Quote Originally Posted by beaverfan23:
Nice comeback win for Usf tonight!
Also, good thoughts by everyone in this thread. I am going to continue just keeping track of the win/loss ytd records, but if someone wants to implement the Kelly system and keep track of +/- units that would be fantastic. This is my first full season keeping track of this system so I am learning as I go and hopefully fine tuning the system for future seasons. I have heard of the Kelly units system, but never used it; is it true that it is a % of current bankroll so that amount placed on each game changes as the bankroll grows and shrinks?
John Kelley was an engineer at Ma Bell in the early 1950s and was given the mandate to write a report on the efficiency of power distribution along transmission lines.
Senator Estes Kefauver (who later became a losing VP candidate) was conducting hearings a few months earlier into organized crime's involvement in sports gambling.
Kelly may have taken his cue from that
Regardless, Kelly was tasked with explaining to his superiors a highly-technical subject... so he decided to analogize.
He wrote his report on transmission-line inefficiencies by setting up "a [horse-racing] wire" like in the Robert Redford/Paul Newman movie, "The Sting" where past-posting the field was alleged/promised to gangster, Robert Shaw by con-man, Harold Gould
When gamblers read the paper they realized that it quite correctly applied to wagering on events where some measure of positive expectation existed and could be quantified.
For even-money wagers like football, the formula is simple.. the 11/10 translates to .91
-105 betting (21/20) translates to .96.
If someone were somehow magically pre-aware that after 304 wagering events that one would end up with 54.9% winners then that individual at -110 would have wagered 54.9 - (45.1/.91) = 5.3% of one's bankroll on every wagering event.
As far as KEEPING TRACK of Kelly Bets therefore.... you are doing that already with your overall percentage...
...what is being done in the prior posts is to estimate how far from that centerpoint of 54.9% the true advantage lies and with what level of certitude we can rely on that "true advantage" figure.
If you are a 58.3% handicapper then one should bet over 12% on each play [ 58.3 - (41.7/.91) = 12.4%.
trabs9 talked about not leaving money "on the table" which means getting into every bet that carries some positive expectation.
Kelly carries with it extreme volatility which can cause an upset stomach.
One approach might be to divide one's bankroll into 2 portions... assume SOME "reversion to the mean" since there are very-knowledgable people who'd swear on their Calculus textbooks that no "power rating system" could pick better than 54%, if that.
but what the heck...we should not be betting money that we cannot afford to lose...
...so use 1/2 of your bankroll to bet those 59% plays at an assumjed advantage of 56%...
...and then bet 2% on the other plays with the other 1/2 of your bankroll.
The Kelly advantage is described by an upside-down parabola where the advantage curve sort of flattens out at the peak... so if you're at 56% your collecting most of that advantage and providing a nice measure of safety
Hot Forum Topics
