# MAIN FORUM

• GLAlexander

Kelly Criterion Calculator Questions

After reading numerous bankroll articles, I am wanting to incorporate Kelly Criterion into my daily play. I have found several calculators on the web. I am a little confused on how to use them. Here is the information that I have.

Bankroll is \$1200 and I play the \$5 Double Ups and Triple Ups, Shot and Rebound entries. I have my cash percentages for all four types of games. How would I input these numbers into the calculator using FD payout structure. I only play the large field +200 in my double and triple ups.

Once I do that would I have four separate amounts of \$ I should play in each game that day?

My current philosophy was playing roughly 10% of my bankroll daily with a mix of roughly 60% cash games and 40% GPP. Having read so much about Kelly, I wanted to give that a try.

• Tippy

The most useful, and hard to obtain, piece of information for using a Kelly is your winning percentage. If you don’t have the right winning percentage to plug into the formula, then the whole thing is pretty much useless.

You will probably have to split your Kelly into four different pieces for H2H, 50/50, Triple up and GPP. Each of these would be considered a different type of bet with different payout odds AND different winning percentage figures.

Probably the easiest place to start is H2H games since it is a lot closer to an actual sports bet, which is what Kelly is useful for. To calculate your bet size you need the payout, which is bet 5 to win 4 on Fanduel (not considering FDP or bonus). The second piece of information you need is your winning percentage. What do your records say your percentage is over a large sample, say 1000 games? If you have won 600 and lost 400, then your winning percentage is 60%, so plug that number into your Kelly formula. The minimum bet on FanDuel is \$1. Use this calculator and simply plug your numbers in and it will give you the bet size. http://www.albionresearch.com/kelly/default.php

Do the same calculation for 50/50. Then for triple up. Then for GPP. Once your get to GPP you’ll see the payout odds are very large, while the winning percentage is very small. I think calculating each one separately and then adding them together should give you a good estimate of your bet size for the day. And remember, if you are using a single roster, all of your bets are highly correlated.

• GLAlexander

Tippy

Thanks, I inputted the numbers and got this, does it look correct to you?

Gambling Bankroll: \$1200
Odds offered: 2.23 to 1 (based on 100 cashing in 223 man double up)

Probability of Winning: 58% Use a conservative (low) estimate.
Bets must be multiples of: 5 \$
The minimum bet allowed is: 5 \$

Results
On 58% of similar occasions, you would expect to gain \$1,036.95 in addition to your stake of \$465.00 being returned.
But on those occasions when you lose, you will lose your stake of \$465.00.
Your fortune will grow, on average, by about 15.53% on each bet.
Bets have been rounded down to the nearest multiple of \$5.00.
If you do not bet exactly \$465.00, you should bet less than \$465.00.
The outcome of this bet is assumed to have no relationship to any other bet you make.
The Kelly criterion is maximally aggressive â€” it seeks to increase capital at the maximum rate possible. Professional gamblers typically take a less aggressive approach, and generally won’t bet more than about 2.5% of their bankroll on any wager. In this case that would be \$30.00.
A common strategy (see discussion below) is to wager half the Kelly amount, which in this case would be \$230.00.
If your estimated probability of 58% is too high, you will bet too much and lose over time. Make sure you are using a conservative (low) estimate.

• agendaman

/

the kelly criterion can be highly volatile meaning it is for long range returns there can be bad losing streaks i.e. variance as we see here almost daily

• Tippy

@GLAlexander, the payout odds offered isn’t right. In the double up you are still betting \$5 to win \$4 (entered as 4 to 5 in the formula).

I’m not real familiar with that kind of large field double up, but you seem to be playing a game where 223 people enter and only 100 get paid. That is only about 45% of the field getting paid. I think the payout odds might be different than 5:4? You are saying that you are getting into the top 45% about 58% of the time? That’s pretty damn good. That would probably translate into 63-65% winning percentage in a doubleup that paid 50% of the field. Like I said, the winning percentage is the hardest part of the formula to get right.

In the Kelly formula, you use the payout odds of the bet. For example, if you applied the Kelly to a football bet you would use 11:10 (10 to 11) in the formula. The winning percentage of your formula is where you estimate how likely you are to actually win (and this is where you work in the 2.23 type payout numbers of the double up).

Edit: is the payout in your double up pay \$5 and you win \$5 (rather than the pay \$5 to win \$4 i suggested)? The problem here is you have to calculate what the exact rake is in this type of game so you can enter the proper payout odds. The rake amount is kind of hidden when the game is set up like this and kind of gets shifted from the payout odds portion of the formula to the winning percentage portion of the formula. I think the rake in this game is probably a little worse than the standard bet \$5 to win \$4. A payout of \$5 to win \$5 combined with a 58% winning percentage would yield about 17% of your roll being bet every night, which seems high to me.

You could actually have a positive profit if you won this particular game with just a 51% winning percentage, whereas you would be losing money with that 51% winning percentage in a regular 50/50 with a bet \$5 to win \$4 setup with exactly 50% of the field getting paid.

• GLAlexander

Tippy

Thanks for the feedback, the 58% is based on when the NBA started in Nov. You are correct that is top 100 in a 223 person field wager 5 win 10.

How would I calculate this for GPP? The shot is a \$2 entry and you win in incremental scales of \$6 and going up?

I also thought the amount of bankroll played each night seemed really high.

• dailyleaguer

@Tippy said…

The most useful, and hard to obtain, piece of information for using a Kelly is your winning percentage. If you donâ€™t have the right winning percentage to plug into the formula, then the whole thing is pretty much useless.

LOL. You mean, if you don’t have a profitable winning percentage?

• Tippy

@GLAlexander said…

Tippy

Thanks for the feedback, the 58% is based on when the NBA started in Nov. You are correct that is top 100 in a 223 person field wager 5 win 10.

How would I calculate this for GPP? The shot is a \$2 entry and you win in incremental scales of \$6 and going up?

I also thought the amount of bankroll played each night seemed really high.

I’m not sure on a GPP. The payout structure is variable depending on where you finish. The winning percentage is also pretty hard to determine because it’s hard to get a large enough sample. I’m pretty sure it would take some serious math to get the payout odds and winning percentage. However, you could try to run a few estimates just plugging in some numbers into the formula just to get a ballpark figure. I’d guess since the winning percentage is so small in GPP’s that the bet size would also be pretty small. Seems like determining your GPP size bet is more of an exercise in how much you can comfortably lose without it affecting your other game entries/bets. Sorry I’m not more help on the GPP calculation.

@dailyleaguer, LOL, yeah that helps for sure.

• GLAlexander

Tippy

Thanks for your help, many more nights like tonight, and I won’t need a Kelly calculator…lol