Data on frequency of slams/games/partscores; and what it means
dmo...@webone.com.au
> Has OKBridge or BBO published a histogram? There have
> been millions of hands played, perhaps billions. What
> percent are slam hands, game hands, and part-score
> hands.
> My guess is less than 10% slams, less than 30% games
> and the rest are part-scores.
As this the original thread was denigrating into invective . . .
That guess is way out, at least for IMPs play between teams, and the
conclusions drawn from them are consequently misplaced.
I have compiled data based on almost 3500 boards played in World
Championship finals and semi finals, finals of major US and Australian
tournaments, and high-quality events in which I participated. All
were teams events; all were scored at IMPs (but two of the old WC
finals were played under the old IMPs scale which I converted to the
current one). The events are from the 1950s, 1960s, 1970s, 1980s,
1990s and 2000s; the data from the 1990s and 2000s includes a small
number of women's events.
Surprisingly, there was little variation between the data from
different decades, notwithstanding that a number of the modern events
include Meckwell and other modern aggressors.
Of these 3419 boards:
-- 10.4% were played in slam at one or both tables
-- 59.6% were played in game at one or both tables (and not in slam at
either table)
-- 29.8% were played in partscores at both tables
-- 7 boards were passed in at both tables
-- 19.6% of the total IMPs turned over occurred on deals where at
least one team bid slam
-- 61.3% of the total IMPs turned over occurred on deals where at
least one team bid game and neither team bid slam
-- 19.1% of the total IMPs turned over occurred on deals where both
teams bid partscores
-- 62.6% of boards involved competition (defined as calls other than
pass and a double of the final contract) at one or both tables
-- 66.4% of the total IMPs turned over occurred on boards involving
competition at one or both tables
For system designers and thinking bidders this data contains much food
for thought.
David
danny...@gmail.com
Great post. Thanks for that info. A tiny quibble, I wouldn't
consider hands with a double of an artificial call as 'competitive'.
Some statistics I'd like to see is what happens when one team opens
the bidding at both tables, or buys the contract at both tables.
Also, choice of games would be a great area for information;
Especially 3NT versus 4 of a major.
Danny
dak...@aol.com
> different decades, notwithstanding that a number of the modern events
> include Meckwell and other modern aggressors.
Not surprising to as a dominating majority has little changed in
bidding philosophy. The effects of those seeing this and applying it
is minuscule.
> Of these 3419 boards:
> -- 10.4% were played in slam at one or both tables
> -- 59.6% were played in game at one or both tables (and not in slam at
> either table)
> -- 29.8% were played in partscores at both tables
> -- 7 boards were passed in at both tables
Closely jibing with my analysis: 60 game (30 us/30 them); 3+slam each
way(another 3 explore slam); 40% not game.
Thanks for the figures from another direction.
-- 19.6% of the total IMPs turned over occurred on deals where at
> least one team bid slam
> -- 61.3% of the total IMPs turned over occurred on deals where at
> least one team bid game and neither team bid slam
> -- 19.1% of the total IMPs turned over occurred on deals where both
> teams bid partscores
> -- 62.6% of boards involved competition (defined as calls other than
> pass and a double of the final contract) at one or both tables
> -- 66.4% of the total IMPs turned over occurred on boards involving
> competition at one or both tables
>
> For system designers and thinking bidders this data contains much food
> for thought.
>
> David
Exactly where my research intends. You from the top; i from individual
modules developing expected IMP.
Again thanks.
This research is intimate to the goal (thus meaning) of bids. I
applaude you.
Larry
Would be good to compare to simulations (was it 6% slams and 60%
games?).
I would like to know about significant IMP swings:
5+ in partials, 7+ in games, and 11+ in slams in order to redesign my
various systems.
My hypothesis is that partials are more important then we give them
credit for, but good slam bidding is the best choice for system design
@ IMPs.
Larry
Lorne
news:63da8cd0-c5b9-4253...@q24g2000prf.googlegroups.com...
Having established that one of the most interesting areas is where one team
bid game and the other did not is there any chance of exploring that
further?
ie did the team bidding game win or lose over the team not bidding it and by
how much - preferably split by decade. I am interested to know if the
current trend of bidding game whenever there is a remote chance it might
make is showing a profit or if picking up lots of 5-7 IMP swings from stying
in a part score on marginal hands is the winner.
vsp...@hotmail.com
> <dmor...@webone.com.au> wrote in message
A further breakdown of this group would be interesting.
*game at both tables.
*game at only one table.
Is the aggressive pair(the one in game) winning or losing imps?
thanks,
jogs
bboin...@gmail.com
are more than 300 million hands in bridgebrowser from both okbridge
and bbo. The okbridge data is no longer being added and hasn't been
for quite some time. I did not look at all the hands, instead choosing
17,124 hands played in jimmy cayne matches from june 30, 2005 until
Feb 2, 2008. I choose this range because i had those hands is a single
file, easy and quick to access on my local drive.
Cont times Percentage
P 118 0.69
1C 22 0.13
1D 31 0.18
1H 53 0.31
1S 134 0.78
1NT 998 5.83
2C 188 1.1
2D 356 2.08
2H 705 4.12
2S 1006 5.87
2NT 331 1.93
3C 524 3.06
3D 578 3.38
3H 629 3.67
3S 606 3.54
3NT 3046 17.79
4C 142 0.83
4D 179 1.05
4H 2216 12.94
4S 2599 15.18
4NT 62 0.36
5C 433 2.53
5D 475 2.77
5H 247 1.44
5S 192 1.12
5NT 2 0.01
6C 199 1.16
6D 216 1.26
6H 278 1.62
6S 295 1.72
6NT 95 0.55
7C 33 0.19
7D 23 0.13
7H 44 0.26
7S 49 0.29
7NT 20 0.12
17124
Games 54.14
Slams (all) 7.3
Small slams 6.31
Grand slams 0.99
That was for all contract, here are the numbers for opening side and
non-opening side...
Cntr Opener Overcaller
Pass 118 0
1C 22 0
1D 28 3
1H 39 14
1S 85 49
1NT 862 136
2C 139 49
2D 262 94
2H 502 203
2S 674 332
2NT 259 72
3C 325 199
3D 349 229
3H 372 257
3S 348 258
3NT 2627 419
4C 66 76
4D 104 75
4H 1726 490
4S 1981 618
4NT 55 7
5C 274 159
5D 322 153
5H 152 95
5S 117 75
5NT 2 0
6C 175 24
6D 176 40
6H 250 28
6S 246 49
6NT 93 2
7C 26 7
7D 20 3
7H 37 7
7S 46 3
7NT 20 0
Bridgebrowser can do a lot of other interesting things, for instance,
here is the results based on TEAM hcp for 4S contracts played by the
opening side....
HCP Hands Average imps
13 0
14 3 -0.67
15 11 -1.36
16 14 2.5
17 20 0.05
18 42 -2.57
19 54 -1.2
20 83 0.25
21 141 0.16
22 189 0.66
23 270 -0.14
24 304 0.33
25 276 -0.09
26 209 -0.88
27 151 -0.12
28 121 -0.63
29 53 -0.38
30 26 -1.69
31 10 -5.5
32 3 8
33 0
vsp...@hotmail.com
Let's see if I understand these stats.
28-31 Does this mean Jimmy Cayne's pair holds
between 28 to 31 points on this board?
And Cayne lost imps whenever his pair was in
slam range?
TIA
jogs
bboin...@gmail.com
I did not break the hands down to jimmy's being the opening side, so
you can't draw that conclusion directly. This was simply by final
contract no matter which side opened. I should note these were all 17k
hands played AT JEC's table, as that is the way i built this
particular database. This means that when 4S was bid by the opening
side 53 times when the combined partnership holding was 29 hcp (for
example), and this contract averaged minus 0.63 imps...
For example, here are the three hands where 4S was played by the
opening side with a combined partnership holding of 32 hcp. Only once
was Jimmy's side the declarer (they won 12 imps as amazingly 3S was
the contract at the other table), and twice he was on defense. On one
of the hands were jimmy was on defense, there was a 13 imp swing for
4S -- so jimmy won 12, lost 13, and won one (on defense on the other),
but the declarer's side was +12, +13, -1 = +24/3 = 8 imps/bd on
average.
BTW, of the 17,124 auctions, 9058 were uncontested (that includes the
118 passouts). This is the major difference with what someone posted
above that 63% hands involved competition.
Richard Pavlicek
> ... 17,124 hands played in Jimmy Cayne matches
> from June 30, 2005 until Feb 2, 2008 ...
Bear Stearns inquiry wrote:
Now you tell me!
--
RP
Jürgen R.
"Richard Pavlicek" <ric...@rpbridge.net> schrieb im Newsbeitrag
news:5KidnVTjO4cdz_fV...@comcast.com...
If a hand takes 8 minutes that's an average of 114 hours per month
during these 2.5 years, almost 4h per day. Probably left him little
time for other addictions.
>
>
>
henry...@yahoo.com
> "Richard Pavlicek" <rich...@rpbridge.net> schrieb im Newsbeitragnews:5KidnVTjO4cdz_fV...@comcast.com...
>
> > BBO inquiry wrote:
>
> >> ... 17,124 hands played in Jimmy Cayne matches
> >> from June 30, 2005 until Feb 2, 2008 ...
>
> > Bear Stearns inquiry wrote:
> > Now you tell me!
>
> > --
> > RP
>
> If a hand takes 8 minutes that's an average of 114 hours per month
> during these 2.5 years, almost 4h per day. Probably left him little
> time for other addictions.
>
>
8 minutes per hand in an online context?
That's a bit of an overbid.
I haven't watched Cayne a lot, but my guess is that with fast
claiming, etc., they probably easily fit 2-3 hands into an 8-10 minute
stretch.
For example, Steven and I have had the good fortune to play set with
fast players on BBO a half dozen times a month. We can typically fit
a 28-32 board session into 2-ish hours. That would be about 3.75 -
4.28 minutes per board, so say around 4 minutes per hand.
Henrysun909
vsp...@hotmail.com
Been studying this sample of 17124 hands.
The intervenors(overcaller's side) have played 24.7%
of the hands. Intervenors were in game of higher
over 50% of the time when they declared. This looks
and feels much too frequent.
Seems like they must be overbidding.
1. Intervenors in game. Net plus or minus imps?
2. Intervenors in slam. Net plus or minus imps?
Ran my own monte carlo. Opener has 11+ HCPs.
Opponents have 20+ about 33% of the time.
Opponents have 25+ about 4%.
In JEC's games intervenors are bidding game about
13%. That's far too often. Are those games bids
bad boards? Are they losing imps?
Disclaimer. My monte carlo does not exactly
duplicate game conditions. Ignores preempts by
opener.
jogs
Andrew
Not necessarily.
1. Some fairly significant percentage of the non-openers game bids
were intentional overbids (the sacrifices).
2. After the bidding has been opened by an opponent, if your side has
a good fit, then the HCP requirements to make game decline
significantly, so the 4% number (hands where the non-openers have 25
HCP) significantly underestimates the number of hands where the non-
opening side can make a game.
Andrew
dmo...@webone.com.au
> BTW, of the 17,124 auctions, 9058 were uncontested (that includes the
> 118 passouts). This is the major difference with what someone posted
> above that 63% hands involved competition.
If I've understood your data correctly the explanation is simple: the
data I posted comes from an analysis of auctions at two tables for
each deal. As I noted, the figure for the frequency of competition
(and of slam and game) were when the opponents competed (opening side
bid slam/game) at one or both tables. A fair proportion (I didn't
record that data) involved only competition at only one table (i.e.
the opponents were silent when the deal was played in the other
room). The data I recorded is similar to that presented by John
Boeder in his excellent book "Thinking About IMPs" (pp.16-17),
although he differentiates between light and heavy competition and has
a smaller sample.
I should add that I stopped recording the data when it became obvious
that computerised analysis of records would be much quicker and would
enable users to mine the data for the answers to a myriad of
interesting questions. Many of them I was also interested in but
there's a limit to how much information one can manually record for
each deal.
David
OldPalooka
>
> - Show quoted text -
You are consistent in representing game requires close to 25HCP in
your recent postings supporting your notion that accurate partial
bidding is enough to dethrone Meckwell.
You are ignoring the effects of good fits [working short suits and
lots of trumps] where 20HCP is often more than enough.
Game on a balanced 22-23 is often reasonable when you can place all
the cards. If you are playing against a sound opening bidder and
responder shows up with a queen or more, you are on the border of
double dummy land.
You don't have to make game to show a profit. I suspect a sizable
proportion of the games reached by intervenors are saves.
The Meckwell effect [although not really original to them]: if you
push to game after game after game after game, you wear down the opps
faster than you are worn down yourself.
In the specific case of overcalling at the 2 level in a minor at IMPs,
the goal is to end up in 3NT, not 3C. Hence overcalls are based on
good 6 card suits and sound values so partner can bid game on a
stopper and a smattering.
You can make the case that bidding game at matchpoints should not be
so aggressive [in theory]. I believe all experts know this, and they
also know that most opps are palookas, so game may be odds-on when it
ain't really. One of the the biggest factors in expert bidding at any
form of the game is not what you hold, but who you hold it against. I
know in my case I have lost many more matchpoints from overestimating
the opps than underestimating them.
-- Bill Shutts
vsp...@hotmail.com
There should be no preconceived biased notions
for the results. We should not try to subjectively
rationalize the data.
Trying to analyze the data as objectively as
possible.
The finals of the last reported world championship
produced a swing of over 500 imps over only 128
boards. In theory the two best teams in the world
are competing. Shouldn't one expect more flatter
results, more pushes. At over 4 imps per board
the bridge world is far from achieving optimal
bidding accuracy.
In competitive auctions most pairs allow leeway
for entering an auction. But most systems make
finding the right strain difficult. Bids by
advancer are forcing or constructive. Momentum
forces the intervenors too high. At least that's
what the data is showing.
It will require more detailed analysis of the
data to prove or disprove this assertion.
13% of the time the intervenors bid game on at
least one table. These numbers must to broken
down further.
a* game at both tables by the same direction.
b* game at only one table.
c* game at both tables by the same team.
On b* which team is winning the net imps? The
team in game or the other team.
vsp...@hotmail.com
I am well aware that fit games can be made with 19+ points.
But the test of data must start with some controlled basis.
The statement was 4% of the time the intervenors hold 25+
points. Yet 13% of the time game was bid at one or the
other table.
My question was were the aggressive pairs bidding the
games winning or losing imps?
At no time did I claim a pair could only make game with
25 or more points.
jogs
OldPalooka
You posted last week
"32+ to 8- ; less [than] 2%
Less than 2% of boards are in the slam range
based on power alone.
"25-31; about 33% are in game range.
"16-24; about 65% in this range.
"That means a large percentage of boards are
partscore battles. So why are most bidding
systems designed for reaching games and slams?
They should be trying for the best strain."
[presumably in a partial seems to be implied??]
If you look carefully at your distribution you will notice that if
only 1/4 of the 16-24 hands should be bid to game, then at least 50%
of all hands should be bid to game. Also note that a 40% double dummy
play for game is almost surely good enough to have a marginal positive
expectation not vul at IMPs, so I suspect 50% is low.
In this thread you note intervenors hold 25+ about 1 in 25 yet bid
game 1 in 8. You make the vague assertion that is *far too often*,
but only propose the analysis to support this, so at this point it is
just an unsupported assertion. Ok, *far too often* could mean in your
mind that 11% is just right, but I infer you are radical and believe
perhaps 8% is the mark. It is difficult to judge from qualitative
assertions.
You also made a Kaplanesque editorial remark that high variance bridge
is somehow not deserving of world championship competition to support
your assertion. Except your notion of high variance [500 IMPs in 128
boards] is almost exactly on Kaplan's definition of normal variance
championship bridge: if you could hold your losses below 2 IMPs per
board he believed you would win almost any long match. Note that in
the USBC finals, with Meckwell - the poster children of high variance
bridge - playing every session, the Nickell team gave up a
phenominally low number of IMPs, particularly on the last day.
There are a number of other reasons that more games are bid now than
in the past. Opponents preempt more often with high variance 2 bids,
2 suited preempts, Bergen responses, etc.: the most effective defense
has proved to be get to the right strain even if you are beyond the
comfort zone. Methods (fit jumps, mini-splinters, etc.) for finding
critical fits have improved dramatically. Statistical analysis denies
or severely limits the utility of game tries. The understanding that
blasting a big fit game and going set can show a bigger profit than it
used to because they are more aggressive game bidders too.
So carry on with your quest to prove experts bid too many low quality
games and not enough good part scores. I continue to think so too,
but I don't believe we are on the same continent in our assessment of
how often, unless you are unintentionally painting a darker picture
than you mean. But confound it, they make too many, and against
players too, not just palookas. It is like the tale of the man who
tells the shrink "My brother believes he is a chicken." The shrink
states "Bring him to me, I will cure him". The man replies "Out of
the question, we need the eggs!" So the expert complains too, but "he
needs the IMPs!"
-- Bill Shutts
vsp...@hotmail.com
>
> read more »
-
You posted last week
"32+ to 8- ; less [than] 2%
Less than 2% of boards are in the slam range
based on power alone.
"25-31; about 33% are in game range.
"16-24; about 65% in this range.
-
I posted based on power alone. Was I required to
repeat that statement on every sentence?
Haven't drawn any conclusions. I'm only posing
questions and suggesting hypothesis for testing.
Waiting for the empirical evidence.
jogs
OldPalooka
No, but perhaps you should identify the difference between a working
hypothesis and a conclusion. I thought your tone was pretty
consistently conclusive over both threads. I am pleased to find I was
wrong, because I also have some ideas about how to design the
experiments.
I don't really believe you can get conclusive 'objective' evidence.
Even if you find that bidding game on these N deals is a losing
proposition compared to the other table, it does not mean that game
was not a good proposition on the available evidence, so you will
still have to do at least a sampling of case by case subjective
analysis to understand the significance of the statistics.
-- Bill Shutts
vsp...@hotmail.com
There'll be between 1000 to 1500 boards where game was
bid at one table and not the another. If the imp difference
is small between the game bidders and
non game bidders, we can not draw any strong
conclusions.
But what if the difference is large. Then we'll
know if aggressive game bidding is right or wrong.
jogs