11-05-21 OE Friday - got euchred #2

Ask questions, discuss and debate your strategies, euchre polls and more
Post Reply
User avatar
Dlan
Site Admin
Posts: 672
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

11-05-21 OE Friday - got euchred #2

Unread post by Dlan » Sat Nov 06, 2021 6:08 pm

The maker was euchred. Was there a way to make a point os should they have passed?

https://worldofcardgames.com/#!replayer ... %3A1%7D%5D

Image



Tbolt65
Posts: 820
Joined: Sun Jun 16, 2019 9:14 pm
Location: Las Vegas

Unread post by Tbolt65 » Sat Nov 06, 2021 7:00 pm

The maker makes a decent call as dealer. Its just so happens that seat 3 was sitting heavy on a light caller, 😉

No way out as Wes says.

Tbolt65
Edward

Richardb02
Posts: 748
Joined: Mon Dec 10, 2018 8:57 pm
Location: Florida

Unread post by Richardb02 » Sat Nov 06, 2021 7:58 pm

Good order by S4, just adverse card distribution. I do see another area to discuss.

In my simple approach to Euchre, I lead power, when I have power. The power in S4’s hand is 3 off suit Aces. An Ace lead will draw trump, so it acts like an 8th trump. Now the approach I am proposing fails against this hand, but very few hands have adverse card distribution anywhere close to this configuration.

I recommend leading the Ah. (The Ad is a doubleton and less powerful). It draws a trump from S3. S3 leads the R and of course wins the hand. If S3 had the Qc instead of Ac (switch Ac with S1’s Qc), then the hand could be won. There are other avenues to getting a point. S2 could be void in diamonds and his Js would take the needed trick, etc.

The question is, with all 3 off-suit Aces and 2 weak trump, wouldn’t there be more success leading an Ace, at the first opportunity? For comparison, the normal approach, as used in this hand, is to lead trump, to draw trump and protect the Aces. This effectively sacrifices a trump. My recommendation semi-sacrifices an Ace, to draw a trump.


raydog
Posts: 260
Joined: Thu Sep 16, 2021 6:56 pm

Unread post by raydog » Tue Nov 30, 2021 5:11 pm

I tried simulating this hand with my program. 10,000 hands, with S4 and turn card fixed.

1) After winning 1st trick with an Ace, my program would lead the best trump. I compared this with leading worst trump, and the latter proved more favorable; I have changed my program;

2) After winning 1st trick with an Ace, I compared leading another off-Ace vs. leading low trump. Low trump proved the far better lead: 418 / 4445 / 1611 (2 pts. / 1 pt. / -2 pts.) vs. 279 / 4121 / 2074. More 2pt. wins and less euchres, by a significant margin.

3) I also questioned whether this was a biddable hand, so I compared bidding this hand with passing it.
bid: N/S 5158 pts., E/W 3210 pts.
pass: N/S 4380 pts., E/w 4750 pts.
So clearly a biddable hand.

Why is 2) true? Dealer has no control over the initial lead; they hope no one trumps and they can win with an Ace. From there, they realize they cannot win a trump war: they are ill armed. So the idea is to get trump cards played, to remove trumping potential of their aces. Just lead a losing 9C, draw out trump, and hopefully void an opponent in trump who would have otherwise trumped one of their aces. Need 2 more tricks here: make an ace good, then hope the K of trump can win, or partner can win a trick. A logical thought experiment, which appears to play out when simulated.

Note that the EV of this hand is not very high, but better than the alternative - and that's the metric which counts.

While I cannot guarantee my algorithm plays ideally, I am reasonably confident in the results it gives when the disparity between opposing strategies is a wide as it is in this case. And can be reasonably explained.

irishwolf
Posts: 1319
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Wed Dec 01, 2021 1:05 am

Your conclusion does not match your data - REASON IS EAST IS THE DEALER.

bid: N/S 5158 pts., E/W 3210 pts.
pass: N/S 4380 pts., E/w 4750 pts.
So clearly a biddable hand.

Your data does not add to 10,000 hands? I suppose then you pulled out all those other hands that S1/S2/S3 ordered, but did not state that?

I know of no good player that would Pass at S4 with that hand. However, not surprised leading high is better, IMO. Generally, (there are exceptions when you want the lead) S1 should always just cover the dealer's lead and stay behind the dealer to over trump. Many a euchre player does not understand that! True because if it gets to S3 and S2 has nothing, S3 only has to just cover the 9 vs the King. And your data suggests that difference between 9 & King, (or the Ace but not so with the Left,) IMO. Size counts!

You said, "Note that the EV of this hand is not very high, but better than the alternative - and that's the metric which counts." Yet that is NOT THE WAY YOU HAVE POSTED IT. You have Passing E/W scoring more points than Bidding (in Red)? Your data and so I don't want to assume anything.

Which is it? I don't want to be picky because you are posting/confirming good stuff, but that is pretty big Ray.

IRISH

raydog
Posts: 260
Joined: Thu Sep 16, 2021 6:56 pm

Unread post by raydog » Wed Dec 01, 2021 8:55 am

Sorry, sorry! In my program the dealer is always South, and I didn't change the results to reflect that the dealer in this hand is actually East. So when I gave data for N/S it was for the dealing partnership. Apologies for the obvious confusion.

And yes, I didn't post the data for the cases when S1, S2 or S3 calls trump, R1, because it is the same in both cases so cancels out (it only dilutes the scores overall and makes S4's decision seem less significant). But I do have to run the simulation this way because when S1, S2 or S3 calls trump they winnow down the set of possible card distributions which S4 is actually facing when they make their decision.

Very interesting strategy of S1 just covering S4's lead, it makes sense and I'm going to play around with it. My program doesn't currently play that way but I will test some hands to see if it helps the S1/S3 partnership score-wise. If it proves the better strategy and I adapt my program, perhaps I will then find that leading the KC (in this instance) is better than leading the 9C (as my program currently plays, and finds better).

irishwolf
Posts: 1319
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Wed Dec 01, 2021 1:38 pm

RAY,

That was helpful to clarify that the Dealer is actually N/S.

YES, IT Makes sense that Passing is now a negative EV. Howeer, I don't consider an EV = + .30 as MARGINAL. And passing a negative EV. Jump for JOY!
3) I also questioned whether this was a biddable hand, so I compared bidding this hand with passing it.
bid: N/S 5158 pts., E/W 3210 pts.
Thus (N = DEALER) LEADS LOW. NET = (5158 – 3210 = + 1948 / 6474 hands = EV = + .30 MARGINAL.

QUESTION: I have a question why are the scores for bidding per 2) : 418 / 4445 / 1611 (2 pts. [total net points = (LEAD LOW PTS = 836 + 4445 – 3222 = 2059 points] not the same as 3) bid: N/S 5158 pts., E/W 3210 pts?
Why do they differ?

The discussion of leading low vs high, a different discussion. Leading low or leading high makes no difference 30% of the time when the Dealer has 2 trumps 9C/KC.
JC/QC, JC/JS, JC/AC, JC/10C, // JS/QC, JS/AC, JS/10C, AC/QC, AC/10C / QC/10C AND ONLY 6 OF THOSE MIGHT S2 HAVE AND NOT ORDER. That's .60 X 30% = 18% So S2 can help with six of those 2 trump combos (18%) might not Order and can help if has JC 28%. Those are additive - 18% + 28% = 46% S2 can help. So what about the other 54%?

The other 52%, basically S4 on your own unless S2 has a doubleton better than the opponents, rare help I say.
So, when opponents have 2 each or one has 3 trumps is where a euchre is possible. I see that as 11% + (9% x 2) 18% about 29% euchre rate but I estimate will be less than this, mainly when S2 has the highest trump, JC. This is the all depends factor. Leading trump will ALWAYS help in this situation. So I am not surprised at your 24% euchre rate. In good agreement, I think.

Leading the KC: When S1 has one trump, 10C or QC, leading high is Best. But you never know when that might be. 9C & QC is 40% of the 5 unknown trumps. It gets complex and the advantage of leading the KC, I estimate ~5 to7%, - marginal, all things considered.

I wonder what your program might say? So do those 10,000 hands again 9C vs KC.

Anyway, this is why I favor leading KC. Most players I see, lead the 9C. Not bad but . . . Nothing works 100% but it helps when S1 has one low trump and S3 has two trumps and that is where you are focusing your attention. Small marginal advantages do matter, IMO. More review, it will not work when one opponent has one high trump and the other has like 10C/QC combo. You can't have it your way 100%, but take the best route. As stated, is when S1 has 10C or QC as a singleton and S3 has a doubleton JC or JS (JC buried) with either the 10C or QC.

But to each, his own!

IRISHWOLF

raydog
Posts: 260
Joined: Thu Sep 16, 2021 6:56 pm

Unread post by raydog » Wed Dec 01, 2021 6:23 pm

To first answer your question, Irish, when N/S gets euchred it counts as points for E/W.
So 418 / 4445 / 1611 yields (418x2) + 4445 = 5281 pts for N/S and 2x1611 = 3222 pts for E/W.
This is not exactly equal to the pts given in 3) [N/S 5158 pts and E/W 3210 pts] because with my program I can currently only compare 2 scenarios using the same dealt hands. So I ran the comparisons in 2) and 3) with different sets of 10,000 randomly dealt hands. Within one comparison (KC vs 9C lead, for example), the same 10,000 hands are played, so it is a controlled experiment. It is comforting to see that the difference when running a given scenario with a fresh set of 10,000 hands is relatively small; if it weren't I would have to use a larger sample size.

As for whether to lead the KC of the 9C, I did run this comparison in 1) and found that leading the low trump was better. Specifically, 401 / 4500 / 1725 (2 pts / 1 pt / -2 pts) leading the 9C and
297 / 4422 / 1907 leading the KC, trick 2. BUT my program doesn't currently follow the mantra that the second to play should play low - they just play their best if they can beat the lead - and that should have an effect on the results of the hand. How significant a result is yet to be determined; that's what I want to play around with.

irishwolf
Posts: 1319
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Wed Dec 01, 2021 7:24 pm

WOW, just seems like with such a large sample size they would only differ by a fraction. Puzzling results, IMO as when I do tests like 300 hands or 450 hands they are within 3 or 4 percentage points. Any tests or experiments in Science should be Validated with Repeatability that follow Statistically laws.
Just saying!

Irish

"This is not exactly equal to the pts given in 3) [N/S 5158 pts and E/W 3210 pts] because with my program I can currently only compare 2 scenarios using the same dealt hands. So I ran the comparisons in 2) and 3) with different sets of 10,000 randomly dealt hands."

raydog
Posts: 260
Joined: Thu Sep 16, 2021 6:56 pm

Unread post by raydog » Thu Dec 09, 2021 9:08 pm

Hmmm... When I compare N/S scores (5281 v. 5158) I get 2.4% difference.
When I compare E/W scores (3222 v. 3210) I get 0.4% difference.
You CAN'T compare the difference - if the EV were close to zero you could get % swings of several hundred %, and + or -! Which would make no sense, and nonetheless be perfectly expected with sound sampling.

From what I remember of statistics, once you reach a certain sample size, increasing that sample size doesn't improve the accuracy (unless you get close to sampling a significant portion of the population). Your 3 or 4% is probably close to the best accuracy achievable (95% certainty?) given that the population of possible hands is in the hundreds of millions.

irishwolf
Posts: 1319
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Thu Dec 09, 2021 11:43 pm

I certainly agree with your post: (BUT READ ON.)

Hmmm... "When I compare N/S scores (5281 v. 5158) I get 2.4% difference.
When I compare E/W scores (3222 v. 3210) I get 0.4% difference.
You CAN'T compare the difference - if the EV were close to zero you could get % swings of several hundred %, and + or -! Which would make no sense, and nonetheless be perfectly expected with sound sampling.

From what I remember of statistics, once you reach a certain sample size, increasing that sample size doesn't improve the accuracy (unless you get close to sampling a significant portion of the population). Your 3 or 4% is probably close to the best accuracy achievable (95% certainty?) given that the population of possible hands is in the hundreds of millions."

I suppose I read your results incorrectly and were directed to those difference below which appeared to me far greater? Don't recall seeing the results as 5281 posted in your #2) above. Hard to compare apples to apples when not presented exactly in the same manner? Anyway, I have moved on! - IRISH

3) I also questioned whether this was a biddable hand, so I compared bidding this hand with passing it.
bid: N/S 5158 pts., E/W 3210 pts.
Thus (N = DEALER) LEADS LOW. NET = (5158 – 3210 = + 1948 / 6474 hands = EV = + .30 MARGINAL.

QUESTION: I have a question why are the scores for bidding per 2) : 418 / 4445 / 1611 (2 pts. [total net points = (LEAD LOW PTS = 836 + 4445 – 3222 = 2059 points] not the same as 3) bid: N/S 5158 pts., E/W 3210 pts?
Why do they differ?

Post Reply