Just wanted to get some opinions regarding a sound money management system. What % of your bankroll should comprise a standard bet? Assuming a standard bet is 1 unit, what is the highest amont of units that should be wagered on one event, even the POD/poy type events. Thanks and I look forward to hearing your responses. :think2:
Kelly System
My question with regard to the Kelly system, is at what level does a competent capper deem the probability of an outcome to be significantly different than a given line? In other words, what probability must a certain outcome have for a good capper to deem a play to be worthy of a pick. Given this number, you can easily calculate what % of bankroll should be wagered on that pick. Is it 60%, 65%?
My question with regard to the Kelly system, is at what level does a competent capper deem the probability of an outcome to be significantly different than a given line? In other words, what probability must a certain outcome have for a good capper to deem a play to be worthy of a pick. Given this number, you can easily calculate what % of bankroll should be wagered on that pick. Is it 60%, 65%?
Your question is a good one. According to the strict mathematical principles of the Kelly Criterion, the expected value of the bet must only yield a positive return in order to be good enough to bet on. So, to beat a -110 line, a pick must have about a 52.4% (or greater, obviously) chance of winning. The Kelly System then allows you to determine the optimal bet size based on the predicted likelihood of the bet winning.
In theory, a capper should play all bets with +EV for their Kelly amounts to maximize ROI. Of course, it is difficult to predict the exact likelihood of a play winning. Also, it is nearly impossible to evaluate all plays on a given card with enough scrutiny to begin to get so in-depth with each game.
It's pretty much a waste to even try to determine the exact percentage of any one play. Better is to keep track over the long term and determine what percentage you hit. If you hit at 53% in the long term, you should bet what the Kelly Criterion tells you to bet for that percentage, but if you hit 55% you should be betting considerably more, based on Kelly. Don't bother trying to figure out a percentage for every single play, in the long term it will all even out.
kelly is a sure route to disaster for anyone [experienced or not] whom is unable to accurately asses their expected win %.
that is the key element to this system.
if you were, somehow, able to determine your edge accurately, you would be winning vast amounts of $$ regardless the system of wagering employed.
if you overestimate your edge and use kelly you will go broke fast.
that is the key element to this system.
if you were, somehow, able to determine your edge accurately, you would be winning vast amounts of $$ regardless the system of wagering employed.
if you overestimate your edge and use kelly you will go broke fast.
A good capper that is confident on their valuation model will bet a proportion of their bankroll off of any disparity in their fair value and the market price. A lot of good cappers that implement Kelly (knowing that overbeting predicated on overvaluation of edge) will implement a margin of safety, confidence intervals, or implement fractional Kelly, as overbetting due to edge will significantly increase risk of ruin.
A lot of people shun Kelly on forums. Most for the wrong reason. If one has logrithmetic utility function, it makes mathematical sense to implement a Kelly stake betting system (and can accurately quantify their edge). The right reason to discourage those to implementing Kelly is because they can't accurately quantify their edge to the level in which one can benefit from Kelly over another staking system(someone would argue not to bet at all if they can't accurately quanitify their edge).
Some good methods for one to analyze the accuracy of their model and percieved edge is to see how often they beat the closing number, and to measure the level of convergence or divergence of the market price (from open to close) relative to ones fair value.
Last edited:
