
https://worldofcardgames.com/#!replayer ... %3A1%7D%5D
Unread post by LeftyK » Fri Jan 21, 2022 6:14 pm
Unread post by Wes (aka the legend) » Sun Jan 23, 2022 2:45 am
S1 played this hand poorly. He can't pass-pass with that holding. Holding 3 trump in the first round with no where to go in the 2nd round, S1 must grit his teeth and make that call. We have lots of good data on this spot showing that calling in the first round has a higher expected outcome than pass-pass or pass-call next with nothing. I've done some work on this spot and so has Ray. This spot is no longer debatable. S1 has to make that marginal 1st rd call. Yes that play probably has a negative expected outcome but passing has a worse negative expected outcome. This is like spliting 88s vs a T at blackjack. It's a losing play but it loses less than hitting or standing pat on your 16, so you have to do it.
Unread post by raydog » Sun Jan 23, 2022 12:17 pm
Unread post by Wes (aka the legend) » Sun Jan 23, 2022 12:52 pm
Yep. This is a leak that needs to be cleaned up. People just need to get comfortable with +EV negative reinforcement strategies, I.E. strategies that fail most of the time but are nevertheless correct. Humans are pretty much genetically designed to suck at those strategies but with study and practice one can reprogram oneself. This concept is even more important in poker. If one can take their game to that level where they master the negative reinforcement strategies they will crush almost any game.raydog wrote: ↑Sun Jan 23, 2022 12:17 pmResponding to Wes's response:
1) The first thing I also noticed is that S1 should not have passed, R1. I ran a simulation which bore this out: not a great result when calling trump, but still better than when passing [another way of thinking about this: when S1 passes R1, with no reasonable R2 bid, they are giving the opponents 4 chances to bid [twice with the given trump suit, twice with a trump suit of their choice] while only giving their own team 2 opportunities to bid - and the R1 bid from 3rd seat is particularly unfavorable and unlikely. This is defensive play.]
Just wanted to point out that the idea behind trumping in and leading the turned down suit on 2nd street is the fact that that's the suit S2's partner is most likely void in. I'm glad this idea is not "wrong" but I am surprised it's not more "right" as we basically have a statistical tie between that idea and leading the 9H on 2nd street.raydog wrote: ↑Sun Jan 23, 2022 12:17 pm2) I then looked at situations where S2 gets a chance to bid R2 with the given hand and turn card (15,775 times out of 100,000). And, further, where S1 leads a diamond. I tested 3 possible plays by S2:
play 9H: 864 / 11,563 / 3,348) [2 pts / 1 pt / euchred] EV = +0.42
play 10S, then lead QC, trick 2 [if win trick 1]: (966 / 11,801 / 3,008) EV = +0.49
play 10S, then lead 9H, trick 2 [if win trick 1]: (807 / 12,088 / 2,880) EV = +0.50
Yep me too, but at least that idea is not wrong. It's just not as "right" as I thought it would be.
Unread post by raydog » Sun Jan 23, 2022 1:57 pm
Unread post by Wes (aka the legend) » Mon Jan 24, 2022 1:35 am
Good stuff. That makes sense to me. Discarding the QC had to be ruled out cuz it had some plausibility. And I think you did that.raydog wrote: ↑Sun Jan 23, 2022 1:57 pmWes, I just tried discarding the QC instead of the 9H (other two plays are the same - trump, followed by QC or 9H lead). Worse outcome: EV = +0.36, vs. the same for the other two plays. Given that C is a VERY short suit (both 9c and JC gone), I think getting void in a 6-length suit would be far more valuable. So I'm not surprised by the result.
I do think leading the AC in that spot is best. At the very least it will force out enemy trump which is still a positive thing. The fact that the AC is so dirty--only the QC left in the wild--wouldn't deter me from leading it. Just the opposite in fact. Against a 2S call this dirty lead could easily create an overtrump situation for our team. Our dirty AC lead puts S2, the maker, in a virtual squeeze sitsituation. That's a plus for S1-S3. Another thing, the AC is so dirty if we hold it back we'll probably die with it getting no use out of it. Leading the AC to force out enemy trump or to create a situation where the maker gets overtrumped is likely the only strategic value we can milk from it.raydog wrote: ↑Sun Jan 23, 2022 1:57 pm1'm also confused by your suggestion of a AC lead by S1 on 1st trick. With 9C buried, only the QC remains. (the fact that S4 passed is only relevant for the JC, which now counts a s a Spade). 2/3 chance it lies with an opponent BUT if S2 has it and S3 trumps, S4 can (likely) overtrump. So really a 50/50 proposition.
I'll leave it to you to simulate. Need to specify S1 cards and the turn, reach a bid by S2, and perhaps specify how S2 reacts to am AC lead by S1 - too far down the rabbit hole for me. Though I agree it is an interesting question.
Unread post by raydog » Mon Jan 24, 2022 10:06 am
Unread post by raydog » Mon Jan 24, 2022 10:40 am
Unread post by Wes (aka the legend) » Tue Jan 25, 2022 12:24 am
raydog wrote: ↑Mon Jan 24, 2022 10:06 amThe scenario you give is very hard to simulate, because so much depends on how S1 plays subsequent tricks as the hand plays out.
As a first pass, I compared 3 different scenarios :
1) lead JS (R)
2) lead 9H
3) lead 9D
The last 2 should be identical, but because the way my program is set up, if S4 has identical singletons in H and D, they will discard the H [I haven't randomized this; the program goes through suits in order], so leading the 9D is more favorable by about 0.007 of EV. Small, but important to realize that there are biases introduced by my algorithm.
raydog wrote: ↑Mon Jan 24, 2022 10:06 amIf the L is with S3 (or buried), the game is secure. If it's with an opponent, it's more likely to be with S4. So trump in with the 10S? That's how the program currently plays. I would posit that this is the best way to win 5 tricks as well, if the opponents don't have the L.
raydog wrote: ↑Mon Jan 24, 2022 10:06 amIt's possible that this situation really is a toss-up, but I want to be sure my program is playing as the experts would. Let me know if there are specific scenarios that experts would play a certain way, and that my program may be plating incorrectly. It would only take a few rather rare scenarios, played better, to tilt the balance in favor of leading the JS.
raydog wrote: ↑Mon Jan 24, 2022 10:06 amUpon further reflection, the decision tree becomes even more complicated if the 9H or 9D is led. Perhaps with expert play in every scenario, and perfect coordination with partner, one or the other leads by S1 trick 1 would prove a bit better. But my guess is it will be close. So my advice would be to lead the JS. Easier to play (less likely to make a "mistake" - suboptimal play) and less reliance on partner playing correctly. Too complex for me to try to simulate.
Unread post by raydog » Tue Jan 25, 2022 12:07 pm
Unread post by Wes (aka the legend) » Tue Jan 25, 2022 12:55 pm
This is great stuff man!! It leads me to formulate this hypothesis:raydog wrote: ↑Tue Jan 25, 2022 12:07 pmFound interesting results.
In the first scenario (I had QS as turn card), I found:
lead JS: EV = +0.60
lead 9H: EV = +0.68
Which supports your end-play theory.
But with the second scenario I found:
lead JS: EV = +0.06
lead 9H: EV = -0.25
Really need to draw trump! Also tested if lead the 9S: EV = -0.18.
Unread post by Wes (aka the legend) » Tue Jan 25, 2022 1:30 pm
Also it almost goes without saying that we should lead trump (R) with this J-A-X S1-R1 hand:Wes (aka the legend) wrote: ↑Tue Jan 25, 2022 12:55 pmThis is great stuff man!! It leads me to formulate this hypothesis:raydog wrote: ↑Tue Jan 25, 2022 12:07 pmFound interesting results.
In the first scenario (I had QS as turn card), I found:
lead JS: EV = +0.60
lead 9H: EV = +0.68
Which supports your end-play theory.
But with the second scenario I found:
lead JS: EV = +0.06
lead 9H: EV = -0.25
Really need to draw trump! Also tested if lead the 9S: EV = -0.18.
Those times you call from S1-R1, if you have R+2+0 always lead the Right unless you have J-A-X in trump, in that case lead offsuit and try to set up and end play.
That said I would predict this hand would be an exception tho:
Upcard:![]()
S1:![]()
![]()
![]()
![]()
![]()
The idea being one of our off Kings will be boss often enough to swings things back to leading the Right. If true, then I'd wonder what about 1 King:
Upcard:![]()
![]()
![]()
![]()
![]()
![]()
Unread post by Tbolt65 » Tue Jan 25, 2022 9:16 pm
Don't turn into an Eric Zalas. It is these complicated decisions and hand recognition during actual play of the hand that is the difference maker. Something which is hard to program into setting parameters. When considering how and why to trump or play off for example. Or take into consideration your partners likely holdings. There are times that you need your partners help but when to employ that help is by how things play out and when you have no other choice but to trust your p to get 1 or your set. All theses things and more change on the fly. Its recognizing in the moment based on multiple factors. Some of which you just can't program.raydog wrote: ↑Mon Jan 24, 2022 10:40 amUpon further reflection, the decision tree becomes even more complicated if the 9H or 9D is led. Perhaps with expert play in every scenario, and perfect coordination with partner, one or the other leads by S1 trick 1 would prove a bit better. But my guess is it will be close. So my advice would be to lead the JS. Easier to play (less likely to make a "mistake" - suboptimal play) and less reliance on partner playing correctly. Too complex for me to try to simulate.
Unread post by Wes (aka the legend) » Tue Jan 25, 2022 9:42 pm
I think it's actually the opposite. We can only get so far without Ray's simulator. For many gray area spots Ray's simulator is our only hope of finding a reliable answer. We can't get "to the nth degree" without it.Tbolt65 wrote: ↑Tue Jan 25, 2022 9:16 pmDon't turn into an Eric Zalas. It is these complicated decisions and hand recognition during actual play of the hand that is the difference maker. Something which is hard to program into setting parameters. When considering how and why to trump or play off for example. Or take into consideration your partners likely holdings. There are times that you need your partners help but when to employ that help is by how things play out and when you have no other choice but to trust your p to get 1 or your set. All theses things and more change on the fly. Its recognizing in the moment based on multiple factors. Some of which you just can't program.
From what I've been reading. You have been doing a good job. Where rigid parameters can be set. However that can only take one so far.
Tbolt65
Edward
Unread post by Tbolt65 » Tue Jan 25, 2022 9:54 pm
Return to “General Euchre Discussions”