[inquiry]

Ok, I used bridgebrowser for a simple question... How do people with OKBridge Lehman's of 60 to 65 (STRONG) do compared to those with 45 to 55 (AVERAGE), and those with 25 to 40 (WEAK)?

For this study, I forced both declarer and dummy to have an lehman's in the indicated ranges, and looked at the results for 3NT contracts. No partnerships with one strong and one average, one strong and one weak, one average and one strong, or any player with a lehman between 55 and 60 was included in this data. These excluded hands explain why, in the first image below 4,036,155 3NT contracts were found, but only 14,563 had partners where both were 60 or higher.

The short answer to this question is
Strong +1.16 imps, 57.94%MP, average tricks.. imps 9.39, Mp 9.46
Averge +0.40 imps, 52.45%MP, average tricks, imps 9.21, Mp 9.29
Weak  -0.15 imps, 45.18%MP, average tricks, imps 9.08, MP 9.01

Of course, this doesn't tell us too much as to why there is such a big difference. Let's look at full stat's starting with the strong players... Let's look at the following chart and see what it might tell us, could we say weak partnerships are 1/2 a trick worse than the strongest pairs in 3NT? Something else?

First, the left hand side with red column heading is IMP results, the right hand side is Matchpoint results. The rows are numbered 3,4,5,6,7, up to 13 (there is 0,1 and 2 also, you would have to scroll up to see those). Those are the number of tricks won in 3NT. The number in the green columns are then number of hands that won that number of tricks, the white columns in the percentage of hands that played in that contract (in this case 3NT), and the yellow columns the imp or matchpoint results. Also, the first three columns for imp are overall, the next three for undoubled contacts, then next three for doubled and redoubled contracts (grouped together). The header to the columns gives teh tota number and percent, for example at imps only 2.35% of all 3NT contracts played by strong players were doubled.

What else does this summary show? The strong players average 9.39 tricks (Std Err 0.01) this earned an average 365 score (average of vul and nonvul results), and 1.16 imps. The strong players averaged 25.62 hcp when they bid 3NT at imps and 25.71 hcp at MP. I have added red and blue squares around 3NT taking exactly 9 tricks. Note the strong players averaged 63.34% on average at MP, and 4.23 imps.



How did the "average partnerships" do?



How did the "weak" partnerships do?



A couple of things, EVEN when the weak players make exactly 9 tricks, they average very similar (but less) to that of the other pairs (3.79 imps versus 3.97 and 4.15).

There are some questions this might bring to mind. Were the strong players reaching better 3NT contracts, or just playing them better. On could use board files (find hands where strong pairs played 3NT, and then see how other players did on these same hands... that would leave you with better play than better bid). Anyway, I thought an example quantifying some aspect of the difference between abilities of strong versus weak players might be interesting. Of course, you bridgebrowser, you can limit all your searches to that for "strong players" should you want too..... rather they are strong players on BBO or OkB.

Ben

[jikl]

Some of it could be due to rightsiding the contract. This would remove some of the onus from play to bidding.

However, interesting results.

If you moved to the slam zone I think you would find even more of a discrepancy.

Sean

[Free]

Yes, and frequency of when experts actually bid game or slam will probably be even more interesting

[hrothgar]

Personally, I'd be interested in seeing you address the converse...

What type of hands do beginners play better than experts? We all know that bridge involves a fair amount of luck. It should be possible to find hands where relatively poor players score better than very good players. I suspect that many of these will boil down to sheer luck: There are two finesses available, the beginner took one the expert took another. However, there should be a lot of others where inferior play gets rewarded:

1. The expert took a safety play on a hand where the cards behaved
2. The expert played for a squeeze or an end play, the beginner took making finesse.
3. A two way finesse was available, the expert made a discovery play while the beginner made a lucky guess

The reason that I suggest focusing on this question is that I believe that the analysis will be much simplier. You could even consider this as a control for this type of analysis using BrBr...

[pigpenz]

Why do experts do better?
Easy for me they understand the basic card combinations and plan ahead before they play to trick one as to what can go wrong and what if possible to prevent it from going wrong.

they understand basic imps odds
they look at dummy and say gee what can the opps make
tend to have less bidding disasters
they count count count count

[Winstonm]

From my perspective and from experience, the area that shows the most obvious differences between expert and non-expert is defense.

[inquiry]

Winstonm, on Oct 15 2006, 08:32 AM, said:

From my perspective and from experience, the area that shows the most obvious differences between expert and non-expert is defense.

I happen to agree. You want to try to guess how much better when defending, say, 3NT? Anyway, it is as easy to ask that type of question as how well they do as declarer using bridgebrowser. See the image below.

The top half (above the red line) is the part of the player search tab. Just leave the names as *'s so you can look at all players and change the player ratings in the boxes below the names (inside the green box i made)... if you want both players to have a certain range, set both.

Then on the contract search tab (below the red line), pick the contracts you want and choose either all, declarer or defend. To see just defend, push the radio button for defend. Be sure player tab is set to auxillary term and then search.

[reisig]

I always thought it was just luck!

[inquiry]

The strong players make whoever they play against average about what the weak players averaged on their own (first table below). The weak players however, had a defense that was above "Average" in that people playing against them earned more than wha tthe average players earned in the first post...

3NT by pairs playing against STRONG PAIRS as defenders


3NT by pairs playing against WEAK PAIRS as defenders

[kenberg]

Am I reading these charts correctly?

From the first chart, first post I get that strong players (60-65) played 7,336 imp contracts of 3NT. From the first chart, most recent post, I get that 3,715 of these were against weak players (25-40). That's just over 50%.

A: Why are they doing this?

B: Even if they wish to, how are they doing this?

My OK subscription has just run out but I once played there quite a bit. In the tourneys, few playere were either as strong or as weak as needed to fit into your classification. Or at least that is the way I remeber it. It seems most hands, and so presumably most 3NT contracts, in a tourney would be played against pairs that were between your two categories.

In tourneys you play against the pair that appears. In non-tourneys, you have choices and I cannot imagine why a couple of sixties would want to play against a couple of forties, and generally the forties were insistent that their tables not be invaded by sixties. Of course these Lehman ratings were far from perfect, and so there were cases where very strong players had very marginal ratings (if a bunch of strong players played only against each other, their average rating would be around 50) but for the most part, the ratings were roughly indicative of level and players tended to segregate themselves accordingly.

The indication that two sixty pluses played 50% of their 3NT hands against two forty minuses astounds me. I must be misreading this.


Added: As Ben explains below, I am.

Ken

[inquiry]

kenberg, on Oct 16 2006, 07:13 AM, said:

Am I reading these charts correctly?

From the first chart, first post I get that strong players (60-65) played 7,336 imp contracts of 3NT. From the first chart, most recent post, I get that 3,715 of these were against weak players (25-40). That's just over 50%.

A: Why are they doing this?

B: Even if they wish to, how are they doing this?

My OK subscription has just run out but I once played there quite a bit. In the tourneys, few playere were either as strong or as weak as needed to fit into your classification. Or at least that is the way I remeber it. It seems most hands, and so presumably most 3NT contracts, in a tourney would be played against pairs that were between your two categories.

In tourneys you play against the pair that appears. In non-tourneys, you have choices and I cannot imagine why a couple of sixties would want to play against a couple of forties, and generally the forties were insistent that their tables not be invaded by sixties. Of course these Lehman ratings were far from perfect, and so there were cases where very strong players had very marginal ratings (if a bunch of strong players played only against each other, their average rating would be around 50) but for the most part, the ratings were roughly indicative of level and players tended to segregate themselves accordingly.

The indication that two sixty pluses played 50% of their 3NT hands against two forty minuses astounds me. I must be misreading this.

Ken

In the title of the images, you see that there are 23,+ million hands in the database I used (I used an Okbridge one to use the published Lehman ratings... ). I did not search the entire database. Why? Because the database is not indexed by ratings. It is indexed by player name, and by contact, and a few other options. An indexed search is fast, a non-indexed one is slow if you do it online. I only have one BBO database stored locally on my computer, which can be searched quickly.

However, if you look at the charts you will see a standard error. Once the standard error gets low, adding more results to the search doesn't change it more than a very small fraction of a point at best. Thus, I stop these searches when the numbers no longer are changing. The total number of 3NT contracts is much more than those examined in any of these searches...

So, to answer your questons...

Quote

From the first chart, first post  I get that strong players (60-65) played 7,336 imp contracts of 3NT. From the first chart, most recent post, I get that 3,715 of these were against weak players (25-40). That's just over 50%.

What the first chart shows is that of 4 million thirty six thousand and one hundred and fifty five 3NT contacts I examined (there were more in the database), partnerships were BOTH of the partners had a rating of 60+ were only 14,565 (accounting for only 0.36& of the boards). A large numbe of hands had to be searched to find ones that meet this requirement. The Lehman scale is such that the vast majority of players are stuffed between 45 and 55, I think it must me logarithmic. IF you look at the number of hands played where both partners had a rating of 45 or less, you see that I searched 3,035347 contracts to find 5723 hands played where both players were weak. This comes to 0.19% of all 3NT contracts played. Since the distibution of lehman scores is normalized, this either means that few very weak players play with other very weak players, or that stronger partnerships are more likely to bid 3NT. This is testable by seeing the percentage of ALL contacts played by very weak verus very strong players. As you see, the percentage where both players fell within the 45 to 55 range was significantly higher, as I searched only 6 hundred thousand hands to find neearly 60,000 contracts (close to 9%). Players with no rating were of course excluded, and also any player with a rating between 55 and 60 was excluded, and any partnership with only one player in any of these ranges was excluded.

IN NO CASE, did I test how STRONG players did against week players, etc. You could do that if you wanted, but the only limits I placed on the search was the stregnth of the one pair (first group of charts, the stregnth when playing 3NT, second group of charts, stregnth of the partneship Defending 3NT).

Quote

The indication that two sixty pluses played 50% of their 3NT hands against two forty minuses astounds me. I must be misreading this.

It was no doubt my bad wording. What I was trying to say was this... when two weak players play 3NT contract, they averaged winning 9.08 tricks at imps ( for -0.15 imps) and 9.01 tricks at MP (for 45.18). That was against all opponents. When two strong players defended against 3NT (no matter who they played against), their opponents averaged only 9.08 tricks at imps (for -0.10 imps) and 9.15 tricks at MP for 49.19%. These tricks won, and imps/MP scores more closely resembled the scores earned by "weak players" when they played 3NT than the normal result.

[kenberg]

Whew. It seemed impossible the way I was reading it. I'll try reading it again later. I think what you said is a match for what you have, I was confused. My eyes tend to glaze over when reading tables of data.

[kenberg]

I'll try another tack.

Looking at your last tables, in the imp column, and totaling the entries for 8 or fewer tricks, it seems (if this time I am reading right) that weak players beat 3NT 27% of the time and strong players beat 3NT 34% of the time (at imps). Making the reasonable assumption that stronger players are defending (usually) aganst stronger opponents (ie, the selections of opponents are not random), we come to the (hardly startling) conclusion that strong players defend (a lot) better than weaker players. Or maybe their strong opponents bid too many 3NT games.


I think it's tough to extract meat from the data, but tough and impossible are not the same. Or at least not necessarily the same.

[pigpenz]

basically you are looking at real time data from play, so the results are not double dummy etc. If we want to just look at the data from the cards or possible combinations of cards that the opps or partner may have then we move to a deal generator with double dummy analysis.

Bridgebrowser is just what it says, it lets us analyze data from hands that are actually played. if we want to take it a step we can put parts of the hands in a deal generator and take it a step further.

so why does the higher ranked player go against the odds etc. at certain times. The human element will never be in any of these programs unless it is programmed in as a random bug.

[mikeh]

pigpenz, on Oct 17 2006, 11:37 AM, said:

so why does the higher ranked player go against the odds etc. at certain times. The human element will never be in any of these programs unless it is programmed in as a random bug.

Even the strongest players develop blind spots on some hands, and (especially in long tournaments) fatigue causes errors. I am not sure how one could programme a computer to simulate these errors: perhaps a random error rate that increased with the number of hands played within a specified time?

As for going against the odds, a strong player may do so for any one of a number of reasons:

1. fatigue
2. loss of concentration
3. table action (an opp's break in tempo, body language)
4. need to create a swing
5. Especially in on-line, and especially (from my now stale-dated OKBridge experience) from cheating: some of the winning plays I witnessed back then had to be seen to be believed: from players with Lehmans in the 70's...'strong' players by Lehman standards but not in real life.

Furthermore it depends on who is doing the calculation of the odds. One attribute of a strong player is that he or she will be filtering many more factors than would a lesser player.

I once heard that Terence Reese said something to the effect that the average player not only doesn't know what an expert thinks about at the table, but wouldn't believe it if told. That accords with my experience: dummy hits and you appear to have 9 or 10 tricks off the top in 3N and you go into the tank at imps, wondering how to cater to really bad breaks... which do not materialize. Average players think that you are wasting their time.... and can make some pointed remarks. Experts recognize what you are doing and never say anything.

[inquiry]

I have posted an example of how to beat data out of BridgeBrowser that is not too painful (and shows more features of the software) in response to a request from someone who has full access to bridgebrowser. Among the people with full and free access to the full online bridgebrowser are all BBO yellows, and all of homebase directors.... so they at least will find that post of interest, as I discuss why you might choose one search option over the other, and some thought processes that go into building different searches. I choose to post it on the homebase page because most here might not be interested (the few of you who are still considering becoming homebase directors, might want to take a look, however)...

The link is... Searching for auction.. 1m-1M how often it i1M 5 cards?