Statistical Analysis: Relationship Between Tricks and Points

[SirEdge]

It may be too much to easily figure out, but average score for offense lead versus defense lead would be interesting.  I'm curious if defenders are good at saving up pointers to feed each other and whether they are likely to snatch pointers from offensive players.  It may be that offensive tricks tend to be worth closer to 2 while defensive tricks may be worth closer to 3.  Just an interesting theory to explore.

[mickmackusa]

(05-04-2016, 10:11 PM)SirEdge Wrote:  It may be too much to easily figure out, but average score for offense lead versus defense lead would be interesting.  I'm curious if defenders are good at saving up pointers to feed each other and whether they are likely to snatch pointers from offensive players.  It may be that offensive tricks tend to be worth closer to 2 while defensive tricks may be worth closer to 3.  Just an interesting theory to explore.

I agree with your speculation on average trick "weight" in those cases.
However, the necessary additional factor is: How often do those 3-counter tricks get trumped?

In other words, if we dared to estimate that each of the defense's tricks should be worth 3, then we should also estimate that a percentage of the defense's estimated tricks will be lost. 

I look forward to compiling and assessing concrete statistics from a large volume of real game data to PROVE what the truest/most accurate point ratio is!

[TigreLXIX]

(05-04-2016, 10:11 PM)SirEdge Wrote:  It may be too much to easily figure out, but average score for offense lead versus defense lead would be interesting.  I'm curious if defenders are good at saving up pointers to feed each other and whether they are likely to snatch pointers from offensive players.  It may be that offensive tricks tend to be worth closer to 2 while defensive tricks may be worth closer to 3.  Just an interesting theory to explore.

Other analyses that may be interesting/informative...

• average trick value by Declarer / Dummy
  
• average trick value for first to lead (i.e., Declarer), second to lead, third to lead, fourth to lead (regardless of whether those tricks won were during the player's first time leading)
  
• average trick value after all players have had a chance to lead
  

Anyone else have other ideas?

[mickmackusa]

That is a smart array of statistical targets.  I'll be sure to remember this thread when I get the Archive implemented.

[ToreadorElder]

As usual, everyone misses the point.

2.5 points per trick is easy to use at the table, and it's easy to teach. The variance involved in trying to determine the "most accurate" is far too high to make the exercise worth the effort. EXPERIENCE will show you when to make adjustments, both to trick-winning expectations and to the points you'll win.

And OF COURSE you count the 2 points for last trick. If you don't take last trick MOST of the time, you're incompetent, or bidding to play when you shouldn't. And if you feel you MUST bid, on a hand where last trick is more dubious...fine, make it 2.5 points per trick, -2 points for NOT getting the last. Still easier to do this.

[rdwrites]

Having said that, predicting tricks to multiply by 2.5 is a crapshoot. My Quality Count bidding approach (trump*2 and aces *2) is easier to teach and works just as well or better. We had a thread where I posted the QC counts compared to the 2.5 points and compared well.

Plus I know it works. My game I wrote in early 80's where I came up with this for the game bidding algorithm plays well. (been a freeware DOS game for close to 30 years). We will be able to look at the results in the Animator some time this year.

[ToreadorElder]

It's not a crapshoot for me. And the same factors that will skew trick count, will also skew your estimator.

[rdwrites]

It may not be a crapshoot for you, but you were talking about ease of teaching to which I responded.

The same factors do skew both, and I saw that in the comparison results I posted. The point is that a computer or beginner or even advanced player can bid with confidence simply by counting their aces and trump and doubling it to replace the "20" portion of adding up how high they can bid. A person can also count tricks and see if they can push it higher, but it is a crapshoot and one can talk theemselves into anything if they want. The Quality Count takes subjective guessing out of it for those that are not experienced enough to count accurately while also allowing them to bid higher (or lower!!) than the +20 points they must pull they are taught.