What would be normal and Good

[badderzboy]

There was an interesting article in the English Bridge Union Magazine about looking at your historical results to see where u needed to practice your play...

 

As BBO has a nice DB of your played hands it should afford some estimate!

 

However, what would u anticipate the normal values to be for the following

 

Games Bid Declared with Avg Imps gained and % made

Part Scores Declared with Avg Imps gained and % made

 

and the reverse in defence...

 

What would u consider good scores for the above?

 

I was surprised that on avg only 60% of games in a sample of 500 were made (includes sacrifices too)!

 

Steve

[joshs]

Let me ignore competative game bids, and discuss freely bid games.

 

You will contract for 4M when you think your hands will produce on average 9.5-11.5 tricks (the lower end is slightly lower at imps, especially vul). Since the lower end of these are more common I will guess that the average number of tricks produced by a freely bid 4M bid is 10.25.

 

To estimate the probability of making game, we estimate this from the binomial distribution (basically assuming that we win each trick independantly with probability 10.25/13):

 

P(10 tricks)=13 choose 10 * (10.25/13)^10*(2.75/13)^3

P(11 tricks)=13 choose 11 * (10.25/13)^11*(2.75/13)^2

P(12 tricks)=13 choose 12 * (10.25/13)^12*(2.75/13)

P(13 tricks)= (10.25/13)^13

 

P(10)=25.1%

P(11)=25.6%

P(12)=15.9%

P(13)=4.5%

Total=71.1%

 

Thats slightly higher than I expected, but not by much. Maybe my 10.25 guess is slightly high.

[han]

Josh, is there any relationship between your calculation and reality? I think there is none.

[joshs]

Josh, is there any relationship between your calculation and reality? I think there is none.

I don't know if

a. my estimate for the mean number of tricks is right

b. how good an approximation is the independance approximation. Its certainly not true...

I did figure that this should give us a ballpark number assuming a is approximently right.

[han]

I don't think that the main idea is right to start with. For instance, your standard deviation is only dependent on the average. This makes no sense, a good pair will have a much lower standard deviation for freely bid games than a pair of absolute beginners. In your model it is not possible to adjust this, there is no reason that it should be a reasonable model even if your estimation for the average number of tricks is correct.

[inquiry]

Well, there is an easy way to do such a calculation for online bridge if you have access to bridgebrowser. I looked at my own data for January 2006. My partner and I bid game 76 times and made 53 of them (70% success rate). If we count only imp games (main room and touneys).

 

I played game in 64 imp hands, making 47 (73%).

 

In main room, I won 4.15 imps on 21 games making, in tourney room (and team games), I won an average of 1.73 imps per game making (26 hands). In main room, in the 11 imp games that went down I earned -4.38 imps/hand, and in tourneys, the 6 hands that went down earned me -3.66 imps/hand.

 

A couple of issues, bidding game in main room was rewarded more than biddign game in the tourneys, no doubt because I play more team games so the field not getting to game is not an issue. The penalty for going down was roughly the same,

 

At 73% game bidding (bid and made), I netted a total of 61.99 imps over the total 64 hands, or just under one imp per hand in january.

 

To do this kind of analysis is very easy with bridgebrowser. First enter your name under player, then on contract, choose specify contract, and choose 3NT, and then 4H through 5NT, also choose "declarer" (so you only get hands your side plays). Then choose "use as auxiallary term". Next click the bid analysis tab and choose "use as auxiallary term", finally go back to the player tab and press search.

 

When the data shows up, a summary of your average imps and mp will be given along with number of imp game and mp game contracts. Then click on the bid analysis tab. Go to the 3NT contract table line, and right click in the number of 3NT contracts you bid, choose plot. Click on each histogram showing 9 or more tricks won. Then go to 4H and do the same (except 10 or more), etc. you will see a total of all games bid at imps and mp, and the average imps and matchpoints. Write those down, clear the window. amd repeat the process for contracts where you were defeated. Doing this for the data I presented took me well less than 3 minutes (but I am on fiber optic internet II line, so my access to the online database is blazingly fast).

 

So what we did was create a search based upon"

1) Player name

2) Game contracts only

3) Player name side declared the contract

--> The program found and listed all these game contracts very quickly and gave an aggregate average imp score and mp score without regard to rather we made or went down in these contracts.

 

Then we used the cool bid analysis tab and plotting histograms to quickly see how well we did making and going down in these game contracts separately

 

Larry Cohen said, "The possibilities for analysis are mind-boggling. ...The good news is that it’s easy to get the program running in a matter of minutes. Basic search functions can be performed without reading any documentation" and Carl Hudecek said, "This program is the greatest tool to study and research bridge that I have ever seen, or even IMAGINED. It is like a DREAM COME TRUE." And time Bourke simply said bridge browser had "the WOW factor", and all of tha twas before it became so much simplier with the plotting histogram tool which was added in 2005 (before these reviews).

 

Disclaimer. The homebase club offers free subsciption to online bridgebrowser that allows you to check the hands played exclusively in BBO's Homebase club. Homebase players can eventually (As the database grows with their plays) check this stuff on themselves. Bridgebrowser is designed to make short work of this kind of info. So my description of the program and how to use it is a little self-serving.. i want you all to sign up for the free service and see for yourself how easy and fun it is to use. Of course, the more you play in homebase events on the BBO the more valauble the tool will be to you. If you are interested in doing this stuff now, and you have played a lot on the BBO in the past, you may want to pay for a subscription to the unlimited version (and I don't get any of that money). It is worth it if this stuff interest you.

[joshs]

I don't think that the main idea is right to start with. For instance, your standard deviation is only dependent on the average. This makes no sense, a good pair will have a much lower standard deviation for freely bid games than a pair of absolute beginners. In your model it is not possible to adjust this, there is no reason that it should be a reasonable model even if your estimation for the average number of tricks is correct.

Oh I am not doing the "How will actual results vary" calculation. I am attempting the "how many tricks are there under best play" calculation. I would definitely need to introduce a variance in declarer play to the caluculation as well for the actual number of tricks.

[joshs]

Note: In theory, players who are worse at declarer play should not bid game as agressively, so if you interprete "there are 9.5-11.5 tricks available" as on average given the declarers playing skill, then

a. good players would bid game more than bad players

b. they would have the same overall probability of making a game that they bid

So the gain from being a good declarer is that you get to bid game more...

 

Of course, bad players don't really bid game that much less agressively than good players...

[sfbp]

Oh I am not doing the "How will actual results vary" calculation. I am attempting the "how many tricks are there under best play" calculation. I would definitely need to introduce a variance in declarer play to the caluculation as well for the actual number of tricks.

I think that's a moot point Josh. Peter Cheung measured that a long time ago, using about 24 million hand records (all we had at the time), and found that the actual average number of tricks taken at the table agreed pretty closely with the double-dummy result. I forget the details, but his conclusion was that the two are essentially interchangeable.

 

Stephen