'test meld' suggestions

[ToreadorElder]

Putting a suggestion in here, continued over from another suggestion...

Test weight is, I think, being massively and disproportionately valued. The difference in time between counting a hand with (say) just a meld and a pinochle, versus something complex with 5-6 meldable elements, isn't a factor of 3...because you still have to check the first hand for the arounds and the pinochles and the like. A glance at the leaderboard shows how dominant a factor the test weight is. And I'm looking at rak's and mick's scores, comparing test weights for the best time result and for the best score result. mick's best *time* is 105 seconds, on a test weight of 25; the time on his best *score* is 122 seconds, with a test weight of 41. The weight increased by 65%, but the time required increased by 15%.

So I think there's a huge penalty on those hands with only 1 or 2 meld components. There's infinitely many ways to tweak the weighting...I might suggest something like:

a) For hands with 0-3 meld units, Test Weight == 3 for that hand
b) for each meld unit > 3, the hand Test Weight increases by 0.5 or 1

So complex hands that do take longer to count, will score somewhat better...but not massively so. There's some baseline time required to check for everything, present or not.

[mickmackusa]

(07-10-2013, 02:22 PM)ToreadorElder Wrote:  My issue was simply the whole double-counting thing. It's not spelled out very well, that this was the intent, and it's uncommon to do this. I hate bad, confusing rules.

Amen.

[rakbeater]

Didn't I hear a rumor that we only take suggestions about the Meld Test from those who have actually played it?

[richardpaulhall]

This is early days in the science of meld counting.

No matter what the hand you have to check for "Arounds", Marriages, and Pinochles. That's where the time is.

Once a database of thousands of hands was build then the data analysis could tell you which hands were slower to count. Then we could attempt to figure our why the hand was faster or slower to count.

[mickmackusa]

(07-10-2013, 09:23 PM)rakbeater Wrote:  Didn't I hear a rumor that we only take suggestions about the Meld Test from those who have actually played it?

All BS'ing aside, I like the way TE is thinking about the program.
Careful, logical thinking is why I usually value TE's analyses.
Thanks for your thoughts.

Rep +2 for the suggestion.
Rep -2 for never playing my game.
...it all balances out.

[ToreadorElder]

It shouldn't be that hard to realize why I choose not to play.

That aside: am I correct in assuming that the trump suit is chosen at random? Sure looks that way. This isn't pinochle-thought. One hand had, I think, an 8 card dbl ace run in diamonds and only KQ of spades, but the trump suit designated was spades.

If I'm counting meld, I count relative to NO trump suit, or relative to the suit *I* plan to name. If you're doing this for purposes of your being able to count the meld in the code...the way to count relative to no trump suit would be to count the meld for each suit, then take the minimum value. The number of meld units is defined by that suit. What I'd suggest is some kind of control that lets the player designate the trump suit, for those times he's got that nice big trump suit and he's saying, I want to play it...with the additional option of saying NO trump suit, for those support hands.