[PhilKing]

Cyberyeti, on 2013-February-27, 03:36, said:

The biggest loss in my slam bidding is where we investigate a slam, don't bid it and go off in 4N/5M/5m, this is almost impossible to catch by this sort of analysis.

I have some large data files where I can perform some searches, but not that one, so I did it by hand.

I put one hand where one pair bid slam (failing) and one pair bid 4NT (failing) under "both failed in slam", but I may have missed one or two at most. I might have more categories in a larger search.

[jogs]

If this were coin tossing, 45 successes out of 66
attempts would be statistically significant. Case
closed. Bidding theory is evolving. Over 17 years
each observation isn't taken from the same
controlled population.
In 1996 the Spingold was a competition mostly
with Americans. Today it is an international
event. The field is much stronger than ever.
In 1996 online bridge was in its infancy. Americans
found that younger, foreign players are more willing
to interfere with their auctions. And it isn't easy to
penalize them. Today auctions are contested nearly
50% of the time.
Opponents even interfere with slam auctions.
Over the last 17 years slam bidding has probably
improved during uncontested auctions. Only
today many more slam auctions are contested.
There is not one homogeneous population. There
are two separate populations. One for uncontested
auctions. And another for contested auctions. In
today's game a ever greater number of observations
are from the contested auctions. In this second
population slam bidding is more precarious and
less precise.

[Zelandakh]

You did actually read what Phil wrote, yes? Especially the parts about separate statistics for highly contested auctions.

[billw55]

Another interesting data mine would be deals where both sides bid only game, but 12 tricks were available. That would be hard to search for though. Each deal would need analyzed for things like off two aces that defenders didn't cash. Also there may be deals that were favorites for 12 tricks, but declarer took only 10 or 11 on a safety play, since he bid only game ... etc. Stuff like this matters but is very hard to find in searches.

[billw55]

jogs, on 2013-February-26, 09:48, said:

Skewed patterns makes slams possible about one out of 40 boards.

I would like to know your source for this since it seems so intuitively wrong. I have played many 27 board sessions and I really think par is more than one slam per session. Such perception is not always right though, so if you have data I would be interested to see it.

Just this week at a club game, slam was cold on three deals, on a fourth only one lead could beat it, and on a fifth could be made on a strip squeeze but objectively should not be bid (again one lead could beat it). Yes this was somewhat unusual, but it would take several sessions of no slams at all to bring this near your figure .. and those don't seem to happen so often.

[Zelandakh]

Bill, you do realise that 98.3% of internet statistics are made up on the spur of the moment? Do you honestly expect this statement to be any more reliable than any of the others that this writer has provided over the last year or so?

[barmar]

Zelandakh, on 2013-February-28, 09:55, said:

Bill, you do realise that 98.3% of internet statistics are made up on the spur of the moment?

Do you have a cite for that?

[PhilKing]

Zelandakh, on 2013-February-28, 08:59, said:

You did actually read what Phil wrote, yes? Especially the parts about separate statistics for highly contested auctions.

My definition of highly contested was rather strict - it only included hands where the opponents bid up to the five level. Having to counter a preempt or overcall plus a raise was just considered as "normal", and the stats suggest that it was not that big of a deal, although obviously competing vigorously on hands where the opponents can make slam would tend to be profitable, but there is no evidence whatsoever that young Europeans have started a trend in this regarded. I may separate them out at some stage.

[Cyberyeti]

billw55, on 2013-February-28, 09:23, said:

I would like to know your source for this since it seems so intuitively wrong. I have played many 27 board sessions and I really think par is more than one slam per session. Such perception is not always right though, so if you have data I would be interested to see it.

Just this week at a club game, slam was cold on three deals, on a fourth only one lead could beat it, and on a fifth could be made on a strip squeeze but objectively should not be bid (again one lead could beat it). Yes this was somewhat unusual, but it would take several sessions of no slams at all to bring this near your figure .. and those don't seem to happen so often.

I had a look by hand at the last 4 sessions at my local club (106 boards, computer dealt)

3 grands 2 of which only made from one side due to ruff

5 good small slams that make, one only good (and only making) from one side
2 plausible small slams 35-55% that make
6 plausible small slams that don't make
3 bad slams that can be made

Will look more later, but 1/40 seems way out.

Edit - Done the next 4 sessions 102 boards.

1 grand
1 good small that makes, grand is bad but makes
6 good smalls that make, one of them only from one side
1 good small that fails
1 plausible small slams 35-55% that make
4 plausible small slams that don't make
1 plausible small slam that makes double dummy but fails single dummy

So 1 in 40 looking way off.

[WellSpyder]

Cyberyeti, on 2013-February-28, 10:32, said:

I had a look by hand at the last 4 sessions at my local club (116 boards, computer dealt)

3 grands 2 of which only made from one side due to ruff

5 good small slams that make, one only good (and only making) from one side
2 plausible small slams 35-55% that make
6 plausible small slams that don't make
3 bad slams that can be made

Will look more later, but 1/40 seems way out.

Seems much more in line with the (no doubt made-up) stat I heard a long long time ago that one in 8 deals offer someone a more or less reasonable play for slam.

[Cyberyeti]

barmar, on 2013-February-28, 10:19, said:

Do you have a cite for that?

Comes from the same place that states that 85.7% of all dwarves are not happy.

[phoenix214]

I remember that last weekend in Latvia's open pair Imps championship there were a few interesting slams. One was with a 9 card spade fit, missing the king and an ace. Two more were making because something like 2 finesses worked. Our opps bid one of those even on a combined 29 point count. It made, although they played badly and we got about 27 imps during the next 3 boards against them

[PhilKing]

I've cleaned up, checked and cross-referenced the stats for the Spingold hands (1996-2012, available finals and semi finals).

The rawest figures:

Total hands: just under 1696 (there are a few hands missing or with incomplete auctions).
Slam was bid on 162 boards.
Slam was bid and made at at least one table on 126 boards. (one in 13.46 deals) That translates to slightly more than one time in forty.

Efficiency over time:

In the 52 sets included from 1996 to 2005 there were 36 failing slam ventures. In the 54 sets I have for 2006 onwards (approximately 20 Mercury years), 24 slams failed (more semi final records are available in the more recent years hence the shorter time-frame). In the first period there were 85 successful slam ventures, whilst from 2006 there were 100 (there is an apparent discrepancy, but that is because I count playing Six when the other side makes grand as a "fail"). Slam bidding appears to have improved a lot. Caveat: there were a lot lucky making slams in recent years which have skewed things slightly.

Competitive auctions:

140 out of the 324 auctions where at least one table bid slam were competitive (24 highly so, with the villains getting to at least the five level). 86 were contested in the first period and a mere 56 from 2006 onwards. Auctions are less competitive. It's not rocket science to suggest that this may be a key reason for the improved performance, but it is far from the whole story. I am still analysing some of the figures, but slam bidding was not that much less accurate on the contested hands.

Tough Europeans:

Few of the regular European semi finalists and finalists joust around. Helgemo/Helness, Fantunes, and Lauria/Versace (all recent regulars) just bid every game in sight and then try to make them.Top Europeans do not preempt all that much. For that matter, few of the top Americans are that wild either. The highest percentage of contested auctions was in 2005 when the semi final I covered was Jacobs v Ekeblad and the final was Ekeblad v Carmichael - basically all-American affairs.

I have everything in a word file, which I can email (6 columns with the hands all referenced) if anyone is interested. As I said before, I would want to have nearer to 10k hands from the top events to draw firm conclusions, but a few things are already very clear.