Right.Quote:AxelWolfIt means I change my mind and made a temporary exception and I reserve the right to do so in the future since either way it's not a big deal. I also don't have the willingness to do crack. I'll stick to my willingness when it comes to the things that really count.

Only fools never change their minds.

Balance: $33.84

Running Total: -.08

Sport: MLB

Game: Texas @ Houston 9/17/19

Pick: Astros-297

Wager: $1.61

Quote:es330tdOkay, so you say I have roughly a 1/3rd chance of doubling my bankroll. What happens if I do it three times? Am I just lucky hitting a 1/27 chance?

That is what I'd say. Same thing I'd say to a craps player who had a system of combining various small negative EV bets. Remember, if your selection system does pick bets that are positive EV, then I'd change my tune and not be surprised that you'd be a winner. If your picks are positive EV, you should use a Kelly system for determining your bet size, not a chase a loss bet size.

And my 1/3 is only a guess approximation by me. If it is 2/5, then you have a 8/125 (1/16 ish) chance to succeed 3 times in a row.

And all the above is predicated on betting lines that you could really get if you were betting real money.

Edit. I just checked. If you shopped, you could actually do a little better than -297. So I will assume you are fairly posting the odds you are placing your wagers at.

That is an unfounded assertion.Quote:SOOPOOIf your picks are positive EV, you should use a Kelly system for determining your bet size, not a chase a loss bet size.

Please elaborate and prove.

(In decision theory, there is no ‘should’. Just a lottery, a prob distribution of results and individual preferences or objectives.)

(Kelly formula is notoriously not applicable to sports betting.)

Quote:kubikulannThat is an unfounded assertion.

Please elaborate and prove.

(In decision theory, there is no ‘should’. Just a lottery, a prob distribution of results and individual preferences or objectives.)

(Kelly formula is notoriously not applicable to sports betting.)

I'm no expert on Kelly, nor sports betting for that matter. But let's say you found a reproducible pattern in sports betting (road teams traveling less than 200 miles against home teams in cold weather cities after a loss by less than two touchdowns win 60% of the time versus the spread) Wouldn't Kelly help you find the best bet size to grow your bankroll while minimizing risk of ruin?

- You might have a globally +EV pattern, but to apply Kelly, each bet must be +EV. (I’m not certain, but I think the Kelly development requires that the probabilities be objectively measured. A subjective -Bayesian type- probability does not ensure positive EV.)
- You must know what the relative EV is, in order to compute the optimal bet. Sports betting rarely gives you that knowledge.
- Teams’ relative efficiency can be influenced by what happened in previous rounds (confidence of the team, changing the trainer).
- And the offered odds depend on the gamblers actions.

All of this prevents the derivation of Kelly’s criterion.

===

On the other hand, I don’t see why recouping losses is a bad strategy, if EV is positive?

Quote:SOOPOOI'm no expert on Kelly, nor sports betting for that matter. But let's say you found a reproducible pattern in sports betting (road teams traveling less than 200 miles against home teams in cold weather cities after a loss by less than two touchdowns win 60% of the time versus the spread)

this is the kind of thing that sports betting sites are always putting out - often with only a hundred or a few hundred results

when blackjack pros test out a strategy against a new rule they typically run a sim of one billion hands

IMHO this kind of pattern cannot be counted on to produce similar results in the future at least when there is a relatively small sample size

I'm aware that you only put this out as an example and I'm not criticizing your post

Let us assume that es330td finds a reproducible situation, where he is certain of the true odds. Let the offered odds be 1 to 2.80, while his trusty info is that the real odds are 1 to 3.Quote:SOOPOOI'm no expert on Kelly, nor sports betting for that matter. But let's say you found a reproducible pattern in sports betting.(…) Wouldn't Kelly help you find the best bet size to grow your bankroll while minimizing risk of ruin?

With a bankroll B and a bet xB, the outcomes G(x) are (B+xB/2.8) and (B-xB).

G’(x)/G(x) are (B/2.8) / (B+xB/2.8) = 1/(2.8+x) and (-B) / (B-xB) = -1/(1-x) respectively,

and the expectation equals 0.75/(2.8+x) - 0.25/(1-x) = (*0.05-x)/(2.8+x)(1-x)

Setting that to zero gives x* = 0.05. With a bankroll of ~34$ this represents a Kelly bet of 1.7$ , whatever the previous win/loss pattern.

BUT

Kelly assumes a priori that x is constant. His formula says: « IF you play a constant portion of your capital, then this is the optimal portion, in terms of long-term growth. »

It does not say that playing a constant portion is optimal in any sense.

So I don’t see what rejects martingale-like strategies when EV is positive, and even less if the objective is short-term gains.

I do not understand why we must assume X is constant. As X fluctuates from bet to bet then the Kelly bet sizing fluctuates. Recall that Kelly’s original analogy for his algorithm was horse racing when the better knows the true odds are different than the posted odds. I must be missing the point you are making.Quote:kubikulannLet us assume that es330td finds a reproducible situation, where he is certain of the true odds. Let the offered odds be 1 to 2.80, while his trusty info is that the real odds are 1 to 3.

With a bankroll B and a bet xB, the outcomes G(x) are (B+xB/2.8) and (B-xB).

G’(x)/G(x) are (B/2.8) / (B+xB/2.8) = 1/(2.8+x) and (-B) / (B-xB) = -1/(1-x) respectively,

and the expectation equals 0.75/(2.8+x) - 0.25/(1-x) = (*0.05-x)/(2.8+x)(1-x)

Setting that to zero gives x* = 0.05. With a bankroll of ~34$ this represents a Kelly bet of 1.7$ , whatever the previous win/loss pattern.

BUT

Kelly assumes a priori that x is constant. His formula says: « IF you play a constant portion of your capital, then this is the optimal portion, in terms of long-term growth. »

It does not say that playing a constant portion is optimal in any sense.

So I don’t see what rejects martingale-like strategies when EV is positive, and even less if the objective is short-term gains.

On the other hand, the risk of a martingale strategy is what it means for the bettor’s risk of ruin.