How good is good?

[Wayne Russell]

What would be considered a very good win rate on BBO in the Main Club? i.e. Average IMP per Board?

[Elianna]

I would say that it completely depends on who you're playing against, but that's not 100% true, it also depends on which direction the randomness happens (ie is there an insane EW or NS out there). Honestly, it seems fairly random, that it's hard to pin down a good IMP average. If you're averaging +10 per board, maybe you need to find opponents that you're not beating up on. If you're averaging -10, maybe you should find people that you can compete more easily with.

[whereagles]

At world championships, the best pairs score around 0.6 to 0.8 imps per board. So if you're winning by 10 after 12 boards, that's already VERY good :)

[hotShot]

IMP's per board is some meaningless statistical number.

 

There are 3 kind of boards:

1) unselective Boards, (almost)no chance to miss game, (almost)no missefence possible. Take a board where exactly 3 aces are missing, nobody will miss game with 28 HCP, and opps can hardly avoid to make there 3 aces.

Even if you play 3NT instead of 4M it's just 430/420 => 0 imps.

 

2) boards where you can win an overtrick or gain by playing NT instead of a minor or something like that.

Here you make an IMP per board against weaker opps, if opps are about your strength, they will make the overtricks too.

 

3) boards that really make a difference. Is it a game or not? Try for slam or not? Is this a good/bad sacrifice?

Is there a wining/loosing lead?

These boards don't come up often, but they come sooner or later. Good player wait for those boards to win the match.

 

 

If you play 8 boards a type 3 board might not come up, so in the end you might be lucky to win by 2 imps. Even top player won't score more.

 

But take a look at some vugraph final, each team match has at least 70-100+ boards. During that period some type 3 boards will come up.

But does these boards favor the better player? No they don't!

They favor bidding system and bidding style. So for one of these boards one of the pairs might have the perfect treatment in there system, another pair might have a weakness in their system for exactly this hand.

 

So if you win with a big imp per board number, it says you took a high risk and got away with it. Usually this implies that your opps where weaker, but it says nothing about how good you are.

If two world class teams fight 100 boards and one wins with 4 imps they made 0,04 IMPs per board. Does this mean they are not good players?

 

So forget the IMPs per board number, it does not mean anything.

[pdmunro]

Here is an analysis by Jeff Goldsmith:

 

http://www.gg.caltech.edu/~jeff/bridge/study

 

He comments

 

"A rough guess from experience at OKbridge tells me that

a national champion is only about 1 IMP/bd better than a

good flight A player. Flight A players are, on average,

I think, about an IMP/bd better than Flight B players and

the difference between Flight B and Flight C is also about

one IMP."

 

His site index is http://www.gg.caltech.edu/~jeff/

[hotShot]

That is a nice analysis, Peter, thank you 4 pointing that out.

 

It say that national champions are less likely to make a mistake than flight A players, who are less likely to make mistakes than those from flight B. And so on.

 

As i pointed out imps per board tell you how much weaker your opponents are, not how strong you are.

 

If you win a 100m sprint race against a 100 year old by x seconds, will you be fast enough to get world champion?

This x says nothing about how good you are.

[Free]

on BBO, if you bid and make a game, then you ALWAYS win some imps because there's almost always some pair not bidding the game, or overbidding the hand and going down. So it's not really representative, since it just depends on how many game and slam scores each side has.

 

Obviously, if you score +0 against good opps, you're doing fine :)

[Fluffy]

On the main bridge club you would need about 50 deals for the comparison to be more or less equal, if you do that, then 0.3 IMPs/board could mean you have played substantly better than your opponents.

[inquiry]

Online, even when you play well, you can do badly. The reason is clear, and based upon what Free just said. If your average opponents bid an average game against you, and make. You will lose 2 to 3 imps.This is because some yahoos will bid a hopeless slam down two, others will play in some part-score.

 

This means in a short set of 20 boards against fair opponents, the side that wins is usually the side that has the majority of the hands. Take a look at your bad sessions, and see how many -2 to -3 imps you got for what should be 0 imps in a "good game".

 

Likewise, you score will be inflated for your normal games for the same reason. Now as luck would have it, if you play long enough (say 100 boards over many days), these +2 to +3 things and minus 2 to minus 3 things average out. That is the randomness due to the field is removed. So you are down to who is better, you or your opponents. If you insist on playing against gold stars, your will be doing well to break even or be slightly plus. IF you play against beginners and consider yourself accomplished, you will be in trouble if you are not averaging +1.8/board or more.

 

Most people gravitate to their own skill level or if htey want to improve, try to play against those slighlty better than you (if they are much better than you, they will tire of playing your and look for better opponents themselves). In a field like this, +0.5 to +0.8 is very good. +1.0 is super. And +2.0 is suspecious... I suspect if you are averging +2 imps against good competition over hundreds of boards, your partner might be cheating.

 

Things like lehmans calculate ability not by raw imps, but by comparing the quality of your opponents along with imp score.

 

Ben

[hrothgar]

For what its worth... I became very interested in this topic a couple years back and spent a fair amount of time researching this and related topics.

 

I strongly recommend the following article: Selection In The Presence of Noise: The Design of Playoff Systems. Micah Adler, Peter Gemmel, Mor Harchol, Richard M. Karp, and Claire Kenyon. The article takes a number of the issues that various posters (Inquity, Free, etc.) have already noted and wraps them in a fairly concrete implementation.

 

The article doesn't provide any specific data that you can use to judge your performance on BBO. With this said and done, it does a great job explaining why your scores on BBO probably aren't a good mechanism to measure performance.

[pdmunro]

When bridge was whist, you scored a point for every odd trick, i.e. tricks in excess of your book of 6. Then Vanderbilt took a cruise from the US West Coast to Havana, Cuba through the 10-year-old Panama Canal. And he really changed the scoring: what with more points per trick (did he add the zero?), more bonuses for contracts bid and made, the idea of vulnerability, slam bonuses, etc. Now along come computers and online bridge and everyone's getting their IMPS routinely calculated. And at roughly 40 points an imp, it all seems to say that if you are an imp a board better, on average, then, on average, you are taking a trick more per board than your opponents. And we have gone full circle, back to the whistful days of yore. (But hey this argument doesn't really allow for game bonuses. Or does it?) But just to think that we are a trick better per board than the opps. To dream ... :-)

 

PS Harold Vanderbilt took his cruise in October 1925. 80th birthday coming up.

 

PPS I just checked in "Culbertson" by J Clay (1985). He reckons the birth date of Contract as 1 Nov 1925. That day, "as their ship made its way along the Panama Canal .... Vanderbilt stayed in his cabin putting the finishing touches to his new game." (page 78)

[Gerardo]

Then Vanderbilt ... really changed the scoring: what with more points per trick (did he add the zero?).

Yes. Originally first NT trick (or mayber all NT tricks?) scored 35 instead of 40.

[mikestar]

Then Vanderbilt ... really changed the scoring: what with more points per trick (did he add the zero?).

Yes. Originally first NT trick (or mayber all NT tricks?) scored 35 instead of 40.

In Vanderbuilt's original scoring table, all NT odd tricks were 35 each. A revision in the early thirties changed it to 30 for the first, third, fifth, and seventh odd tricks and 40 for the second, foiurth and sixth odd tricks. The next revision established the modern values.

[pclayton]

When I played on OKB; the general thought was: .5/ brd was good, 1.0 was really good. My Lehman was around a 60; I think my average was .5 - .7. I played against all types - idiots to world class.

 

If I remember right, to make the top 10, you had to average roughly 2.2 - 2.5 / brd over 60+ boards.

 

My last week in 1999 I had my best - 3.2/ brd. It was less to do with my ability and more to do with hands going my way and opponents tripping over their own feet.