What is the correct Kelly formula?

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • kokky
    SBR Hustler
    • 04-29-09
    • 63

    #1
    What is the correct Kelly formula?
    I see many versions of Kelly formula on the internet. Some of them say i should bet 10 or 20% of my bankroll, that seems too much. Others never give percentage over 1%, that seems too little.

    In some formulas there is some Kelly divider, what is that?

    So what is true, or best Kelly formula?
  • FourLengthsClear
    SBR MVP
    • 12-29-10
    • 3808

    #2
    Read the notes below this calculator.
    Sports betting and handicapping forum: discuss picks, odds, and predictions for upcoming games and results on latest bets.
    Comment
    • LT Profits
      SBR Aristocracy
      • 10-27-06
      • 90963

      #3
      Or even more simplified, bet to win whatever your edge is, i.e., if you cap a game as having a 3% edge, bet to win 3% of bankroll.
      Comment
      • splash
        SBR Rookie
        • 05-25-09
        • 38

        #4
        If kelly says bet 20%, then that is the optimal amount. Just means your edge must be pretty huge. Don't doubt the equation just cause it tells you to bet a large percentage unless you think that an edge that large isn't possible.
        Comment
        • goblue12
          SBR MVP
          • 02-08-09
          • 1316

          #5
          Betting half Kelly cuts your "Risk of Ruin" from around 13-14% to 1-2%.
          Comment
          • splash
            SBR Rookie
            • 05-25-09
            • 38

            #6
            Originally posted by goblue12
            Betting half Kelly cuts your "Risk of Ruin" from around 13-14% to 1-2%.
            Uh no. Your risk of ruin betting full kelly is 0%. This is not factoring in that when your roll gets so low kelly tells you to bet less than one penny.
            Comment
            • FourLengthsClear
              SBR MVP
              • 12-29-10
              • 3808

              #7
              Originally posted by splash
              Uh no. Your risk of ruin betting full kelly is 0%. This is not factoring in that when your roll gets so low kelly tells you to bet less than one penny.
              If I was to lose 95% of my bankroll, I would consider that as "ruin".
              Comment
              • sharpcat
                Restricted User
                • 12-19-09
                • 4516

                #8
                Originally posted by kokky
                I see many versions of Kelly formula on the internet. Some of them say i should bet 10 or 20% of my bankroll, that seems too much. Others never give percentage over 1%, that seems too little.

                In some formulas there is some Kelly divider, what is that?

                So what is true, or best Kelly formula?
                You should stay far away from kelly until you have a much better understanding of it.
                Comment
                • mjespoz
                  SBR Rookie
                  • 02-15-11
                  • 42

                  #9
                  If you need to ask these sorts of questions then you shouldn't even think of using Kelly.
                  Comment
                  • slickeddie
                    SBR Sharp
                    • 01-03-09
                    • 365

                    #10
                    Originally posted by FourLengthsClear
                    If I was to lose 95% of my bankroll, I would consider that as "ruin".
                    LMAO.

                    Not sure I understand the others telling him if he doesn't understand then he shouldn't be using it tho. I'm sure he is asking because he wants to learn more about it.

                    Wondering myself if I have a baseball RL system that has a record of 797-752, and the average odds were +113, is the Kelly a good thing, and if so, what would my "edge" be?

                    Thanks for any help.
                    Comment
                    • FourLengthsClear
                      SBR MVP
                      • 12-29-10
                      • 3808

                      #11
                      Originally posted by slickeddie
                      LMAO.

                      Not sure I understand the others telling him if he doesn't understand then he shouldn't be using it tho. I'm sure he is asking because he wants to learn more about it.

                      Wondering myself if I have a baseball RL system that has a record of 797-752, and the average odds were +113, is the Kelly a good thing, and if so, what would my "edge" be?

                      Thanks for any help.
                      In simple terms, yes. Kelly gives the optimal return compared to the risk of going bust.

                      That risk of going bust when utilising Kelly is not insignificant and is why some people opt for a fractional approach which still produces better results than flat staking.

                      In terms of the baseball record, the average edge per game is:

                      51.45% (winning percentage)
                      divided by
                      46.95% (implied winning probability of +113)

                      = 1.0959
                      Edge was therefore 9.59%

                      The next thing you need to establish is the wisdom (or otherwise) of using that going forward!
                      Comment
                      • slickeddie
                        SBR Sharp
                        • 01-03-09
                        • 365

                        #12
                        Thanks 4Lengths. How do you come up with the 46.95%? I have other systems and knowing which one has the biggest edge, will help me a ton.

                        As far as "wisdom" is concerned, I'm shiot outta luck on that one, lol.
                        Comment
                        • FourLengthsClear
                          SBR MVP
                          • 12-29-10
                          • 3808

                          #13
                          Originally posted by slickeddie
                          Thanks 4Lengths. How do you come up with the 46.95%? I have other systems and knowing which one has the biggest edge, will help me a ton.

                          As far as "wisdom" is concerned, I'm shiot outta luck on that one, lol.
                          Convert US style odds to the implied win probability as follows:

                          + odds e.g. +113
                          100 divided by (odds + 100)
                          So +113 is
                          100/213 = 0.4695 or 46.95%

                          - odds (e.g. -113)
                          odds divided by (odds + 100)
                          So -113 is
                          113/213 = 0.5305 or 53.05%
                          Comment
                          • slickeddie
                            SBR Sharp
                            • 01-03-09
                            • 365

                            #14
                            Great Thanks.
                            Comment
                            • slickeddie
                              SBR Sharp
                              • 01-03-09
                              • 365

                              #15
                              One more question if you don't mind. Can we use the "average odds" to get a reasonable edge number?

                              For example, if I have a RL system that is 1108-843, with the average odds of +104, I have an edge of 15.87 (I think, based on the formula given), but a few of the odds may have been -120 and some could have been +160, etc. Will this variance in odds make a difference in my edge result?
                              Comment
                              • FourLengthsClear
                                SBR MVP
                                • 12-29-10
                                • 3808

                                #16
                                Originally posted by slickeddie
                                One more question if you don't mind. Can we use the "average odds" to get a reasonable edge number?

                                For example, if I have a RL system that is 1108-843, with the average odds of +104, I have an edge of 15.87 (I think, based on the formula given), but a few of the odds may have been -120 and some could have been +160, etc. Will this variance in odds make a difference in my edge result?
                                With Kelly the idea is to bet based your edge for a particular matchup and as such using a historical average of any description is not optimal.

                                With that said, what I would do is to break the overall sample into smaller groups (subsets).
                                Lets say you subdivide in to odds of +150 and higher, from +120 to +149, etc.

                                Also split up home teams and away teams (very important in baseball RL betting).

                                More divisions should be made for anything that is important to how you make your picks.

                                Of course each of these subsets represents a smaller sample and is therefore more prone to variance but it should nonetheless help you identify where you have been extracting the greatest value.

                                Some in here would object but I would be comfortable utilising half Kelly on that subsetted and constantly updated data. The more data you have the more you can have confidence in it and the closer you should be to full Kelly.
                                Comment
                                • splash
                                  SBR Rookie
                                  • 05-25-09
                                  • 38

                                  #17
                                  Originally posted by FourLengthsClear
                                  If I was to lose 95% of my bankroll, I would consider that as "ruin".
                                  Which is fine but the guy I quoted gave actual percentages. It doesn't make any sense to say RoR is 13% at full kelly and only 1% at half kelly without defining what parameters and definitions you are using. Since he didn't define "ruin", the absolute sense of the word is assumed.
                                  Comment
                                  • Wrecktangle
                                    SBR MVP
                                    • 03-01-09
                                    • 1524

                                    #18
                                    A lemming and his money are soon parted.
                                    Comment
                                    • WendysRox
                                      SBR High Roller
                                      • 07-22-10
                                      • 184

                                      #19
                                      I made a little excel spreadsheet where you input the odds and your win percentage. It give you everything else.

                                      Here: http://www.4shared.com/document/SfYF...el_2003_V.html
                                      Comment
                                      SBR Contests
                                      Collapse
                                      Top-Rated US Sportsbooks
                                      Collapse
                                      Working...