| Money management strategies help the punter to | | | | wager by the corresponding ratio. For example, if the |
| choose an optimal stake, and even though Kelly is | | | | accuracy ratio of the betting tips for a home win is |
| known to be the best strategy of them all, it is not | | | | 0.55, then when one bets on a home win, he should |
| necessarily true for all types of betting tips. | | | | multiply his wager by 0.55 to lower the risk. |
| This article summarizes possible ways to improve | | | | If one's betting tips don't include a probability of a win, |
| Kelly's strategy for different types of betting tips. | | | | he should start with estimating this probability. One |
| First, let's emphasize Kelly's strengths and | | | | simple way to do that is by calculating the ratio of |
| weaknesses: | | | | home wins, draws and away wins to the total number |
| According to Kelly's, the wager for each outcome is | | | | of games in the season. These proportions can be |
| calculated in the following way: w * (p-(1-p)/(odds-1)), | | | | used as probabilities of a win and no other correction |
| where w represents the wealth of the punter and p is | | | | factor will be necessary. This method is less accurate |
| the probability of a win. For example, if your wealth = | | | | then the previous one, since it uses average statistics |
| 1000$, p = 50% and odds = 2.5, then the wager your | | | | to estimate the probability of a win. |
| have to place is equal to 1000$*(0.5 - 0.5/(2.5-1))=166$. | | | | The methods described above help punters to |
| If the probability of a win (p) is estimated correctly, then | | | | calculate their optimal wager. However, these methods |
| the calculated wager will be optimal in a long run. If, | | | | won't help you choose the outcome to bet on. For |
| however, it is wrongly estimated, then using Kelly's | | | | example, when you distribute your wager between |
| won't do the trick. | | | | two of three possible outcomes, the long run profit will |
| What are the possible ways to improve Kelly's | | | | be maximal. The only way to calculate the optimal |
| strategy when the probability of a win is incorrect or | | | | wager for each outcome is by using data-driven |
| unknown? | | | | algorithms. In this case, the wager should be distributed |
| When one's betting tips include a probability of a win, it | | | | between matches whose outcomes can be estimated |
| is necessary to determine how accurate this | | | | automatically given historical data only. The output of |
| probability is for home/draw/away wins. It can be done | | | | this algorithm is the value of the wager for each |
| by ratio of several successive tips to the total number | | | | outcome. The optimization criterion for those algorithms |
| of tips provided. This ratio can be used as a correction | | | | is, of course, a maximal profit when betting odds are |
| factor for a wager estimated according to Kelly's. To | | | | known. |
| use this correction factor it is enough to multiply the | | | | |