[dogsbreath]

Hi Uday..
Hand records from 'live' tourneys I play recently include a 'par' score derived by the program. Is there any possibility of introducing 'par' tourneys so we score -v- computer-derived result rather than table comparisons?
Rgds, Dog

Quote

furnulum pani nolo

[mink]

Such a par should not be set by software. Instead, the tourney host would be requried to use predealt boards (already possible) and manually assign a par to each board (not yet possible), and the software would be required to use this par for calculating the IMPs (not yet possible). Of course, finding a par for a board is a matter of judgement, and it is possible that you overlook something and assign a poor par. On the other hand, IMPs calculated with well-chosen pars would enhance the meaning of the scores.

E.g. if a non-vul slam could be bid and made, you would assign 980 as the par. If in a tourney without assigned par only 1 pair of 11 bids the slam, it would get 9 IMPs for that, and the opps would get -9 IMPs though they did nothing wrong. With par assigned they would both get 0 IMPs which is more appropriate. Also fair that all those who did not bid the slam get -9 IMPs (-0.9 IMPs without par). Not so fair still that the opps of those who did not bid the slam get 9 IMPs.

If the par was assigned by software, using the average of the scores reached in the tourney, this would be Butler and inferior to the scoring method used right now.

Karl

[McBruce]

Take this deal:

Scoring: MP

---

Double dummy analysis by the computer program Jack reveals that:

In notrump, East can make 11 tricks.

In spades, East or West can make 8 tricks.

In hearts, East can make 12 tricks.

In diamonds, East can make 11 tricks.

In clubs, East can make 11 tricks.

The Par score for this deal is 980 to E-W but few will get to this poor slam. A more difficult thing to do is to find deals where the par score is acheived if everyone makes reasonable decisions, and a less than optimum decision usually leads to a sub-par result. This would be an interesting game!

[DrTodd13]

I'm not sure I'd want to see tourney results compared against par but it would be interesting for every hand that is played to determine and show the par result.

Todd

[uday]

Well, if someone has a way to determine par score on a board.....i'd try it, just for kicks. The server runs on a linux box, fwiw.

[paulhar]

McBruce, on Jul 19 2004, 04:01 AM, said:

A more difficult thing to do is to find deals where the par score is acheived if everyone makes reasonable decisions, and a less than optimum decision usually leads to a sub-par result. This would be an interesting game!

Yes, this is quite difficult. I know because I have to create 8 such deals for every lesson my wife teaches. The rule is: no bad bridge decision can be rewarded, and those bad decisions likely to be made by the students should be punished.

As the level of the class increases, this becomes a much harder task. I can typically create a beginner's hand illustrating a specified bidding and play concept in 10-15 minutes - with the contracts in the wrong strain at the correct level going down even on bad defense.

As the targeted students improve to where these simple hands are simply not appropriate anymore, it becomes a more difficult construction to punish bad bridge. This is especially true in deals where both sides are bidding, which are more interesting for the students. In some of the hands created for a higher-level intermediate class, it has sometimes taken me a couple of hours to create a hand to teach a certain concept where someone gets a worse result (or at least not a better one) for any error.

A tough thing about building these par hands is the fact that sometimes one side's error might induce the other side into a greater error that wasn't available if the first side's error hadn't happened. Ideally, you wouldn't want the first side to gain from their error. A typical example is when N/S overbid to reach an awful suit contract and East leads fourth best from AKQx to let it make! Granted, N/S were going to be well over average for getting that lead anyway, but N/S are getting a lot more than they deserve in any kind of par contest. In theory, if the hands are properly constructed, overbidding against a weak pair should pay off a lot less in this optimized par contest than it should in real life.

If it's that difficult to create such par deals for intermediate classes, then I would think it would be a total nightmare to create such deals at the difficulty level that would interest BBO members! What makes it even more difficult is that something I would call reasonable play you might call an error, and vice versa.

[McBruce]

Good post by paulhar, confirms my experience with pars computed by Jack's double-dummy engine. For awhile I listed the 'Jack Par' scores (pun intended) during my tourneys when all tables finished the board. Players for the most part enjoyed it.

I imagine the same could be done with Deep Finesse, or any double-dummy program. Players might enjoy seeing what can be done with perfect play. I always included a disclaimer that the Jack Par results were based on perfect double-dummy play and your mileage may vary.

The trouble with a par contest is that it really is virtually impossible to create a perfect deal where mistakes will always cost.

Another thing Jack can do is play the deals, single-dummy, against itself many times, using a wide variety of systems and thinking times. You can generate a set of results for any hand and hold a different type of par contest where your score is matchpointed against the scores acheived by the computer. For example, on the board above, when played 25 times by Jack, nine computer pairs made 450 and seventeen made 480. A chart could be constructed from this:

N-S -440 or better: 100%
N-S -450: 86%
N-S -460 or -470: 68%
N-S -480: 34%
N-S -490 or worse: 0%

It would be like the old Epson Pairs games. To do this on BBO we'd need a mechanism to input the deals and the pre-played results (a special type of .lin file?)
The advantage is that the movie would be able to display immediate results, even in an unclocked game. One disadvantage might be that some sets would favour N-S over E-W by random chance -- quite often the normal result scores very low or very high.

[dogsbreath]

hi ..thanks for yr interesting replies..
I concurr with many of the comments.. the 'par' score is unlikely to be achieved on many deals, but would possibly provide an interesting comparison against current tourney scoring which is arguably even more erratic.
The thought arose when i played a tourney recently where a hand arose where good play is needed to make 9 tricks in a major suit .. 5 pairs made 10 tricks including 2 who bid game.
The current scoring encourages gross overbidding because of the generally awful standard of defence ..hence the idea that a 'par' tourney, even with it's occassionally unachievable par results, might be a refreshing change from having the average determined by the worst defenders in the field
Rgds Dog

Quote

furnulum pani nolo

[paulhar]

I don't understand - how would creating a par datum change the fact that you want to overbid against weak defense? Unless the hands used are defender-proof!

Normal deals with par determined by Deep Finesse or whatever will not change the fact that the tournaments are determined by the worst defenders in the field.

In fact, the structure of tournaments is such that you must depend on poor defense to win. You will not win any cross IMPs tournaments on BBO averaging +2 IMPs per board. You may place well, but you won't win. If you win, you have to assume your opponents aren't going to play well and act accordingly. If they play well, you couldn't win this tourney anyway, better luck next time.

You want to reward playing well instead of this wild swinging necessary to win? In a 12 board X-imp tourney, only count your 5th thru 10th best IMP scores, throwing out the top 4 and the bottom 2. Then this wild swinging hoping to score a huge number of IMPs against a weak pair will cease, as your windfall will be thrown out. The steady Eddies will win. It will also throw out the speculative doubles that are necessary to win these events too. I remember subbing for the last two boards in a cross IMP event. I came in negative a couple of IMPs. On the last two boards, we played a pair that were wildly trying to catch up (they must have been down more than a couple of IMPs.) Having nothing close to a penalty double on either board, I doubled based on nothing more than the sound of their auction (they basically bid as if they were overbidding, stuff like 1NT-2H signoff-3NT) We picked up 25 IMPs on these two boards and my subbed partner went from under average to 2nd. Unfortunately, these are the types of things you have to do to win these things. You must play against weak defenders and overbid, and you have to hope others are doing the same against you so you can double 'em for a number. Look at the IMP scores required to win a tournament. They're phenominal! I tried playing steady bridge to win these things, it just doesn't work. I'll take my steady game into the Main Bridge Club, if you don't mind.

I just don't see how having a par datum is going to help. People will still be out there trying to beat up on the weak pairs, or the shooting pairs. And the wildest bidders and the awful defenders will still determine the event. What you need is some scoring method that reduces the effect of these random results.

Some tournaments, the leaders are playing against leaders. You're probably not going to get weak defense against a leader.