Scores, Decisions, MPs

[awm]

I'd like to try and quantify a matchpoint score based on how many decisions a pair got right during a session. Obviously there's a lot of randomness involved here (how many decisions did opponents get right with this pair at the table, how good is the field in general, how flat are the boards, etc). But in any case, it seems like you can try to quantify an expected score by determining:

 

(1) What is the average number of "decisions per board"?

(2) What percentage of these decisions does the field get right, on average?

(3) What percentage did this particular pair get right?

 

Note that getting all your decisions right doesn't mean you had a 100% game -- some part of the field is also getting decisions right (diluting your tops), and on some boards you just won't have any decisions to make (either the board will be flat, or you will simply be handed a result based on your opponents decisionmaking). Of course there is some inaccuracy here too, because some pairs are capable of getting "non-decisions" wrong, and other pairs are capable of creating decisions most of the field won't see, but I'll try and ignore that for simplicity.

 

So the questions I'd like to ask are:

 

(1) Suppose you play a session where it seems like you do nothing right. What range of scores do you expect to see in a typical club field? How about in a national or international caliber field?

 

(2) Suppose you play a session "perfectly" without any errors. Again, what range of scores do you expect to see in a typical club field? How about a national or international caliber field?

[hrothgar]

There already is a measure of how many decisions that I get "right" durng a session. Its called my MP score. I think that you are headed into a world of hurt if you try to move much beyond this.

 

The very notion of getting a decision "right" is incredibly problematic. I suspect that you can find a number of cases - primarily related to declarer play - where there is clearly a dominant technical line. However, there are also an enormous number of cases in which a line that is technically inferior might yield superior results. (The opposition is far from perfect and they can be lead down the primrose path)

 

Worse yet is the amount of time and effort required to implement such a system. I know that professional teams like the Dallas Aces spent a lot of time analyzing hand records trying to quantify the same types of information you're talking about. I can't imagine every trying to automate something similar to this.(Speaking of which, I would expect that Bob Hamman probably has good perspective on the pros and cons of this type of system. In theory, someone could try to solicit his opinion regarding an appropriate choice of parameters). If you really want to use this type of methodology for a rating system, you'll need some way to automate things. I can't see this ever happening.

 

Long term, I suspect that you might be better off looking at par contests, Bridge Master hands, and other such systems that are deliberately engineered to test one's ability to get things "right"

[Vilgan]

This seems way too complex to ever be quantifiable. Suppose 1 line is technically 2% better, is that the correct decision? What if there are inferences that can be made that suggest the other line is the one that is actually correct? Also, table feel can frequently sway a decision significantly more than a slight superiority of percentages.

 

I agree with the other poster.. this is one thing that bridge master excels at. You can polish your skills and figure out the technically superior choice every time. However, knowing the technically correct play does not guarantee superior performance at the bridge table. There is a lot more to bridge than just wrong/right decisions and knowing %'s.

 

Then again.. maybe I'm missing the point of this <_<

 

Eric

[BebopKid]

I think the best I can relate is the scores I get at our local club playing either 0-500 MPs or Open MPs, many 2000+ players and at least 3 10000+ players.

 

When a partnership does everything right, expect 60 to 70 %, though I did get 71% a couple of weeks ago in a 10 table game.

 

When a partnership does nothing right, expect 20 to 30 %, though usually even though I felt everything went wrong. my bad days are usually 38 to 44%

 

I would also say that about 50% of the time, the scores at our club range from 35 to 59 % and the rest 30 to 68 %

 

It really does depend on how the whole field is playing. With two pairs playing equally well, your score will not be as high. The more tables you have, generally the lower the top score.

 

I hope I've helped a little.

[awm]

Let me try to explain what I mean....

 

Obviously my MP score will depend on all three of: (1) How well partner and I played. (2) How strong the field is. (3) Luck. We can try to reduce the luck factor by averaging over many trials. The question I was trying to resolve is "how does strength of field effect my matchpoint score, assuming that I play roughly the same way?"

 

My feeling is that it's pretty close to a 5% difference between scores in a local club game and a regional tournament, and about another 5% difference between a regional pairs and day two of a national pairs. If this is true it's a good sign for a rating system -- it indicates that the "strength of field" adjustment would be basically linear.

 

Of course, it could be that I'm dead wrong here, that there are some pairs who routinely score around 55% at their local club and totally bomb at the regional (say 40% typical) whereas another pair that also routinely scores around 55% at the club does just as well in regional or national fields. I do think there are exceptions like this (for example there's one pair that always wins at the local club but doesn't do much of anything in tournaments) but I also find that my own scores reflect approximately this 5% difference.

[Fluffy]

Well if I got everything right my expected percentages would be:

 

Local: 72-76% (weak opponents)

National: 63-75% (very random opponents)

International: 55-64% (strong opponents, much less experience than the other 2)

[jtfanclub]

I usually get 55%-60% with my regular partners, whether it's a club game, an ACBL online game, or a regional game. The level of competition just doesn't seem to make much of a difference.

 

I've only played in one upper echelon game, and that was to prevent a sitout at the Chicago Nationals. Not much of a sample size.