Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Power Pinochle Player Ratings
#1
I'm not saying I plan to do this but I suppose we could...

If PP archived every game from every site that allowed me to, then we could potentially create our own rating formula.
What would that look like?

I remember using an independent website to track my poker statistics (and others) called SharkScope.
Thoughts?
It's unbelievable how much you don't know about the game you've been playing all your life. -- Mickey Mantle
Reply
#2
No chance, IMO. Not from any automated formula.

The only way to really rate is, does the player win? Against quality opponents, and over time?

SharkScope can gauge with some accuracy the quality of a poker tournament by the buy-in. If a seat's costing $500, it's safe to say that the large majority of players will be fairly good...certainly larger than the percent of players that actually cash. A $3 event? LOTS of total newbies. And anyone can bust out just from luck, so you need lots of events.

And, in a partnership game, the skill level of your partner matters enormously. Your carefully considered plan can get blown away by partner's thoughtless plays. The major luck factor obscures things, so you need a larger sample...or a way to diminish the luck element, that is to say, playing duplicate.
Reply
#3
Here's a nice example hand from a game I just finished.

           S KQQQJT
           H AAKQT
           D AAKKJT
           C AKQ

S AAAKQJT  +-------+  S AKT
H AKKJTT   |   N   |  H AKQQJJJT
D QQT      | W   E |  D AAQJT
C AQQJ     |   S   |  C KKQJ
           +-------+

           S KJJT
           H Q
           D KKQJJT
           C AAKJJTTTT
}
[Auction "S"]
50 52 60 Pass
Pass Pass

The blame is 1000% with East;  60 is egregious.  They were board set, but East has probably 7 tricks...ok, say 8.  With 17 meld, that's 37 at best, and he's gonna have to get 23 from his partner, who did NOT give meld.  5 of those trumps are non-pointers, so 2.5 points per trick is *probably* an overestimate;  that makes things just a bit worse.

So if we gauge on that...if West loses, should this count against his personal rating as a player?  Against an East that makes such a godawful bid, should N/S be credited for winning?

The reason I remember this hand, beyond that it was...20 minutes ago?  East LAID INTO his partner for bidding 50 with 2 meld.

Uhhh...yeah.  Right.
Reply
#4
I'll state again that this is a purely theoretical thread.

...but if we have complete hand histories to pull data from, we could be as scrutinizing as we wish.

"Penalties" for:
  • voluntary bidding and getting No Trump set
  • voluntary bidding +50 more than what was earned in the Meld + Play phases.
  • getting No Board set when partner tabled, say, 10 meld.
  • not saving partner who gets stuck and No Board or No Trump set.
"Rewards" for:
  • setting the Offense
  • pulling 31 counters
  • winning a game by more than 250 points
Obviously I am just spitballing and haven't thought things through.
What sort of occurrences, which could be collected from hand histories, would you look for as favorable/unfavorable?

This is an opportunity to think way outside the box.
It's unbelievable how much you don't know about the game you've been playing all your life. -- Mickey Mantle
Reply
#5
I greatly dislike someone taking the no-marriage intentional set, but it *might* be the only possible way that side can ultimately win. Same with the impossible overbid. I don't save if my suits with marriages are too weak, so that's not anything worth penalizing. Is it good defense to set the opponents...or was it just bad luck, or terrible declarer play...or actually good, effective defense? Same with stopping the defensive save. And winning a game by 250 may have absolutely nothing to do with skill.

Trying to use individual hands is, IMO, awful. It's micromanaging, and all the criteria you mention become imposing a particular set of 'good plays.' It's not thinking outside the box; it's creating the box.

There might be something to aggregate stats...percentage of hands made as declarer, percentage of hands played as declarer where defense doesn't save, the obvious converse conditions for the defense. These numbers would be pinochle measures of effectiveness...they encompass many aspects of bidding and play, and directly speak to winning and losing. Luck *should* to a large degree be removed...everyone gets no-brainer 31s and no-chance saves or drops. They focus on the whole hand, not any super-narrow aspect like bidding with no marriage.

They can also be computed *quickly*. My analysis for the save/dump hand statistics needs only the Dealer, Declarer, Contract, Result, and Score lines...I can ignore everything else. For the 4 stats above, I'd need the players, declarer, and result...and that's it. The result in PBN is shown as N:E...the N/S result, a separator (":"), the E/W result. Declarer is given as a direction. The 4 stats above are, at that point, trivial.

Then, a 'rating' could be done in a wide variety of ways...perhaps starting with just adding the 4 numbers up and maybe multiplying by 100. Or analyze the distributions for modes...clusters and gaps. Say there's 100 players. The top 10 are all within 1% of each other on percent of hands made as declarer. Then there's a gap...#11 is 3% lower, and the gap between 11 and 18 is just 1%. That gap is big, suddenly.

In analysis, a finer step down is the explanatory metric. As the name implies, they attempt to explain WHY the measures of effectiveness come out the way they do, by looking at individual factors. An explanatory metric can help identify specific aspects to improve. (For anyone else who understands the terms...OK, the MoE should probably be winning percentage, and the 4 offensive/defensive are arguably explanatories...but that's a bit too boring. I think the 4 measures I started with are the best area to focus on.)

Some explanatory metrics...percentage of hands played as declarer, and percentage of the hands played *by your side* that you play as declarer. The first might be the Aggressiveness factor; the second, the Hand Hog. Smile Another aggressiveness/hand hog metric might be the percentage of hands using a shut-out bid...bidding 60+ when a) partner has not called so far, and b) there has not been a shut-out or massive meld bid by your RHO.

Interesting *field* metrics (based on all the deals, and not caring about the players) might come out about shut-out bids, altho one will probably have to be careful with them.
Reply
#6
Wink 
Here's a really coarse bar graph:

[Image: 15y9n2d.jpg]

As noted:  it's the percentage of hands declared by a player, that made.  The data's sparse because I required 10 hands played, and one generally plays QUITE a few different people online.  1460 hands...close to 300 different players.  Player 1 is "all players"...just a reference tally that can be used as a baseline.  (The spreadsheet I used, doesn't make it all that easy to throw on a separate horizontal line.)  10 hands played is still too small to say much, given that 84% of the bids made.  There is the one standout line...player 228.  Played 10, made 3.  Yeah, he really is that bad...I remember him.  A very desirable opponent! Smile
Reply
#7
Geez, three out of ten!
That makes me want to see a new thread where we discuss each set, and find the pattern.
I want to know if it is mostly due to ignorance, arrogance, pure chance, or something else.
I think that player could use some guidance.
...it's probably best to keep this player's identity anonymous as the discussion may be rather carping.

But we don't only have to scrutinize the unsuccessful players.  If player 106 is a standout for satisfying contract, let's have a look at why.
Is it because he/she is conservative, lucky, skillful in play, cheating, or something else?

I'd like to see these players under the microscope.
And I'd like to know what their PlayOK ratings are -- if it is indicative of their success or way off base.
It's unbelievable how much you don't know about the game you've been playing all your life. -- Mickey Mantle
Reply
#8
(07-22-2016, 10:58 PM)mickmackusa Wrote:  Geez, three out of ten!
That makes me want to see a new thread where we discuss each set, and find the pattern.
I want to know if it is mostly due to ignorance, arrogance, pure chance, or something else.
I think that player could use some guidance.
...it's probably best to keep this player's identity anonymous as the discussion may be rather carping.

But we don't only have to scrutinize the unsuccessful players.  If player 106 is a standout for satisfying contract, let's have a look at why.
Is it because he/she is conservative, lucky, skillful in play, cheating, or something else?

I'd like to see these players under the microscope.
And I'd like to know what their PlayOK ratings are -- if it is indicative of their success or way off base.

Overbidding routinely, and particularly being the type of player that will take major fliers if the opponents *might* go out if they take the bid.

And, no.  There might be value to it, but that is not clear...especially with small sample sizes.  The luck factor is still quite large.  For one thing, I'm not throwing out the save and dump sets, that probably should be dropped.  Most of the time they have nothing to do with declarer play (and of course no marriage or no 20, on a dump/save, is pure luck.)  And it's not worth the effort to do it.  You can't analyze one hand;  you have to analyze a good-sized set of them.  That is a LOT of work.

If you want to do it...be my guest, once you have your own data sets.  I would say you definitely need to anonymize most detailed analyses.
Reply
#9
(07-22-2016, 02:15 PM)ToreadorElder Wrote:  Trying to use individual hands is, IMO, awful.  It's micromanaging, and all the criteria you mention become imposing a particular set of 'good plays.'  It's not thinking outside the box;  it's creating the box.

By "outside the box" I meant something more clever/thoughtful than what is being done else where.  I was angling for something beyond just wins, losses, and abandons.

I am happy that you put forward some ideas.

The ideas that I mentioned were only intended so spur creativity and acknowledge the broad capabilities that accessing complete hand histories affords.
It's unbelievable how much you don't know about the game you've been playing all your life. -- Mickey Mantle
Reply
#10
Here is a thought for the statistics.

First, we need a 'declarer trick count' based only on his hand. One trick for every side-suit ace...that's easy. The trump suit is trickier to gauge the various cases, but we can work something out.

Then you can run a general stat...when declarer has N tricks in hand, he'll make his bid P% of the time. Yes, I get that there's meld issues, and major lack of support issues...but those should largely even out. There's also cases where dummy has shown aces...so bidding with something like 8 trump and 2 aces total may well be the best choice.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)