Quote:djatcThis doesn't seem much of an edge, considering she would only check with a ten. Maybe if she fudged up tucking the second card underneath and you could see paint or pips it might have been good.

Without taking the time to do any math, I would guess that the overall edge is around 2%. I'm not sure exactly what it is because I'm not sure how much the hole card edge is biased towards different upcards (ie, you can't just multiply it by 5/13). Maybe teliot will write a blog post about it.

You can reduce swings a bit by counting as well as hole-carding -- a TC of -4 will wipe out a 2% edge, which is not that uncommon of a situation.

I consider this a healthy edge, but you don't want to overbet your bankroll, of course. The Kelly Criterion always applies.

Quote:AxiomOfChoice

I consider this a healthy edge, but you don't want to overbet your bankroll, of course. The Kelly Criterion always applies.

Overbetting is ok if the edge is huge and the hit frequency is pretty high. For example if you were to have your machine "salted" after a taxable playing a $100 video poker machine would make a lot of sense since you'd only need to hit 3oak to lock it up on a non wild game. I don't know the math off the top of my head but adding 5 units to every pay except 2 pair and high pair would be huge.

EDIT: 7/5 jacks would become 8/6 jacks with a healthy 7.3% edge.

Quote:djatcOverbetting is ok if the edge is huge and the hit frequency is pretty high. For example if you were to have your machine "salted" after a taxable playing a $100 video poker machine would make a lot of sense since you'd only need to hit 3oak to lock it up on a non wild game. I don't know the math off the top of my head but adding 5 units to every pay except 2 pair and high pair would be huge.

EDIT: 7/5 jacks would become 8/6 jacks with a healthy 7.3% edge.

Overbetting is never ok. If the edge is huge and the hit frequency is high, then the amount that you can bet goes up. But that is not overbetting. The Kelly Criterion always applies to find out how much you should bet.

Betting more than Kelly is ok if it's necessary to play -- although betting exactly Kelly is better, betting (for example) 1.5x Kelly is better than not playing if those are your only 2 choices. Betting more than 2x Kelly leads to disaster, though -- it's better not to play.

I've never heard this before. Can you say more?Quote:AxiomOfChoicebetting (for example) 1.5x Kelly is better than not playing if those are your only 2 choices. Betting more than 2x Kelly leads to disaster, though -- it's better not to play.

Quote:teliotI've never heard this before. Can you say more?

It's well-known that bankroll growth is optimized by using the Kelly fraction to size your bets.

My understanding was that bankroll growth was also symmetrical around this point, ie, for a given bankroll B, if your optimal bet is f * B (f is the Kelly fraction) then it was also true that betting (f - x) * B had the same long-term expectation as betting (f + x) * B. So betting 2f * B has 0 long-term growth, and betting > 2f * B has negative long-term growth, ie, 100% RoR.

Didn't Wong publish this? I thought that he did but I can't find it. Am I wrong here? Should be easy to verify.

http://www.sbrforum.com/betting-tools/kelly-calculator/

Quote:teliotI've never heard this before. Can you say more?

Just calculate CEV (certainty equivalent value) which is CEV = exp(<log(B+X)>)-B, where B is your bankroll, X is the random variable outcome of your bet to make, and <.> is the expectation value.

CEV is (over the full positive bet range) maximal for a Kelly criterion bet size. However if that bet size is not available, one should maximize CEV over the available left range of bets.

CEV turns negative for bets roughly twice the Kelly bet. In the extreme case, where only twice a Kelly bet (or larger) or not playing is an option, you would not want to play as CEV=0 for not playing. If 1.5 x Kelly bet is available but Kelly bet (or less) isn't, you should still play it. It is not "overbetting" in the technical sense, because you still maximize CEV.