Ganch, can you do Kelly criterion for dumbies?

[robzilla]

Last night I read Ganch's articles on Kelly criterion.

I'm not very good at math, but I think I get the basic concept.

my question is how do u calculate Win Probability/Edge?

[pat venditto]

kelly is a bullshit theory. its not handicapping it's a modified martingale.

[MonkeyF0cker]

kelly is a bullshit theory. its not handicapping it's a modified martingale.

That's so far from the truth, it's silly.

[durito]

my question is how do u calculate Win Probability/Edge?

that's the million $ question

[robzilla]

does it mean that the Red Sox 61-46 edge is 75.4% vs. a .500 team?
what is it based on?

[Sinister Cat]

does it mean that the Red Sox 61-46 edge is 75.4% vs. a .500 team?
what is it based on?

it's based on whatever model you develop-- you still have to cap the games to determine the probability of the outcome. Kelly just tells you how much of your bankroll to bet once you've figured that out.

[durito]

No.

Kelly is a staking strategy, it tells you nothing about your edge/win %, those are factors you have to figure out on your own (with a model or some other handicapping method)

[LT Profits]

In other words robzilla, you have to handicap the games yourself and determine what you feel the proper odds should be. How this is done varies from capper to capper.

Once you have estimated what you feel the proper odds should be, THEN you turn to Kelly if it looks like you have an edge, to determine what bet size should be.

[robzilla]

In other words robzilla, you have to handicap the games yourself and determine what you feel the proper odds should be. How this is done varies from capper to capper.

Once you have estimated what you feel the proper odds should be, THEN you turn to Kelly if it looks like you have an edge, to determine what bet size should be.

But depending on what ur system is, the % is gonna be different for alot of people. Some people might think a 2% edge is awesome.

[pat venditto]

same ammount on every bet. no ****ing kelly or john or eugene criterion.

[MonkeyF0cker]

Rob,

Google for sports win probability models. That will get you pointed in the right direction.

[MonkeyF0cker]

But depending on what ur system is, the % is gonna be different for alot of people. Some people might think a 2% edge is awesome.

As long as your model isn't inherently flawed, any edge is awesome.

[durito]

But depending on what ur system is, the % is gonna be different for alot of people. Some people might think a 2% edge is awesome.

Here's what you do rob:

You are a 75% picker, so at -110 your edge is 43.182%, a 1/2 kelly bet will be 25.4% of your bankroll. Starting with a roll of $1,000 you should have $38,081,436,386,372,656 after just 300 bets.

[SBR Lou]

I assume one has to HAVE an edge to calculate it?

[donjuan]

kelly is a bullshit theory. its not handicapping it's a modified martingale.

Who claimed the Kelly criterion was handicapping? It's a staking strategy for advantage gamblers who are looking to grow a sum of money as quickly as possible.

[reno cool]

I think it is kind of inversely proportional to Martingale. If you overbet Kelly you will almost bust most of the time, unless you run particularly good and then youll make a fortune.

[LT Profits]

If you overbet Kelly, it probably means you overestimated your edge. If you know what you are doing handicapping-wise, it is comical to put Kelly in the same sentence as Martingdale.

[reno cool]

don't think its comical at all.

The reason everyone is so down on Martingale is because people primarily want to beat a losing game with it (which is impossible longterm)

Overbetting Kelly seems to turn a winning game into a looser, (which is also impossible )

Since those betting Kelly are primarily concerned with pos expectancy games they of course get more credit.

[max_asdf]

[MrX]

Let's put it this way:

Those familiar with Kelly know that by betting more than double your Kelly stake, you will have negative expected growth. +EV bets, but negative expected growth. Therefore:

Say you had a much larger bankroll than casino X (no affiliation to Mr. X) and casino X offered a -2% EV blackjack game with limits such that you could do the following:

Always bet an amount such that casino X had 6% of it's bankroll staked.

You are now forcing casino X to over-bet Kelly by more than 2x Kelly, therefore casino X will experience negative growth, therefore you will experience positive growth.

Now you're expecting to make money longterm at a -EV game, right?

[Sinister Cat]

Let's put it this way:

Now you're expecting to make money longterm at a -EV game, right?

An amusing scenario. I've often wondered how it would work if someone offered a +EV casino, with fairly high minimum bets, where they could somehow filter out any players with large bankrolls and any common sense; e.g., only let in people who are really drunk and don't look like they have much cash on them, or something. Might think they could still turn a profit just by busting most of their players, or at least have some expectation of doing so in the short-term?

[HedgeHog]

Last night I read Ganch's articles on Kelly criterion.

I'm not very good at math, but I think I get the basic concept.

my question is how do u calculate Win Probability/Edge?

Check out "tools". Many of us dummies are using them. As several have pointed out, the handicapping is up to you. Kelly just tells you the proper bet size if you're somewhat accurate.

[HeeeHAWWWW]

If you overbet Kelly, it probably means you overestimated your edge.

That's my favourite thing about Kelly and anything related to it: if it tells you to bet big, you know to go back and have a really damn good look at your edge calculations. Really helps with the discipline.

[Bet Shooter]

Kelly Calc

Is the Kelly calc in SBR working properly? I tried using it tonight for the first time and I got a series like this:

1 NaN%
2 NaN%
3 NaN%
4 NaN%
5 NaN%

I then tried entering the example given below the calc to see what I was doing wrong, and I received the same series. Could someone give me a hand with this? Thank You in advance.

[Ganchrow]

What browser are you using and exactly what data did you enter in each box?

If you entered the win probability did you append a '%'? If so, then that could be the issue.

[Bet Shooter]

Sorry for the late post Ganch. Was crazy today!

I am using internet explorer and this is what I entered: It's the same as the example that is posted at the bottom of the Calc page.

Independent
5
17
15000
100%

Listed for all five events:
US odds
-110
Win Prob
55.0000

Then I get this after clicking calculate Kelley:

1 NaN%
2 NaN%
3 NaN%
4 NaN%
5 NaN%

[LT Profits]

I think you mean a 15K roll at full Kelly. If so, Multiplier should be 1, and you should put 15000 in Starting Bankroll. Any number you put in that field without a "%" at the end will automatically convert to dollars.

[Ganchrow]

I think you mean a 15K roll at full Kelly. If so, Multiplier should be 1, and you should put 15000 in Starting Bankroll. Any number you put in that field without a "%" at the end will automatically convert to dollars.

I think LT nailed it. You're not really interested in a Kelly multiplier of 17, are you?

From the Calc page:

Set the event type selector to independent events, the # of events to "5", consecutive series to "17", starting bankroll to $15,000.00 (the two places after the decimal point will yield results to the nearest penny).

And then just use the default value of 1 for "Kelly Multiplier".

[Bet Shooter]

I got it LT, thank You. I had the bankroll 15K in the Multiplier spot. I will experiment with this for a few days.

[LT Profits]

Ganch,

I think 17 was his number of series to coincide with the number of weeks of NFL.

[Bet Shooter]

I think LT nailed it. You're not really interested in a Kelly multiplier of 17, are you?

From the Calc page:And just use the default value of 1 for "Kelly Multiplier".

Ok, now the hard question: Can you point me in the right direction to read about this black box I am using and why this is the optimum strategy for managing bankroll? Once I read up on this theory, I will save my questions till the end.

[Bet Shooter]

Ganch,

I think 17 was his number of series to coincide with the number of weeks of NFL.

Yes, I tried to use it on my own and it failed. Then I tried to insert the #'s in the example posted below the calc to see if I got the same results to be sure it was working. The example used 17 as the NFL regular season weeks.

[Bet Shooter]

I see the link to A quantitative introduction to the Kelly Criterion. Thanks. Let me digest this.