 1
MinVal= -999 MaxVal= 999
MinGap= -999 MaxGap= 999


Data Window Settings:
Divisor = 999
Filters Applied: LT0 (1st time starters removed)

Surface: (ALL*) Distance: (All*)
From Index File: D:\2007\Q1_2007\pl_Complete_History_06.txt


Data Summary Win Place Show
Mutuel Totals 50583.60 47178.80 45908.50
Bet -56174.00 -56174.00 -56174.00
Gain -5590.40 -8995.20 -10265.50

Wins 4813 8897 12376
Plays 28087 28087 28087
PCT .1714 .3168 .4406

ROI 0.9005 0.8399 0.8173
Avg Mut 10.51 5.30 3.71


By: MLOR3

>=Min <Max Gain Bet Roi Wins Plays Pct Impact
-999.00 0.00 0.00 0.00 0.0000 0 0 .0000 0.0000
0.00 0.25 -183.10 526.00 0.6519 29 263 .1103 0.6435
0.25 0.50 -263.80 6040.00 0.9563 429 3020 .1421 0.8290
0.50 0.75 -1704.70 14878.00 0.8854 1142 7439 .1535 0.8959
0.75 1.00 -1359.00 11888.00 0.8857 1042 5944 .1753 1.0230
1.00 1.25 -1099.30 7324.00 0.8499 724 3662 .1977 1.1537
1.25 1.50 -701.70 4744.00 0.8521 477 2372 .2011 1.1735
1.50 1.75 -284.30 2886.00 0.9015 294 1443 .2037 1.1890
1.75 2.00 -266.50 2112.00 0.8738 201 1056 .1903 1.1108
2.00 2.25 -265.10 1498.00 0.8230 122 749 .1629 0.9505
2.25 2.50 57.00 986.00 1.0578 102 493 .2069 1.2074
2.50 2.75 -90.40 742.00 0.8782 55 371 .1482 0.8651

2.75 3.00 98.30 578.00 1.1701 41 289 .1419 0.8279
3.00 3.25 198.30 368.00 1.5389 37 184 .2011 1.1735
3.25 3.50 -68.20 294.00 0.7680 18 147 .1224 0.7146
3.50 3.75 125.20 298.00 1.4201 25 149 .1678 0.9791
3.75 4.00 55.40 170.00 1.3259 17 85 .2000 1.1671
4.00 4.25 -28.00 164.00 0.8293 13 82 .1585 0.9252
4.25 4.50 -6.60 122.00 0.9459 10 61 .1639 0.9567
4.50 999999.00 196.10 556.00 1.3527 35 278 .1259 0.7347
Notice what happens when MLOR3 is 2.75 or higher? The public bets enough of these horses in a fashion close enough to the morning line... resulting in actual post time odds that are high enough to produce a flat win bet profit for that tiny segment (4.5 pct of all selections) of the model's plays.

Please understand that this is an overly simple model. I'm using it for example purposes only. The models used in my own live play are a lot more involved (and hopefully a lot more effective) (and produce a lot more plays) than this model. But this example does illustrate how the ML can be used to good effect for identifying value in your own live play.



-jp

.

[InFront]

Yes you're right I shouldn't say MLO are useless but maybe as a stand alone factor they are. In many of my own spot plays I include a MLO rule just to try and get a higher average payoff on winners. This doesn't mean what kind of payoff I will get on a specific winner but in general the higher the MLO setting is the higher the average post time odds. But of course this has nothing to do with bottomline ROI.

But one type of rule I never use in any of my spot plays is a POST ODDS rule. Simple reasoning is if I am not willing to implement that rule by being as you said a "toteboard slave" race after race no reason to have one in there. Sure for backtesting queries using a odds rule is OK just for research but not for the real day to day actual betting.

[GaryG]

Linemakers have always hesitated to hang up realistic odds on standout favorites. How often do you see 8/5 or 9/5 on the ML off at 3-5? One exception is the linemaker at Pomona (is it Brad Free?) who doesn't hesitate to put up a 3/5 ML.

[dutchboy]

Does anyone know who makes the ML at various tracks? Several years ago the rumor was that the ML maker for Chicago area tracks also ran a private tip service. Don't know who started the rumor or why. My guess they were implying the ML was adjusted to hide horses. Seems like the ML in the Maryland area are never close. Doesn't matter to me. Just curious.



I was wondering if anyone has done these studies through a large database.

1. What percentage of all races are won by horses going off lower than their morning line odds?

2. What percentage of horses go off lower than their MLO vs those higher than their MLO?

3. What is the ROI if you bet all those horses to win that went off lower than their MLO vs what the ROI for those that went off higher than their MLO?

Any help would be appreciated. Thanks.