Kelly-How many use it?

[JustMissed]

I was reading in one of Quinn's books this past weekend about the Kelly Criterion. The part about earning a middle managers salary at the track got my attention. If you don't mind sharing, how many of you use the Kelly or something similiar.I am also curious as to how many of you use software to calculate bets-as to amount and type?

Although Kelly was demonstrated using win bets, I think I may try it with exactas.

I have been using a fixed 2% of my beginning week bankroll but my capping is getting better and I think I might do well to increase the bet size and think Kelly might be the answer. Your response will be appreciated.

Thanks,

JustMissed
:)

[JimL]

Just Missed, I think variance will play tricks with Kelly. Just because your ROI and edge may be whatever in your first 100 races does not mean it will stay the same in the next 100.

[JustMissed]

Jim, since I posted I did a web search of Kelly and found lots of sites. Some of the sites show the Kelly to be calculated like you're talking about where you calculate your edge and ROI to get your bet size as a percentage of your bankroll.

At some of the sites, they are talking about calculating a fair odds line, compariing that to the tote odds and doing a Kelly calculation to bet a particuliar horse in a particuliar race. It look almost like the "value betting" we see talked so much about here on this board.

I guess I'll have to do some more research as I thought the Kelly percentage to bet was calculated periodically based on the players win percentage and ROI but it appears it is also calculated to come up with bet sizes for value plays.

WHAT A GAME!!!!!

JustMissed
:)

[JustMissed]

The is where the article is that I read which talks about calculating your own fair odds line and using Kelly to determine the bet amount:

http://www.dmtc.com/tools/wt_kelly.html

There is so much interest on this board regarding "value plays", I thought this to be a very interesting use of Kelly.

I hope some of you more advanced players will read this and let me know what you think.

JustMissed

:eek:

[Lefty]

The "Kelly" is for any type betting you care to do. This article just happens to be about value plays but it can be used for any type bets; just make sure your bets are in "flat bet profit teritory"
At one point he writes something like you may be called on to bet 20% but you have to do it.(something like that)
To my mind Huey Mahl was the foremost exoert on "kelly " and he always advocated NEVER betting more than 8% no matter what bet is called for.
He also writes that every 15 bets or so you have to refigure your win pcyg and avg. payoff and adjust "Kelly" accordingly.

[Jaguar]

JustMissed,

You may find that Kelly is excellent for flat bets, as long as you cut the Criterion's determined amount in half. Otherwise, the Kelly amounts go up precipitously- not for the faint of heart.

But, for exactas, dutch 'em out to your win goal- as long as no combination pays less than $13.00.

All the best,

Jaguar

[JustMissed]

Thanks for responding to my post and for you good advice.

I went back and studied the Value/Kelly program some more and believe that all they are doing is using the value odds(difference between fair odds & tote odds) as the 'edge' percentage. They use this value to calculate the Kelly bet size.

I usually play w/p, exactas & trifectas and occasionally just a win bet. I think I might just up my 2% to 3% so I could go to a max of $120 per bet which is pretty good for a small player like me.

I didn't hit it but the 2nd race at Calder Saturday paid $607.80 for the $2 exacta and $8,744.80 for the $2 trifecta. If I had played that race I would have had a $10 exacta part-wheel and probably a $2 trifecta part wheel. If a guy could cash for $11,783.80 every now and then-Who needs Kelly anyway?

I hope both of you guys keep up with your posting.

Thanks,

JustMissed

:)

[Rick]

I use something approximately equivalent to a fractional Kelly method for bankroll building with the goal of making fixed size bets when the bankroll grows to the desired level. The bankroll building bets are about 2% and the fixed bets are about 1% of the target level. I don't adjust bet sizes until the bankroll changes by at least 20%. For example, with a $1000 bankroll, I'd bet $20, but I wouldn't raise the bet size until it was up to at least $1200, calling for a $24 bet. Actually, I'd more likely wait until $1250/$25 just because it's a more convenient amount.

The 2% level is somewhat less than half of what would be suggested by the usual Kelly calculation based on average odds because that calculation is only accurate for fixed odds payoffs. When you have variable payoffs, the correct solution depends on the variance of the payoffs, and can be substantially less than that recommended by the constant payoff formula. The correct amounts for several of my methods ranged from 0.25 to 0.66 as much as the constant payoff percentages.

Another method you can use (thanks to so.cal.fan for this one) is to bet a percentage of the maximum level your bankroll has been at. Testing of this method using my home-grown simulator and results for my main method showed that it results in about 20% higher bets than the percent of current bankroll method. So, for my method it performs pretty much the same as betting 2.4% of current bankroll. The growth rate is somewhat higher and the risk is somewhat higher but it still seems to be within acceptable limits.

[so.cal.fan]

Does anyone reading these posts bet different amounts according to how much you like the horse/race?
Put another way, is anyone successful doing it?
For years I struggled with this problem.
It made sense to me that if I liked a horse a lot and/or it was a generous price, a larger bet would be in order.
However............it never seemed to work in practice.
I would bet $100 to win on a horse I really liked......it would lose, I would bet $50 on a "marginal type selection" and it would win, often at a big price..........
Isn't the goal to get to a flat bet level we are comfortable with and that shows a decent profit?

[JustMissed]

Great question.

The answer might be in the Value Betting/Kelly method referred to in my post above, or an alternative that I propose.

If you like a horse better than the public, then it would make sense to place a larger bet, especially since it would be an overlay and of course have value.

The big question is how much to bet and I would like to run this by you.

Suppose you normally bet $100 to win, which is 4% of your $2,500 bankroll. For this race you like horse A, and make your fair odds to be 5-2. Suppose the public has the horse at 3-1. The tote payoff would be $8 while your expected payoff would be $7, a profit difference of $1 or 14%. So you would bet $115(114 rounded). If you made the horse at 3-1 and took the bet anyway, you bet your normal $100.

Since a value bet gives you an edge already, betting this additional amount would give you a double edge.

I have been doubious of value betting in the past, not because of its function, but because of the method of coming up with the fair odds line(read:your opinion of the horse-not the computer's opinion). The better you can make your odds line(better handicapping), the better the play all around.

Please note that I just made this system up and have not tried or tested it myself.

JustMissed


:)

[so.cal.fan]

JM
Your idea is good, but again, I go back to, will it work in reality?
I doubt it.
I am certainly not going to argue against the merits of making a good line, but it has it's limitations.
Example:
Let's use a big pool track like Santa Anita. I like a horse I put up as 3/1. A minuet before post time it is 4/1. I make my play.
The horses go into the gate......oh oh....my selection goes down to 7/2, the flag goes up........oh s**t ! I'm seeing 3/1. They go past the wire.....damn it.....5/2! An underlay.
To add insult to injury......the horse loses! Damn it again, I should have even bet it or at best, I should have bet less on it.
What to do?

We have to take a stand........decide what we are going to do....and stick to it in the face of all arguements, and that includes
public opinion/odds.

[JustMissed]

Just another thought.

If you are playing a longshot horse, let's say going off at 6-1 and your fair odds line has her at 5-2, your bet increase factor gets pretty large-$700/$350=200%. 200% x $100 bet is $200 which is double your percentage bet amount.

Someone posted here that Mr. Mahl said to never bet over 8% of your bankroll. 8% of your $2,500 would be $200 so you would be right in line with his thinking.

If you are a pretty good capper I don't see how you can get hurt going up to 8% every now and then.

Let me know what you think.

JustMissed


:)

[so.cal.fan]

Just Missed:
Yes, I do think that is a workable idea.
Provided.....you have the courage to stick to it.
Look, if we bet to win.....we don't make much profit on our 3/1 and under winners. It is the occassional long ones that make our meet for us.
Has anyone ever tried this? Let us know if it is workable, please.
I may just try it at Oak Tree.
1/2 a bet on all horses under 5/1 and a full bet on all over.
It will take some courage on my part, because it is totally against my character....but I'm willing to try, unless someone convinces me that it is a terrible plan. LOL

[JustMissed]

Excellant point. Like ole Bill Clinton used to say, "I feel your pain".

I used to be very confused about betting value horses for this one reason which is very, very important. I used to think that the value(or amount of overlay) had something to do with the horses ability to win the race. Once I got the concept correct in my mind, I have looked at races in a completely different way.

If I could make a bold statement, Tote Odds or Public Odds or Track Handicapper's Odds have absolutely nothing to do with horse winning the race. (I say this with the exception to trainer, jockey, or owner manipulation based upon the tote).

If I could go forward with this I think it might be good therapy for me.

1. When you make your odds line you are doing nothing more that ranking the order of finish that YOU, THE HANDICAPPER, believe the horses will finish. Nothing more, nothing less.

You express this order of finish as odds in order to be able to more easily compare your opinion to that of the track capper and the tote. Nothing more, nothing less.

2. Now your have your betting consideration. These consideration are numerous as you know and can include such things as:

a. You gave two horses the same low odds. Which do you bet? Both or one which the tote gave higher odds. This is a value consideration.

b. Your odds on the best horse are the same tote odds. Do you bet or pass? This is a value consideration.

c. Your odds on the best horse are greater that the tote odds. You know this answer.

d. Your odds on the best horse are much, much lower that the public's odds? What is it that you saw or know that the public missed? How much do you bet? Will they take my car title through the window.

To sum up the process, I think once you have your fair odds completed, your handicapping is complete except for identifying the reason that your fair odds are different than the publics' odds. In other words you know why your odds are more or less than everyone elses'. If you don't know why then you need to keep on till you do know why.

Then you move on to Step 2, which is the betting decision where you really are making a risk/reward analysis. How much to bet, if any. What kind of bet. And so forth.

I have not proved this to myself mathematically, but the reason value betting works so well that if your handicapping ability stays about the same, you are only going to win so many races and are going to lose so many races. But with value betting, when you do win the cashier will give you more money back than you should get back based on the risk. That's it.

Let me know what you think.

JustMissed
:)

[Rick]

socal,

The theoretical answer to how much you should bet would be according to what your edge is on that particular bet. If the ROI was the same on all bets, then a bet at twice the odds would call for a bet half the size. BUT, that's not usually the case. In fact almost everyone has a higher ROI on their longer odds bets. With enough data you could probably determine how your ROI varys by odds and that would provide the most accurate answer. But if you assume that your ROI is twice as much at twice the odds (a pretty reasonable assumption I think for most bets), then a constant bet size is the correct solution anyway.

[GameTheory]

Good record-keeping helps with these questions. First thing you need to discover is if horses you "like more" actually win more. Do they still win more when they are "big overlays"?

In my opinion one's personal odds line should not be fixed, but take into account the actual odds as well as the reasons you like a particular horse. If a horse is 2-1 on your line and it is going off at 10-1, it is probably not an overlay but a bad bet UNLESS the reasons you favor that horse are completely hidden from the public -- it looks bad on paper. If you are scratching your head wondering how the public can let this great horse off at 10-1, then it is you who are missing something and you best stay away. I find that if the public isn't acting as I would expect that they would, then neither will the runners.

[Aussieplayer]

Game Theory's post is the most pertinent so far in this thread, IMO. I can't believe record keeping hasn't been mentioned yet!
Oops! Rick already has!! lol

Records (could) in fact surprise you totally!! For example: most (good) players I know down here are always talking about "low risk," and betting on "standouts." Gives you a nice high win%, which is good. And if you are good enough, you can get a profit.
Thus, such players would probably pass a race that they have rated as "tight" or you might say, "open."

I am working on a particular set of ratings at the moment. Last Saturday I had a classic pass race for people who bet standouts, or "best horse" bettors. I had something like 4 horses virtually rated together, like, less than a length (from memory - don't have my records in front of me). Top (VERY marginal) horse is weak 3/1 fav. on the board. Closely rated second choice is 25/1. 3rd and 4th choice also closely rated are both floating around the 9/1 or so mark.
Forgetting exotics at the moment, and looking at either playing one horse or passing, most "best horse" players I know would pass. I say, bet the second choice at 25/1!! This open race was a no brainer IMO, for the one horse to win player. Of course, I use this particular race as the second choice won, LOL!

Anyway....I digress (although the above probably shows how you can play "value" without a "formal" betting line).

The point is, as I build up more race records of these particular ratings, I want to not only be able to answer the usual questions, "how do they go at 7f compared to 10f, how about wet tracks, etc. etc."
I want to know.....if I have a standout rated 3 lengths ahead of the field, and it's favourite on the board - is that a good bet? What about when it's not fav?
What about when I have half the field rated within 3 lengths? Is that a pass? Or is it a bet for a price? Or is it a bet on the fav?
What about when my top 4 doesn't include the fav??? Is this the put the money down situation we would imagine it to be? Or have I stuffed up the race too much, and the fav will win????

I know that the answers are going to be UNIQUE for these ratings alone, and this is why personal records are the only "real" answers you will ever get. Everything else is just nice theoretical discussion material. But, it can give you new ideas at least! :)

Cheers
AP

[JustMissed]

[QUOTE]Originally posted by Aussieplayer
[B]Game Theory's post is the most pertinent so far in this thread, IMO. I can't believe record keeping hasn't been mentioned yet!
Oops! Rick already has!! lol

Since I am the one that started the thread with a question about 'Kelly' and bet size, I find it odd that you did not find my, and the others', subsequent post about the subject to be "pertinent".

Maybe I'm not as an accomplished player as you are but I do try to put some thought and effort in my post and believe my questions and the others' responses are very "pertinent", at least they are to me.

JustMissed
:(

[Rick]

AP,

You're SO right about the answers being unique to the method you're using. Most of us can provide only answers that relate to the exact method we're using, so others should be careful about blindly applying those recommendations.

[Aussieplayer]

JustMissed,

Sorry - by pertinent I meant, "being of striking appropriateness" rather than, "being the most relevant."

Hope you didn't take it the wrong way. I was wanting to post, but hadn't had the time, and Game Theory's post was one that made me say, "YES!!" lol

I just have strong feeling about records providing the answers to some of these questions. Rick puts it far better than I did, where he says, "Most of us can provide only answers that relate to the exact method we're using, so others should be careful about blindly applying those recommendations."

That's what I was trying to say, lol

Example: Steve Fierro uses a variant of Kelly Betting, as found in Barry Meadows book. If you check out his book, you will in fact find (IMHO) no real justification for kelly betting over flat betting. Six one way, half a dozen the other. Whatever pickles your gherkin.

Now, that's for HIS betting.

The idea is (IMO), to read about other's methods (books, here on PA etc), and their experiences - to get ideas that you mightn't have thought of. Then check 'em out over YOUR betting style.

Sorry if I came out the wrong way, by the way.....I have enjoyed all your posts.

Cheers
AP

[Aussieplayer]

Another example is exactas. Rick provides a cool, simple little betting method. I haven't thought of it before, and I'll check it out. Something I have found out on the ratings I am developing/using at the moment, is that whilst win% is okay, but not spectacular - I am noticing that I am getting the exacta in the top few frequently. Some people have ratings where the winner is in the contenders or it's not - and exactas are a no go. I (hopefully) have the opposite here. A reasonable win betting set of ratings, with (hopefully) a good exacta betting set of ratings.

Another one: I think you guys generally learn that "boxing is for losers," just as we do down here. BUT: I've found with these ratings that when the fav. (and preferably 2nd fav.) is out of my top 3, or even top 4 (which is not too often of course) - I can box away!! Stuff trying to put tickets together in order!!!! I'll take a box exacta & box trifecata any day, (1) IN THIS SPECIFIC SITUATION, and (2) WITH THESE PARTICULAR RATINGS.

For "normal" races, perhaps something like what Rick suggested will work for me? I'll certainly find out, as I have all the races ready to check for such ideas.

Sorry I've wandered a bit away from Kelly betting. It's only my opinion, but I do not think that many players would benefit from using Kelly (I'm talking here of odds line Kelly, not working out a percentage of bank to bet on). I don't anyway.

Have you (JustMissed) read, http://www.netcapper.com/TrackTractsArchive/TT010302.htm

Might be of some interest.

Cheers
AP

PS. I posted this ages ago - which is pseudo Kelly:

Just thought I'd post this in case it's of any use to personal odds line users who haven't considered "grading" there bets. I think you get the same sort of idea from Barry M. (I think he bases it on kelly).

Anyway, 10 years back, one of our biggest bookies (Mark Read) launched his product, "Read's Ratings." These give you the contenders in each race with the rated odds. Average of 4 per race. Sometimes 5 (esp. big fields). Sometimes 3 (esp. short fields). You get them either by subscription, or free from a betting outlet that subscribes, which several do.

Profits made from the launch were donated to a charity. Can't remember exactly how much - but I think it was something like $30,000.

The starting bank was $50,000 and the units were $100.

There was a scale of "base uinits" according to rated price, which follows:

< 2/1 10.0
=2/1 9.0
9/4 to 5/2 8.0
11/4 to 3/1 7.5
13/4 to 7/2 7.0
15/4 to 4/1 6.5
4/1 to 5/1 6.0
11/2 to 6/1 5.5
13/2 to 8/1 5.0
9/1 to 10/1 4.5
over 10/1 4.0

(sorry if the table comes out bad!).

The formula used to determine a bet was simply: "available odds" divided by "rated odds" multiplied by "base units".

Example: Redelva had a rated price of 5/4 and an available price of 2/1

2/1 divided by 5/4 = 1.6
Look up the table and you see that 5/4 has 10 base units.
1.6 * 10 = 16

If you remember, units were $100, so the bet on Redelva was $1600.

As you can see, very kelly, but it might be of use to someone!

Cheers
AP

[JustMissed]

Thank you for your post and for the link to Gordon Pines article. I enjoy Gordon's writing and try to read as much of his material as possible.

I am going to investigate the possibility of a fault in the use of "bankroll" .The dictionary definition of bankroll is as follows:

1. A roll of paper money
2. A person's ready cash
3. To underwrite the expense of(a show, for example).

Something tells me that the use of a bankroll by modern horseplayers is a carryover from the early part of the century when a player had X amout of cash to take to the track and if he lost that amount he could not play the rest of that day, week or season.

I suspect that the use of a "bankroll" as a determination for bet size may penalize the modern player by limiting his bet size due to the use of an antiquated methology.

You obviously are a great record keeper and I would suspect you never "tap out". Suppose your records indicate that your win percentage is 35% with a $2.60 average mutuel. You play with a $1,000 bankroll and make $100 bets and show a profit and never tap out. My question would be, if you are making money with $1,000, why not increase your bankroll to $10,000 or $100,000 and make 10 times or 100 times more profit than you are making now. Assuming your handicapping and betting methods remain the same, it would appear that all most good players need to do to make more money is to increase their bankroll(or rather their bet size) and keep playing the same way.

I am making the assumption that the bettor does not have to have $10,000 actual cash in his pocket but has access to it via savings, his job, credit, equity, etc. In order words if you are a winning player then bet like you have $10,000 in your pocket instead of only $1,000.

I will go back and look at some of the "tap out" studies but I am a little suspect of those because I suspect the modern player does not have as long of losing streaks as the computer might randomly generate.

I am a new and learning player so unfortunately I do not have a long enough history to do this test, but is someone who is a winning player with a flat bet history could go back a recalculate their performance as if they had bet 10 times the amount, I would love to see the results.

Please let me know what you think.

JustMissed
:)

[Lefty]

Aussie, I think there's a HUGE diff in Kelly or other run-it-up methods and flat betting. It's how you get to large flat bets by starting small. It's better to use a small br of a hundred or two or three and use a kelly or similar method to runitup so you can get to those big $100 $200 bets. It's mighty hard to start with say, a $5 flat bet and make enough to get to those large bets. If a small
player has any hope to get to large bets it's with kelly or similar.
I'd rather lose a small amt trying to runitup than start with large flat bets and lose that BR.

[Dave Schwartz]

Just Missed,

>>>I am making the assumption that the bettor does not have to have $10,000 actual cash in his pocket but has access to it via savings, his job, credit, equity, etc. In order words if you are a winning player then bet like you have $10,000 in your pocket instead of only $1,000.<<<

Actually, most people do this in reverse... That is, they earmark (say) $1,000 to their "bankroll" and designate their unit size at (say) $20. Then they say, "I've got 50 units."

The problem is that when the lose $400, they say, "This isn't working," and quit.

My point is that their "effective" bankroll may be far less than what they designated as their "bankroll" to begin with.


Regards,
Dave Schwartz

[JustMissed]

Excellant observation Dave. Maybe horseplayers need sports psychologist just like the professional golfers do.

[Dave Schwartz]

JM,

LOL - Boy did you get that right. We could all use a little of that.

Dave

[Rick]

Lefty,

I agree 100% with what you're saying. Use Kelly only to build the bankroll and select your level of risk according to your own resources and risk tolerance. Once you reach your desired level of flat betting you want to take profits, and that changes the whole situation anyway. I prefer using simulation to verify my opinions in this area


Dave,

I've seen many go down the same road you mentioned. That's why I like to start small and play with "other people's money". My betting is much more aggressive when I'm risking only small amounts. The first professional that I knew started out with $10 bets and a $400 bankroll and was betting $200 when I knew him. I have known others that started with as little as $200 and $2 bets. If your method is a good one, you'll get there. I'm currently getting 28% winners and 30% profit (483 plays) and my bankroll is growing exponentially. If this continues, in 6 months I should be in the "big leagues".