A statistic about bidding in 3rd seat

[ToreadorElder]

Or actually, about PASSING in third seat.

Question:  when the auction pass-50-pass-pass, or pass-pass-pass-50...the basic save and the dump auctions...how often does dealer's side go set?

I have a growing library of deals I've played on PlayOK...just under 1300 at the moment.  WIDE variety of partners and opponents.  It wasn't hard to put in a filter to check for these conditions:

Contract == 50
declarer == dealer (for a dump) or dealer's partner (for a simple save)

Then Result == 0 (or NOMELDS, in the case that declarer has no marriage) shows dealer side went set.

There were 17 sets out of 87 simple-save auctions;  there were 13 sets out of 50 dump auctions.  So:  30 out of 137, or basically 20% of the time, the dealer's going set.

A followon question:  clearly, if the dealer's side is going down, the other side has a good result.  Now, obviously, some of the time 2nd seat or dealer will have a nice hand.  When dealer's side saves...how often does the defense ALSO save?  That doesn't mean they ever had meld...but they do get at least 20.  Throw in this check...46 of the 107 remaining hands.  

So what does this show?  When you're in 3rd seat, and the auction starts pass-50 or pass-pass to you:  recognize that the other side may well be in trouble.  They're going set 20% of the time if you pass.  That doesn't mean you semi-automatically pass...it means you need all factors in your favor:

--fairly good offensive orientation
--at least 6 tricks with high confidence (ATKKQQJ trump suit with 2 side aces does NOT count)
--meld + trick points has to be at least 45, IMO.  If you bid 51 and give up at 52, you have taken the risk of taking them off the hook, for no gain.  Also, you have to be conservative about partner's support...if YOU go set, when they were about to go set, that's a complete disaster.

There are some exceptions.  Perhaps:



This isn't all that offensively oriented.  You should have 6 tricks, tho, and you can attack trump with spade leads to try to protect yourself.  Certainly you have tons of meld.  Finally, in clubs or diamonds, saving this meld is not automatic.  It also helps that you can clearly compete to 75.  But something like this:



Man...this one's tough.  LOTS of defense.  Is this a save...declarer on my right is good, but also dealer can still bid...or a possible dump, where I'm wrong-sided?  I don't much like being in 4th seat.  What about the opponents' play skills?  I'd probably bid.  BUT...turn one club ace into a low club and I think I'm passing.  I don't like the trick count, I don't like the timing any more.

[mickmackusa]

Marek told me yesterday that PlayOK holds 100,000 pinochle games per month.
That's over 3000 games per day.
(He doesn't really have any reason to lie about that, and I can't prove that claim to be inaccurate.)

With that much volume, I am making new considerations about what I am going to archive on Power Pinochle for statistics.
Perhaps I should selectively collect hand histories from PlayOK that include players over 1800 rating.
I would probably need to ask Marek to include the ratings minimum on the annotations for this.
If I were to archive every single game at PlayOK (and other sites) I would see the archive grow very, very quickly -- but to what benefit?
At some point, more is not better.
What do you think, shall I set a minimum on ratings to qualify for statistical queries?

I like what you've done with your statistics, TE.
This is valuable information for players to consider when deciding to save or dump.

[ToreadorElder]

I have about 1300 hands. It's all in ASCII...just cut and paste from the PBN feature Marek's using. And the format could be reduced...Marek has a hand diagram that's really nice for quick review, but totally unnecessary for archiving. It's a good-sized chunk...actually about 30% of the total size. (Copy one hand into a new file...check length. Delete diagram, recheck length.)

With no hand diagram, tho, it's about 700 bytes per hand. You'd add some things like the points per trick and who won it...that's 3 bytes per trick line, or 60 bytes. You also wouldn't have some of the punctuation he uses. Figure 750 bytes per hand, more or less.

Remember something: for my meld stats, from random hand creation, I run 25 million deals, to have 100 million hands. I'm also doing this for hand suit distributions...you actually see that one side does get all 20 cards in one suit, maybe once in 10 million deals. More is better...unless you're dealing with a really bad random number generator and all you're seeing is repeats of the pattern.

For other things...WHAT queries? How do you think you'll use the info?

I will agree that harvesting all the hands is not a good idea. It is NOT!!!!!! NOT!!!! NOT!!!! true that players with higher ratings are even remotely competent...I posted about a player who asked for meld with a 6-5-5-4, 6 card single ace run, 1 side ace hand, and a couple other extremely STUPID bids a few days ago. Guess what...he's an 1800. Higher ratings NEVER necessarily relate to skill. That said...LOW ratings (like below 1300) *often* do bid and play very badly...it's something that concerns me in those stats, but right now, I'm hoping that the cases where the dealer or his partner misbid, are mostly offset by hands where 3rd seat *should not* have bid in those save/dump auctions.

[TigreLXIX]

To TE's point...

I think the quality of the game (i.e., all players fairly competent and making mostly smart decisions) should outweigh the players' rankings, which can be somewhat misleading.  The difficulty is in determining how to measure quality without human intervention.  Maybe some sort of formula including the number of hands played (to prevent what used to be known as "Yahoo specials" - completely lopsided games where one partnership got tons of meld in a few hands) (the more, the better), percent of hands where the Declarer was set (the lower, the better), and final score differential (the lower, the better)?  So the game quality could be calculated as

GQ = x*NumberOfHands - y*SetPercent - z*ScoreDifferential

Just thinking outside the box.

[ToreadorElder]

Nothing there says whether a game was bid and/or played well.  If the winning declarers play the hands well to pull 33 instead of 29 a couple of times, it could make for a larger final margin.  

What defines good bidding?  That's probably impractical.  It might be somewhat more reasonable to point out bad bids, but even then the most you might be able to find is the most egregious.  What criteria indicate a bid was really bad?  This might illustrate;  I'll offer my personal rating, on a scale of 1 to 10, for various scenarios.  The hand:



Scenario 1:  First hand of the game, first to bid.  
50:  4.  This is more of a support hand, but may well have the best trump suit.
52:  2.  This is a bad underbid.  
53:  8.  

Scenario 2:  early in the game.  RHO opens 52.
53:  3.  With support orientation, you need to give the meld.
54:  2.  Same underbid issue.  
55:  9.

Scenario 3:  early in the game.  Partner opens 50, RHO bids 52.
53:  1.  Partner has at least as good a hand.
54:  2   As before.
55.  9

Scenario 4:  double bidder out.  As 3, bidding is 50-52 to you.
53:  -10.  You must *assume* LHO is about to jump.  If you don't give meld, your side will only be guessing for the rest of the hand...with the game on the line.
54 and 55:  the choice here is subtle.  In double bidder out:  will partner *push*, and be optimistic that you'll have extras?  If so, then giving 30 might be problematic.  I'd rather give 30...but the 20 here might be understandable, particularly with an unfamiliar partner.

As a sidebar, giving 20 meld only might be reasonable if partner's an overbidder or weak declarer.  Bidding to play in the first 2 scenarios becomes more attractive if partner's a weak declarer.

So there's lots of complications. :