
https://worldofcardgames.com/#!replayer ... %3A1%7D%5D
Unread post by Tbolt65 » Tue Nov 23, 2021 11:34 pm
Unread post by jblowery » Wed Nov 24, 2021 10:07 am
Unread post by Wes (aka the legend) » Wed Nov 24, 2021 2:02 pm
I'm calling Next from S1-R2 as you guys probably already know. I don't pass in that spot if I don't block reverse Next except at certain scores where I have to tighten up (9-8, 8-8, and arguably 7-8, 6-8, 8-7). Would love to have that spot tested someday but for that we'd need a real simulator comparing one bot team with my philosophy to another bot team with a passing range. Score matters too. I understand that this spot is debatable but if EG you're up by 3 then you gotta call something if you don't block all suits.jblowery wrote: ↑Wed Nov 24, 2021 10:07 amI generally agree. I'd definitely call diamonds from seat 2. Left +2 and it is reverse next, even if I have the other suites blocked like he does.
I'd actually pass in seat 1, round 2. I know a lot of people on here wouldn't though.
I almost always call with 3 of anything in round 2 but I can see seat 3 passing if it is STD and he has the other 2 suites blocked also.
Unread post by raydog » Wed Dec 29, 2021 10:44 pm
Unread post by irishwolf » Thu Dec 30, 2021 12:29 am
Unread post by irishwolf » Thu Dec 30, 2021 12:18 pm
Unread post by Dlan » Thu Dec 30, 2021 12:27 pm
Unread post by irishwolf » Thu Dec 30, 2021 1:47 pm
Unread post by raydog » Thu Dec 30, 2021 2:47 pm
Unread post by Wes (aka the legend) » Thu Dec 30, 2021 3:23 pm
Makes sense. That's what I would've predicted--probably everyone else too--that option 3 would win out. I would not have predicted that it would be +EV tho. Good stuff.raydog wrote: ↑Wed Dec 29, 2021 10:44 pmSo many bidding options with this hand, which ones give good results? I tried to simulate all the scenarios.
First, I looked at S1 bidding R1. And to do that I needed to know which was the best lead for S1 if they bid. I tried 3 different leads: (100,000 hands dealt - I did this several times, not much variance)
[note: I fixed S1's cards and the turn card, all other cards randomized - the situation as seen by S1]
1) KH: 3,362 / 56,795 / 39,843 (2 pts. / 1 pt. / euchred) EV = -0.16
2) JS, then KH: 5,635 / 52,550 / 41,815 EV = -0.20 [more sweeps but more euchres]
3) JS, then AD: 6,223 / 53,100 / 40,677 EV = 0.16
Yep that's what I would've done.raydog wrote: ↑Wed Dec 29, 2021 10:44 pmI compared 1) to 2), then 2) to 3) [can only compare 2 scenarios at a time]. Since 1) and 3) were close, I compared them head to head:
1) KH led: 3,313 / 56,581 / 40,106 EV = -0.17
3) JS led, then AD: 6,133 / 53,108 / 407,759 EV = -0.16
A slight edge to 3), but too close to call definitely. I used JS lead, AD 2nd trick lead for the rest of my comparisons.
I would've predicted that 3) would be the best leads. That's an interesting result to me. But I'm not surprised the choice between lines is so close given that we are starting from a dubious hand to begin with where we really don't have any good lead choices.raydog wrote: ↑Wed Dec 29, 2021 10:44 pmOf course, S1 has the option of passing R1 and bidding Clubs, R2. So I looked at what would be the best lead if S1 called next, R2: (100,000 hands dealt, approx. 63,000 make it to R2)
1') KH: EV = -0.55 [these EVs are based ONLY on the 63,000 hand played]
2') AD, then 10D (if A wins): EV = -0.51
3') JS, then AD (if JS wins): EV = -0.55
It looks like AD is the best lead, so I used this lead in the rest of my comparisons.
THIS is BY FAR the your most fascinating finding. Your results suggest that there are A LOT of R+1+A combos from S1 that should be calling hands instead of bagging hands. That's a pretty HUGE finding as almost every player I know--good or bad--doesn't call from S1-R1 with just R+1 unless they have 2 aces to go along with it. People are very timid, including myself, when it comes to ordering up the enemy--giving them a trump and a chance to create a void--with just 2 trump.
When it's a statistical tie, I would always choose the line that leads to less opponent loners. Controlling loner variance is especially important in tournament play when one bad game can sink your ship. Therefore, calling in the first round still wins in my book. Instead of "tie goes to the runner" we can say "tie goes to the loner stopper".
Although we would expect passing to beat out calling for S2-R1, I am surprised at how effective calling was in the absolute since. +.44 for L+1+0 is better than I would've thought. From your findings I would actually create this hypothesis:raydog wrote: ↑Wed Dec 29, 2021 10:44 pmWhat if S1 passes, R1 - should S2 bid, R1? (1,000,000 hands played; 4% are bid by S1, R1)
[note: I fixed S2's cards and the turn card, all other cards randomized - the situation as seen by S2]
7) S2 bids, R1: EV = +0.44
8) S2 passes R1, bids D, R2: EV = +0.53
It seems that S2 has a good defensive / supporting hand, and a good bid R2, so better to pass R1.
This is exactly what I would've predicted. S4-R1 call has a negative expected outcome but is +EV overall due to the high cost of passing this hand. Good stuff. Glad to see my play here being confirmed. I will even say the word "confirmed" here cuz that's a pretty wide spread (-.34 vs -.73). I think we can feel safe about that one.raydog wrote: ↑Wed Dec 29, 2021 10:44 pmWhat if everyone passes to the dealer? (100,000 hands, ~63,700 make it to dealer for a bid)
9) S4 bids, R1: EV = -0.34
10) S4 passes, R1: EV = -0.73
Here, S4 does not have a very good bid, BUT it's still better to bid than pass. Passing sets up S1 to bid almost 50,000 hands for a net EV of +0.97 (when bidding)
Very good stuff! I favored calling but I actually thought the results would be a little bit closer given that S2 blocks all suits.raydog wrote: ↑Wed Dec 29, 2021 10:44 pmIf everyone somehow passes R1, I determined earlier that S1 should also pass, R2. What if it makes it to S2, R2? I tested 100,000 hands, and only ~14,000 made it this far, so the interpretation of the results starts to be a bit too reliant on knowing for certain the bidding behavior of the other players. But for what it's worth:
11) S2 bids D, R2: EV = +0.30
12) S2 passes, R2: EV = -0.19
No surprise here, it's the result I found earlier. If S2 passes, S3 calls trump about 80% of the time, with good results.
Yep makes sense that passing is best in that spot. This is especially true in a tough game. If the action gets to S3 in a tough game it's a safe bet that someone has all suits blocked or at least 2 out of 3 suits blocked. Whether that's your P or S2, that's even more reason to pass. In the former case the probability of euchring the dealer goes up, in the latter case the probability of something going wrong when you call goes up.raydog wrote: ↑Wed Dec 29, 2021 10:44 pmFor completeness, what should S3 do, R2, if it makes it that far? Of 100,000 hands, only about 4.5% made it that far. I ran the comparison 3 times and averaged the results (so 300,000 hands):
13) S3 bids Hearts, R2: EV = -0.13
14) S3 passes, R3: EV = +0.25
Same caveat with the small sample of hands that make it this far, but the result does seem reasonable, since S4 will be forced to bid (STD) and S3 has some good stoppers and euchre potential.
Yep. Score definitely matters becuz my assertion is that in euchre it is often correct to make -EV calls to protect our equity. Classic example:raydog wrote: ↑Wed Dec 29, 2021 10:44 pmOf course, as Wes pointed out, score matters, as does the aggressiveness of the players; that can change the subset of hands which make it to various stages of the bidding and affect the results. I can't simulate all that. But hopefully something useful can be drawn from this analysis.
Unread post by Wes (aka the legend) » Thu Dec 30, 2021 3:35 pm
One thing I will add: So many times I have made thin, dubious Next calls and have gotten euchred becuz my amateur partner didn't know what to play or lead on subsequent streets. Whereas if I had an expert partner my call would've made it. Well 95-99%+ of the time you WILL be playing with an amateur P if you're playing on the app or in a tournament where your partners are picked at random. This reality makes it EVEN MORE likely that your finding--that calling with that R+1+A from S1-R1 is better than bagging and calling Next--is correct. I am very excited about that!raydog wrote: ↑Thu Dec 30, 2021 2:47 pmIt's a typo, dude! - but you knew that. Should read 40,759 euchres. And I'm glad you are looking at the details, that is why I included them.
My program, as you say, is not Big Blue, and the card led on the 3rd trick will be based on some aggregate of hands with similar characteristics, NOT the exact cards specifically led in this scenario. So an expert player should conceivably play better. But that also depends on the expert's intuition being infallible, never making mistakes, having tens of thousands of situations memorized or being able to accurately assess the odds of different plays in a few seconds. A tall order.
I obviously don't know a thing about your program, but man it seems legit from my perspective. It either spits out results I already strongly agree with or those times it spits out results that suggest my claim is wrong I'm' like "Yeah I can see that". It didn't have to be that way, your program could've been producing results that had me going "NO F***ING WAY" all the time. That's not happening. Only once did your program spit out a result I strongly disagreed with and it was on that loner where the program wanted the maker to keep the lower ranking garbage card. I see that as a minor glitch. But overall I am extremely impressed with your work.raydog wrote: ↑Thu Dec 30, 2021 2:47 pmBecause I don't examine and take account of the minutia, I don't claim my results to be definitive unless there is a comfortably large margin between two different scenarios. In the example you mention, I don't think an expert's knowledge would change the outcome of the game more than 5-10% of the time, max. [quite often they would play the same as my program; and in the majority of cases it wouldn't make any difference]. Yes, that's significant, but in the example you cited, getting 10% more sweeps (an extra 600) instead of 1 pt wins changes the EV by 0.006 - hardly moves the needle. Now, 5% less euchres would improve EV by 0.06 - much, much better, but I question how likely such a big change is. Such a big part of the game is just luck.
We could debate this ad infinitum. But I won't. I believe I understand the limitations of my program, and I'm not touting it as any more than what it is. I, naturally, have far more confidence in it than an outsider, because I programmed it, know intimately how it works and have seen how much that labor of love has improved my own game. I'm just having a little fun and challenging my brain cells.
Unread post by raydog » Thu Dec 30, 2021 6:01 pm
Return to “General Euchre Discussions”