Today we test computer dealt hands against live deals.

Ask questions, discuss and debate your strategies, euchre polls and more
Post Reply
User avatar
Dlan
Site Admin
Posts: 588
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

Today we test computer dealt hands against live deals.

Unread post by Dlan » Thu Apr 01, 2021 9:59 am

Many times I’ve heard that this euchre site or that euchre site’s deals are messed up. Having played a lot of live euchre as well as a fair amount on-line, I don’t believe there is a big difference. I’ve seen new players have incredible winning streaks and great players lose over and over. There’s no question, better players win more in the long run, but short-term results have little meaning.

There is a tendency to remember the extremes, those days when the cards were dealt in your favor, or the times when it seemed the computer was out to get you. Yet the times when the cards were average tend to fade away. Such is the case with online euchre games. It’s easy to think something must be amiss here. I wasn’t getting these types of crap cards the time when I made 9 lones in a single night.

Up until now, there has never been a way to test whether computer deals are the same as live deals.

Here I’m using the euchre workshop to deal 70 hands and see how close my results would match which Wes got using a standard deck of cards.

The results may surprise you, the numbers came out relatively close.
Here is my data.

Calling gets euchred: 29/70 = 41.42%
Calling gets 1 pt: 34/70 = 48.57%
Calling gets sweep 7/70= 10%

And here is what Wes got
viewtopic.php?f=8&t=540&sid=7aa5af39ace ... e65a4dc09c

Calling gets euchred: 27/70 = 38.57%
Calling gets 1 pt: 33/70 = 47.14%
Calling gets sweep: 10/70 = 14.29%

I’ve added counters on the workshop that you can use to track euchres, points made, and sweeps.

Image



irishwolf
Posts: 1050
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Thu Apr 01, 2021 10:29 am

Good job Don! Very close and well within the margin of error against Wes test, especially with just 70 hands. Sweep are higher.

So my question would be how long did it take you to do your 70 hands? I LIKE IT!

User avatar
Dlan
Site Admin
Posts: 588
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

Unread post by Dlan » Thu Apr 01, 2021 10:58 am

A little over 3 1/2 hours

Wes (aka the legend)
Posts: 1423
Joined: Wed Jun 13, 2018 3:03 pm

Unread post by Wes (aka the legend) » Thu Apr 01, 2021 8:58 pm

irishwolf wrote:
Thu Apr 01, 2021 10:29 am
Good job Don! Very close and well within the margin of error against Wes test, especially with just 70 hands. Sweep are higher.

So my question would be how long did it take you to do your 70 hands? I LIKE IT!
Wow this is really F-ing cool. Like an independent way to test my sample to make sure nothing's out of line. Love it. The 3.5 hours bit is disappointing tho. I do around 20 samples per hour so it would take me 3.5 hours too! Thought there would be a significant time improvement. Well at least Don didn't have to grind at the kitchen table!

User avatar
Dlan
Site Admin
Posts: 588
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

Unread post by Dlan » Fri Apr 02, 2021 11:02 am

Using this hand and the data sets provided, let us take it a step further, we would make 48 points, but we give up 58.

Wes’s data shows us making 53 points but giving up 54.

At what point would a hand be a recommended call vs a pass?

irishwolf
Posts: 1050
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Fri Apr 02, 2021 2:01 pm

My personal guide on points for and against is what I call the breaking point which is a 38 to 40% euchre rate. Over 40%, you just cannot make enough points, sweeps and 1 pointers, to make it +EV in the long-run.

And if the points are more projected to be more negative in round 2, decide if a Defensive call fits the situation. Don't try to over complicate it or try to out think the cards. If you don't know be conservative as aggression does not pay unless you have HARD facts about the expected outcome. He, he!

I will let Wes interpret his test results.

~Irishwolf

Wes (aka the legend)
Posts: 1423
Joined: Wed Jun 13, 2018 3:03 pm

Unread post by Wes (aka the legend) » Fri Apr 02, 2021 4:45 pm

Dlan wrote:
Fri Apr 02, 2021 11:02 am
Using this hand and the data sets provided, let us take it a step further, we would make 48 points, but we give up 58.

Wes’s data shows us making 53 points but giving up 54.

At what point would a hand be a recommended call vs a pass?
IMO we can't safely recommend a call vs a pass until the point difference between competing strategies becomes statistically significant, I.E. our P value gets below .05. The smaller the likely difference is the larger our sample would have to be to reach that level of significance.

And Ideally we'd have at least 1 other person confirming the results with their own sample. After all, independent replicated results is the cornerstone of science.

irishwolf
Posts: 1050
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Fri Apr 02, 2021 10:00 pm

So give me some time and I will see what the results might be. I looked at it QH 10H 9H AS 9S, and think it close to having a +EV by itself. I will investigate euchres, sweeps and points at S3 ordering the AH.

IMO we can't safely recommend a call vs a pass until the point difference between competing strategies becomes statistically significant, I.E. our P value gets below .05. The smaller the likely difference is the larger our sample would have to be to reach that level of significance.

And Ideally we'd have at least 1 other person confirming the results with their own sample. After all, independent replicated results is the cornerstone of science.

Wes (aka the legend)
Posts: 1423
Joined: Wed Jun 13, 2018 3:03 pm

Unread post by Wes (aka the legend) » Sat Apr 03, 2021 2:54 am

irishwolf wrote:
Fri Apr 02, 2021 10:00 pm
So give me some time and I will see what the results might be. I looked at it QH 10H 9H AS 9S, and think it close to having a +EV by itself. I will investigate euchres, sweeps and points at S3 ordering the AH.
Awesome man. Love it!

irishwolf
Posts: 1050
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Sat Apr 03, 2021 10:12 pm

Here is what I got on 150 hands. I did pull out 8 hands that S2 assisted for a total of 142. However, there were 5 others that were border line that could have gone either way. A hand like this JD AH KD 9D XX. a HAND like that S2 might want to bag on next. So there were five like that that I did not pull out. 40% of the time S3 could slough his 9S to off suit ace led (AC OR AD).

Any way the results:

Calling gets euchred: 52/142 = 36.61%
Calling gets 1 pt: 74/142 = 52.11%
Calling gets sweep 16/142= 11.27%

-74 + 32 + 52 = EV +10 PER 100 HANDS

Irishwolf

User avatar
Dlan
Site Admin
Posts: 588
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

Unread post by Dlan » Sun Apr 04, 2021 4:14 pm

The idea of calling from 3rd with 3 small trump was not a call I’d even try, at least not before now.

First-round calls from 3rd are notorious for getting one euchred. I think we’ve all been told that you only call from 3rd with a really strong hand.

Holding the 3 lowest possible trump doesn’t seem to qualify. You also have to consider that you’re putting a trump in the dealer’s hand.

After Irishwolf’s initial post, and the testing both Irish and Wes have done, it may be time to reconsider.

Here is one I’ve just tested. 3 bottom trump with two off-suit aces.

Calling gets euchred: 13/50 = 26%

Calling gets 1 pt: 26/50 = 52%

Calling gets sweep 11/50 = 22%

46 points made, 26 points given up

Looks like this 3rd seat hand is a must-call.


Image

irishwolf
Posts: 1050
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Sun Apr 04, 2021 7:28 pm

I would expect the results of your test, QH 10H 9H AX AX. Should be similar to same I tested. But let me also say, with the hand AS 9S AH up, those 8 hands I pulled out as ASSIST HANDS, if not pulled out my results would be right there between you and Wes, borderline, breakeven. I am not sure what you two did about those like JD KH QD 9D XX or those similar that at times you are bagging on a next call. Those would most likely be euchres.

Here is another hand QH 10H 9H KS QS - those will have a slightly negative EV, and QH 10H 9H TWO SINGLETON OFF SUITS will be a loser hand. Even one Ace with singleton off suit a negative result. But if Next is called, that too will be a losing negative hand unless S1 has a powerful hand.

Just me opinion and testing.

justme
Posts: 22
Joined: Sun Apr 04, 2021 6:06 pm

Unread post by justme » Sun Apr 04, 2021 8:03 pm

irishwolf wrote:
Sat Apr 03, 2021 10:12 pm
Any way the results:
Calling gets euchred: 52/142 = 36.61%
Calling gets 1 pt: 74/142 = 52.11%
Calling gets sweep 16/142= 11.27%

-74 + 32 + 52 = EV +10 PER 100 HANDS

Irishwolf
[quote=irishwolf
Any way the results:
Calling gets euchred: 52/142 = 36.61%
Calling gets 1 pt: 74/142 = 52.11%
Calling gets sweep 16/142= 11.27%

-74 + 32 + 52 = EV +10 PER 100 HANDS
[/quote]

Using your 142 hand test i come to a different conclusion. ( i bold printed the main cause of the difference)

calling gets euchred: 52/142 =36.61%, or otherwise 36.61/100 = 73.22 points
calling gets +1 pt: 74/142 = 52.11%, or otherwise 52.11/100 = 52.11 points
calling gets sweep (+2pts): 16/142 = 11.27%, or otherwise 11.27/100 = 22.54 points

-73.22 + 22.54 + 52.11 = EV of: negative .143 per 100 HANDS

Appears your sweep count was mistyped in your last line. should read = +22 not +32

justme
Posts: 22
Joined: Sun Apr 04, 2021 6:06 pm

Unread post by justme » Sun Apr 04, 2021 8:13 pm

Dlan wrote:
Sun Apr 04, 2021 4:14 pm

Calling gets euchred: 13/50 = 26%

Calling gets 1 pt: 26/50 = 52%

Calling gets sweep 11/50 = 22%

46 points made, 26 points given up

Looks like this 3rd seat hand is a must-call.


Image
Shouldn't it read: 48 points made, 26 points given up.
So EV = 44/100 hands

I'm new at this so........ go easy on me if I'm wrong. :)

User avatar
Dlan
Site Admin
Posts: 588
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

Unread post by Dlan » Sun Apr 04, 2021 9:12 pm

Justme, Welcome to the forum,
Testing was done on 50 hands. the calling team in 3rd seat made 48 points and they donated 26 points to the other team. Not having much of a math background, this EV thing is over my head.

Edit by Dlan 4-5-21
Just to be clear, this forum has readers of all levels. The goal is to help everyone learn a little more about the game of euchre. With that in mind, stating the EV (Expected Value) of a hand plus giving the win/lose percentages of a hand will offer help to the widest group of players.


Irishwolf,

I played every hand as if it was ordered from third. On the first set that I did, there was a couple of hands where 2nd seat could/should have ordered. I didn't kick them but thinking I should have.

I guess we need a standard method of testing. What type of hands should be kicked, left in? Your thoughts?

irishwolf
Posts: 1050
Joined: Tue Apr 24, 2018 9:33 pm

Unread post by irishwolf » Sun Apr 04, 2021 11:52 pm

Not sure, but one thing is that our results suggests all three tests are in the margin of error as expected, meaning close and can be repeated. Total would be close to 300. You can reasonably expect a +/- 5% if you repeated the test. And sometimes, human judgement, one player to the next after winning a trick has a choice of picking the next card. That and random dealing is your variability from test to test. Even flipping a coin, heads or tails will seldom be exactly 50:50 with 400 coin tosses. Off a little bit is normal and expected, even with a fair coin. The larger the sample, the tighter the mean will be to the expected average.

Now to a standard when testing. I think if you were or I were playing, ask yourself would you routinely assist with the holding. Not to bag or do something to another player style. Take that part out of the equation (test). And samples are pulled out, say how many were pulled out of total sample and maybe, give an example of those. To leave hands in that for sure would assist, changes the dynamics and can skew results, IMO. So if you had for example 2 or 3 that would have assisted, your euchre rate comes down a bit??

So I pulled out 8 of 150, netting 142 that would have been euchres (I expect, did not do that part). So 8 of 150 is 5.33% and my euchre rate would be up a bit but I think that is skewed up too high.

I guess we need a standard method of testing. What type of hands should be kicked, left in? Your thoughts?

Wes (aka the legend)
Posts: 1423
Joined: Wed Jun 13, 2018 3:03 pm

Unread post by Wes (aka the legend) » Mon Apr 05, 2021 7:22 am

irishwolf wrote:
Sun Apr 04, 2021 7:28 pm
I would expect the results of your test, QH 10H 9H AX AX. Should be similar to same I tested. But let me also say, with the hand AS 9S AH up, those 8 hands I pulled out as ASSIST HANDS, if not pulled out my results would be right there between you and Wes, borderline, breakeven. I am not sure what you two did about those like JD KH QD 9D XX or those similar that at times you are bagging on a next call. Those would most likely be euchres.

(Card_J-D) (Card_K-H) (Card_Q-D) (Card_9-D), X

That hand is a pass in my sample unless that X = an off Ace (Altho frankly I go back and forth on this one and have called it many times in real life. Someday I will test this hand for sure).

If X = an off ace like this hand:

(Card_J-D) (Card_K-H) (Card_Q-D) (Card_9-D) (Card_A-C)

Then that hand is a call imo. I am not passing L+1+an off ace when I don't block 2 out of 3 second round suits. Not taking the chance S1 is sitting on a black loner. So that latter hand would force a reshuffle.

Tbolt65
Posts: 717
Joined: Sun Jun 16, 2019 9:14 pm
Location: Las Vegas

Unread post by Tbolt65 » Mon Apr 05, 2021 10:14 am

irishwolf wrote:
Sun Apr 04, 2021 7:28 pm
I would expect the results of your test, QH 10H 9H AX AX. Should be similar to same I tested. But let me also say, with the hand AS 9S AH up, those 8 hands I pulled out as ASSIST HANDS, if not pulled out my results would be right there between you and Wes, borderline, breakeven. I am not sure what you two did about those like JD KH QD 9D XX or those similar that at times you are bagging on a next call. Those would most likely be euchres.

Here is another hand QH 10H 9H KS QS - those will have a slightly negative EV, and QH 10H 9H TWO SINGLETON OFF SUITS will be a loser hand. Even one Ace with singleton off suit a negative result. But if Next is called, that too will be a losing negative hand unless S1 has a powerful hand.

Just me opinion and testing.

I personally passed there with the off ace and with the x. Against a passing partner and weak player as dealer. I would order. Playing with people who picks up. Thats always a pass except at close out scores of 9 or always ordering at 8. Any other score I pass. Everyone should be too.

So thats what I did in my q-10-9h, A-9s kitchen table testing last night. That I posted the results in the other thread.

Tbolt65
Edward

User avatar
Dlan
Site Admin
Posts: 588
Joined: Thu Apr 19, 2018 10:08 pm
Location: Ohio

Unread post by Dlan » Mon Apr 05, 2021 10:51 am

Testing standard

So, can we say that our tests will exclude any hand where a player before you would/should have called? Also that we should include that number in the results? Is there anything to add to this?


If anyone else has hands they’ve tested (either with the workshop or kitchen table) please share on a new post. I’m guessing there are a lot of hands that only come up from time to time that players wonder about.

Post Reply