The Fair Odds Line And The Unknown Factor

[trying2win]

I don't personally use a fair odds line. I haven't got a clue how to make one, nor am I really interested in learning how to formulate one. I just use the morning line odds as a guide. Albeit, my critics in the past have admonished me about using the ML, but it works okay for me. Anyways, I didn't start this thread to debate on whether one should use a fair odds line or morning line for handicapping races, but rather to satisfy my curiosity on something.

I've read chapters in some handicapping books, as well as some posts here in the past, about making a fair odds line. One of the more interesting takes on this subject, is the odd author here and there who follow the usual formula in making a fair odds line, but add a little twist to it. Let's say they formulated the following fair odds line for their top three contenders in the following fictious race:

HORSE A -- 2 TO 1

HORSE B -- 4 TO 1

HORSE C -- 7 TO 1


The difference I noticed, was that some author's did things a little bit different. They demanded not only that their contenders be at, or higher than the above odds near post time...BUT, ADDED ANOTHER 50% REQUIREMENT TO THEIR FAIR ODDS LINE BEFORE CONSIDERING A BET. This is because of what they called the 'UNKNOWN FACTOR'.

So, can anyone give me any examples of what these authors possibly meant by the 'UNKNOWN FACTOR'? It's been quite awhile since I read an article on this particular segment of making a fair odds line, so I can't remember exactly what they were referring to. Did they mean unknown or unexpected things might happen in a race such as:

--The horse they want to bet, might possibly get squeezed out at the starting gate, or carried 3-wide at the first turn?

--They don't know the possible effect of first time lasix on a horse?

--The horse got claimed last time, and they're not quite sure if the new trainer will have a good, bad or just marginal effect on a horse's effort on a particular race day?

ETC. ETC.....

It would be interesting as well, to see if any PA members also subscribe to this theory of adding on an extra percentage to their fair odds line before making a bet. Some of the responses by PA members who are well versed on this aspect of making a fair odds line, might even benefit some members who are neophytes on this subject, and would like to learn some more things in this area.

Thanks,

T2W

[yak merchant]

I can't speak for anyone else, but I run monte carlo simulations to figure out what percentage of the time a horse will win. If horse A wins 5000 of my 10000 simulations, I assume his "fair odds" are even money. But because I want to make money and not just break even, and there is no way I can be certain my projections are 100% accurate I add 50% on top of that before I make it a bet. So on that horse I'd require 3 to 2 before I'd consider a bet. I don't know if there is any reason or reasons like bumping, or Lasix or whatever to contribute to this 50%. I just don't think anyone wants to bet into a break even situation when there is no way their "fair odds line" can be 100% correct. If you bet every 3-1 horse that you think has a 25% chance of winning you better hope your odds line is wrong or you'll never win a dime.

[sjk]

I require a 65% overlay percent on all bets. Why?

1). I might be wrong.

2). Backtesting shows me that if I demand that advantage I can expect to be successful.

3). This is mathematically equivalent to using the public odds as an ingredient in my odds.

4). My odds are sufficiently different from the actual that I still get to play almost every race.

[GameTheory]

Yak & sjk --

Do you have a minimum probability or fair odds value to consider a bet (i.e. do you bother to look for overlays on "fair odds" 20/1 horses?). I like the 14% (probability) & up rule -- the margin of error is just too small on longshots, even if they are big overlays...

[alysheba88]

Demand at least a 50% overlay myself. As others have said its to provides a cushion.


Odds making (and a big believer in making your own) is art not science (to me anyway). The more odds lines you make the better you are at it. I use the Fierro templates, much easier for me.

Anyway, many of my win bets end up being 100-200% overlays. I only want to bet situations where I feel I have a real significant advantage. There is just so much to overcome (possible late odds drop, the takeout, just plain bad racing luck, whatever), that I want to have a cushion. Betting with only a 8-10% advantage is cutting it way too close to me. You have to be perfect and make no mistakes. Too stressful.

[sjk]

I bet mostly exactas. Horses in top spot with at least .075 probability to win. Horses in second spot with at least .04 probability to finish 2nd.

[JustMissed]

You guys just crack me up. I'm going to vote "horse A" as Horse of the Century.

When I read the following I laughed so hard I thought I was going to crap in my pants "If horse A wins 5000 of my 10000 simulations, I assume his "fair odds" are even money. But because I want to make money and not just break even, and there is no way I can be certain my projections are 100% accurate I add 50% on top of that before I make it a bet."

I hope the guy from New York that had the heart attack recently doesn't read this thread-he might blow out his stent.

Thanks for the laugh,


JustMissed
;)

[Jeff P]

Have to concur with what others have said here. Oddsmaking is as much art as science. This is a tough game to beat in the first place. The concept of demanding a premium before making a bet is a good one.

For me, the unknown factor is the chance that I am wrong in my assessment of a horse's chances. I like to build that premuim into my odds line before printing it out. Example: I think a horse has a 20 percent chance to win. Fair odds would be 4-1. I've programmed my own software to mark up those fair odds before I see them. The output I get shows the horse at 6-1. This is what I see. When I make my decision to play or pass I don't have to bring the "mark it up 50% first" into my decision making process. All I want to do is compare the printed output against the current odds.

The fewer things there are to think about come crunch time the smoother things seem to go.

[Steve 'StatMan']

I want to get back to making fair odds lines, or at least being more cautious with value estimates, this season.

That unknown factor covers ones self for many things. Many of them were mentioned. Things you didn't know, or couldn't know, the bad trips, etc. It also covers you in case you're not really a good line maker. That may be the most important piece of all.

Let's say you make a horse 2/1 in your line (33% chance to win). Your 50% bonus/overlay would require 3/1 odds to bet it.

Now, remember that true 3/1 horses win 25% of the time. If you are wrong about your 2/1 shots that are 3/1, and they only win 25% of the time, assuming you bet the same amount each time, you'll still break even.

If you're exactly right about your 2/1's that go off at 3/1, you advantage is (.33 * 3) - .67 = . 22, a 22% edge in these situations.

Also note, that you have a difference of 8% between making your estimate (33%) and your break even(25%). This means that you can be off in your estimate by 8%/33% = about 25% and still at least break even in these situations!

I use this thinking everytime I need motivation to make lines, as well as to encourage me that this game can be beaten, and look how much room I have to play with, if only my estimates are reasonably decent.

I haven't made many lines the last few years. Formal lines aren't always necessary, as long as you are doing enough analysis to be certain you are really getting value. The last few years, I've let work/time constraints get in the way of being thorough enough when making my estimates. I don't want to make those same mistakes this year.

Thanks for bringing this up T2W! I needed that mental boost to get me ready for HAW and motivated to make lines again!

[Exactaman]

I agree, the unknown factor is the accuracy of your line.
If your line is on the money, then the unknown factor = your profit margin. Now that is unknown.

I go for 50% on exactas, 100% on win bets. at least natural odds for the win spot, up to 2% points less for place. As Steve says there are other ways to find value--an M1 pick 3 tonight paying over $600 for a 15-1 shot over two odds on faves comes to mind--but the odds line forces you to look for value and is a worthwhile exercise, probably even if your line is wildly off--if your handicapping is that bad, you'll probably lose more without a line!

[jackad]

Exactaman,
When you speak of requiring at least natural odds on a win bet, do you mean that in a six horse field you require odds of at least 6/1 to make a bet, in an 8 horse field odds of 8/1, and so on?
Thanks.
Jack

[trying2win]

Thanks to everyone for your posts on this thread. It gave me a greater understanding about why a handicapper likes to add a percentage onto their fair odds line, because of the unknown factor.

T2W

[yak merchant]

Yep them statistics applications are funny. Next week on pace advantage comedy hour I'll talk talk about discrete time Markov Chains and the Chapman-Kolmogorov equation.

[ranchwest]

Originally posted by JustMissed
You guys just crack me up. Thanks for the laugh,


JustMissed
;)

Look, you may think this is funny, but just watch the paddock judge. He always has a piece of paper in his hand and holds it up near the horses' eyes. The piece of paper has the horse's name and morning line odds, so that the horse knows how much of an overlay or underlay he is.

Really, man, I fail to see the levity of this conversation.

[Exactaman]

Originally posted by jackad
Exactaman,
When you speak of requiring at least natural odds on a win bet, do you mean that in a six horse field you require odds of at least 6/1 to make a bet, in an 8 horse field odds of 8/1, and so on?
Thanks.
Jack

an even shot, all horses being equal. actually 5/1 in a six horse field, 7/1 in an 8 horse field etc. ie. in an 8 horse field, if I rate a horse 8/1 on my line he won't see top spot action no matter what his tote odds.

[andicap]

How in the hell can you differentiate between 8-1 and 10-1 in making a line?
An 8-1 horse has an 11% chance of winning., A 10-1 horse has a 9% chance.

Can anyone honestly estimate it that fine???

Even 8-1 and 12-1 seems iffy, 11% and 7.7%? C'mon! You can honestly tell me you can tell if Horse A will win 1 out of 9 races and Horse B one out of 13.

Now that's funny.


When I do make a line -- time permitting -- I'll try and go through this exercise.
1. How many contenders, based on my records and the competitiveness of the race at hand are there? Lets say 5 in a 10 horse field. I'll then give a rough estimate of the "field's" chances for winning. If it's a race that's tough or seems on the chaotic side I'll give the field 20%; if the rest of the field seems lame, maybe 10%.
I did like CJ's idea for alloting % depending on your number of contenders and the field -- from another thread -- and will play with that idea.

2. I don't agonize and say Horse A will win 1 out of 5 times. How the hell does anyone know how many times a horse will win in 10 runs or 100 runs? And besides they ain't running 100 times. Just once. So I'm just trying to compare the horses.

A marginal contender gets a sliver, say 6-1. A stronger, but still outside choice gets a slight bump, maybe 5-1. If the three main contenders are relatively close, I'll split the remaining points, say I began with 80 and the 6-1 and 5-1 equal about 33, I'm left with 47 pts. That's 15-16 apiece. or 5-1 or so.

Whoa! something's wrong here. Got to readjust. I probably over valued the weaker contenders so I shade them back to about 8-1 or so. That's 22 points freeing up 11 to give me 58 for the 3 main ones or 4-1, 4-1 and 9-2 maybe if I think the first two are tied and the third a shade worse.

I won't try and cut it too fine because I'm not that good and I'm missing a lot of information anyway -- maybe there's a trainer angle which if I was hip to would make the 4-1 horse a 2-1 shot. I can't definitively say one is 3-1 and the other 4-1 so I'll get myself a rough estimate, build in an overlay and hope I'm right. If I'm wrong a lot my bankroll will tell me and I'll go more conservative by demanding larger overlays.

I'm not saying the top two will win 20% of the time, but trying to make a realistic estimate of the differences between the horses -- handicapping them if you will.

3. Then I demand at least a 50% overlay and sometimes more if the race seems more chaotic. I'll also demand a larger overlay on my lower choices, from 3rd down than from my top 2 because I'm not bad in pinpointing the top two, but less accurate from there.

I won't be a horse below 6-1 on my line - excpet maybe an action flier if he's a REAL longshot --(and then I want to know WHY he's not being bet if I think he's a contender. What do I know that others dont? I ain't THAT smart. I'd better have some real insight that's not apparent to the "smarties" in the crowd. Or maybe the c rowd has fallen in love with two horses and really "overlooked" one of the fringe contenders, letting them off at 30-1 or so when they can win with a bit of luck, the right pace scenario, and some form cycle improvement.

[GameTheory]

Originally posted by andicap
How in the hell can you differentiate between 8-1 and 10-1 in making a line?
An 8-1 horse has an 11% chance of winning., A 10-1 horse has a 9% chance.

Can anyone honestly estimate it that fine???

Even 8-1 and 12-1 seems iffy, 11% and 7.7%? C'mon! You can honestly tell me you can tell if Horse A will win 1 out of 9 races and Horse B one out of 13.

Now that's funny.



It's all a matter of granularity. If you have a computer making your lines, maybe it can make finer distinctions more accurately.

Take a white piece of paper and draw a straight line on it with dark ink. Now make a few dots with your pen as close as possible to the line -- some on each side. Are you honestly telling me that you can tell which dots are on which side of that fine line?

Of course you can -- it is plain as day -- even though two dots on opposite sides might be only a millimeter apart. Your eyes can make that distinction quite easily. If the tool you are using to make oddslines can make distinctions as fine as 2-3%, why not?

Nevertheless, your point in general is well taken, and it is certainly harder to tell the difference between a 15-1 horse and a 20-1 horse than between a 2-1 and a 3-1. Which is why I like to stick to horses I have assigned at least 14% to.


But this stuff about how its hilarious that someone would try to make such distinctions and asking questions (or doing simulations) like "If this race were run 100 times..." always comes from a place of ignorance in my opinion. Like it or not (and in defiance of any scoffs), those techniques work if you do them right, and can often make subtle distinctions. Making subtle distinctions (as well as gross ones) is the REASON you resort to such sophistication.

"Keeping it simple" is nice, but if you can gain an extra couple percent profit by making it complex, why not if you're willing? A couple percent can be a whole lotta money...

[alysheba88]

Andicap, I agree. Thats why I assign odds to only three or four horses, using between 79-81% of the odds line and never assigning a contender more than 6-1.

In otherwards the method Fierro talks about in his book

[Exactaman]

Andicap, I agree with some of your points but not with others.

Really assigning any % number to any horses' chances is a leap of faith to begin with. There's an element of unreality to the whole thing, not least of which because you will never know for sure what the real number is.

A line is just an aid to betting. I'm a longshot player. This influences the way I make my line. In the search for longshots, I'll take horses that don't fly off the program. If I feel a horse has an even chance to win, that is 10% in a ten horse field, I'm willing to make a bet provided i get my overlay odds. IMO the tracks I bet are conducive to this kind of style--they feature very competitive races that often leave handicappers scratching their heads, with a length separating the top seven finishers, figure horses finishing out of the money... with my style these are all music to my ears. In fact in the past I would go down to an 8% shot on top, with spectacular results on one occasion, but i have settled on the even chance minimum now. Often it will mean I would consider win action on at least half of the horses in a race--these are the kinds of races i like to bet.

Since I will bet horses going down to an even shot, I make the kind of distinctions you describe as a matter of course. Look at it this way--an 11% shot is 22% stronger than a 9% shot, and in a ten horse field, that is a significant difference. combine two of these in an exacta and the difference is even more significant. Is it exact? No it's just an estimate. For me, it may be almost a subjective procedure--if I think a horse is OK, but for one objective reason or another know that i don't want to bet him, he may get a 9% line, which will see him used underneath only, or even a 7% rating, where he won't get any play. In a sense the line is simply what I think I should bet as much as what i think the horses chances are. if i'm doing it right these should be the same. once i've settled on a line, i'm loathe to second guess it--i'll never back off the 30-1 shot because i can't figure out why he didn't get the play. if he's on the line, and the odds are there, he gets bet, otherwise he shouldn't be there in the first place.

I often go through the kind of adjustment procedure you describe as well. I never make a conscious assignment of a certain number of points to contenders though. I simply consider each horses chances, then adjust. occasionally this leads to fine distinctions, but every horse has to have some number in the end.

[ljb]

I have a question for folks on this thread. Have you ever tracked your win percent for say your 4-1 shots going off at different odds?
That is:
You make a horse 4-1 your play. (with or without your percent overlay)
Now what frequency does this horse win at 4-1,5-1,6-1 etc.
And at what frequency does this horse win as an underlay?
Try tracking and compiling this data you may be suprised at the results and then again you may not.
ljb

[maxwell]

The public & morning line does a better job of 'capping my selections than I do ! The morning line is actually the champ. If the morning line is 3-1 and I peg the horse @ 6-1 and it wins @ 8-1, it's pretty clear who knows up from down. It may be that I am a taskmaster and demand a lot in the way of factors.

I do best when I 'cap horses between 3-1 and 8-1 in the morning line - the ML saves me a great deal of time and grief.

I use two betting lines only : 6-1 and 12-1.

Horses that qualify as the top 3 speed horses at the class, distance, and value in the last six months are playable at 6-1 and up. Those at 4th and up are playable at 12-1 and up. There are other factors to be weighed for sure, but this is where I start the 'capping process.

[Rick]

In my experience, it seems to be much harder to find legitimate overlays among the longer odds horses. That's because you're really trying to estimate probabilities, and there will be a relatively constant error associated with that process that affects low probability horses to a much greater degree percentage wise than the higher probability horses. For example, its probably impossible to tell whether a horse should be 19-1 (5% probability) or 24-1 (4% probability). But if you're comparing a 3-1 (25% probability) against a 4-1 (20% probability), you probably have enough of a difference to call it an overlay. The best way would probably be to set a minimum absolute difference of a few percent and a minimum percentage difference on top of that. Also, most methods are fine-tuned so that the probabilities are most accurate on the top rated horses, which contributes even more to inaccuracies on the lower rated horse.

[PSTfredk]

wouldn't the best answer to the puzzle is to never bet a horse
to win as horses don't bet on people.

well bak to betting on horses when i handicap a race i don't
make my own line, i just look at the race and make a hard
fast rule of NEVER betting a horse to win unless it going off
at 3/1 or better. it cuts down the plays and still gives you
a money making odds. if making your own line was that important
to racing there would be some genius out there just selling
his odds line instead of selling he's horse picks. there are many places where you can buy sevice with there so called odds
line and they don't do any better than a public hanndicapper
so much for my two cents. this will not give you a lot of action
but you should make money.

[alysheba88]

PST,

But the whole beauty of making a line is that is not something that is reproduceable (or should not be). Its subjective. Just like the handicapping process itself.

While most of my plays end up on horses over 3-1, I do catch a decent share of winners on horses I make 8-5 or so on my line and end up going off 5-2 or 3-1.

[andicap]

Originally posted by Rick
. The best way would probably be to set a minimum absolute difference of a few percent and a minimum percentage difference on top of that. Also, most methods are fine-tuned so that the probabilities are most accurate on the top rated horses, which contributes even more to inaccuracies on the lower rated horse.

Can you explain to my alcohol addled mind what you mean here. Maybe an example.

Thanks.

[Rick]

andicap,

Say, for example that you look for at least 2% absolute difference in probability and also a 50% relative overlay. Now, if you're looking at a 29-1 horse (.033 probability) that you think should be 19-1 (.05 probability), you have the 50% overlay (30/20) but you don't have the minimum 2% absolute difference (.05 - .033). But If you're looking at an 8-1 (.111 probability) that you think should be 5-1 (.167 probability), you've satisfied both requirements. Of course, the minimum absolute difference should be set according to how much error there is in your probability estimate and maybe should be even higher than the 2% I mentioned.

The second part of the statement has to do with my observations over the years that most of the time I have a choice between making the top choices better or trying to go for the most accurate probabilities for the field as a whole. I always opt for the most accurate probabilities on the top choices (usually the top two), since those are the ones I'd prefer to be betting. It's just my opinion that you can't have it both ways from what I've seen. It may be possible to get both at once but I've never seen it.

[Rick]

A further elaboration for the mathematically inclined. According to statistical theory, the standard error of a proportion (or probability) is at its least when the proportion is 50%. Since the top two horses in a good method usually get a little better than 50% winners, I try to predict their combined win % since it would have the least error of any combination of horses. This can be really important if you have a small sample.