zapakh
• Posts: 3
Joined: Feb 23, 2010
February 23rd, 2010 at 2:02:51 PM permalink
I've got an optimal play question on a particular hand I faced on the Wizard's applet. I was betting five credits on the default payschedule, so that the payoffs would be 5-10-15-20-30-45-125-250-4000.

Here's the hand:

As 2c Th 3s Ks

I kept the suited AK, but this was recorded as an error. Seems I should have kept the same-suited trey as well. Here is the error entry:

HAND BEST PLAY EV PLAY MADE EV ERROR VALUE
As 2c Th 3s Ks A--3K 2.1869 A---K 2.1341 0.0528

My first thought was that the Th was the culprit. Discarding it reduces the probability of drawing to a straight. But that probability was pretty slim to begin with, and now I'm not so sure. I can't find mention in the wizard's optimal JoB strategy of drawing at 3-to-a-flush, ever.

Of course that error value amounts to 1% of a wager, on a situation that probably doesn't happen very often. But, out of academic curiosity--and before I let that curiosity drive me through a full analysis--does anyone know what's going on here?
teddys
• Posts: 5529
Joined: Nov 14, 2009
February 23rd, 2010 at 5:12:15 PM permalink
That's odd. Must be a glitch. The correct play is to hold the A-Ks. I've never gotten an error on that before.
"Dice, verily, are armed with goads and driving-hooks, deceiving and tormenting, causing grievous woe." -Rig Veda 10.34.4
scotty81
• Posts: 185
Joined: Feb 4, 2010
February 23rd, 2010 at 6:37:49 PM permalink
I had asked the Wizard for a one-coin JorB strategy (which he graciously provided in an earlier thread), and the holding of the third flush card is a legitimate play in that strategy. In order:

...
4 to a Straight (JQKA)
2 to a Royal Flush (JQ, JK, QK) *
3 to a Flush with 2 high cards [this is the case we are talking about]
2 to a Royal Flush (JA, QA, KA)
....

* 3 to a Flush with 2 high cards beats Suited JK or QK if the cards include a Jack, Queen & King

I had thought this was just a special one-coin strategy.

Either this strategy also applies to full-pay 5 coin play, or the Java Script game is determining its strategy based upon the 250 RF payout, not the 800 RF payout.

Only the Wizard knows for sure...
Prediction is very difficult, especially about the future. - Niels Bohr
zapakh
• Posts: 3
Joined: Feb 23, 2010
February 23rd, 2010 at 8:58:42 PM permalink
Thanks for this great find. A royal payout discrepancy between 4000 and 1250 (because I'm betting 5) times the 1-in-16215 royal probability translates to a change in EV of 0.1696, which is bigger than the error value I got.

However, I don't think this is what's going on. I duplicated the program's analysis sheet (on a different hand, since sadly I didn't have the foresight to capture one for the hand in question) and found that the EVs given are valuing the royal at 4000 when I bet 5.

I tried to go one better and calculate the 2.1341 EV that the error report gave me but, working by hand, I can't seem to arrive at the same figure. My calculation yields 2.8403 instead:

A---K nothing JoB 2Pr 3oK Str Fl FH 4oK StrFl Royal TOT
DRAWS 10012 5022 711 281 48 120 18 2 0 1 16215
PAYRATE 0 1 2 3 4 6 9 25 50 800 -
EXPECTED PAYRATE 0.00000 0.30971 0.08770 0.05199 0.01184 0.04440 0.00999 0.00308 0.00000 0.04934 0.56805
EV(on betting 5) 0.0000 1.5486 0.4385 0.2599 0.0592 0.2220 0.0500 0.0154 0.0000 0.2467 2.8403

What I really want is to see the app's analysis sheet on As 2c Th 3s Ks, but it's going to be a long time before that exact hand comes round again.
ZPP
• Posts: 31
Joined: Feb 7, 2010
February 23rd, 2010 at 11:49:10 PM permalink
zapakh, I get the same numbers as you except I get 47 straights and 119 flushes (and 10014 garbage). I believe you are counting the royal flush 3 times. However, the effect on expected value is only 0.0031, so that doesn't help much.

Edit: You were betting 4 coins. If you apply my correction, and also use 4-coin payouts (including 250*4 for a royal), you get an EV of 2.1341. Also, the 4-coin EV of holding AK3 is 2.1869.
zapakh