Idea for MBT's

[pclayton]

I understand that to help prevent cheating, the boards in MBT's are completely random.

 

Sometimes you get a parade of 16 counts, and sometimes you get nothing but drek. While this makes it cheat-proof, it also makes the boards very boring and randomizes the winner. Many of us play duplicate to avoid this randomness. I was leading a MBT a few minutes ago, and then watched my GIB opponents score a vulnerable slam, and 2 games while I held 4 counts.

 

One idea would be to make each hand have the same number of points as the other tables. Board 1 could be a 23 count with the player and the player's GIB, Board 2 could be 17, Board 3 could be 29 points, etc.. However, collusion could occur with this setup as a player could know from another in advance how many points are in a future deal.

 

Perhaps a better idea is to always have the "same" number of hcp per deal in a given set. However, players could manipulate their score by the number of boards they play. If the set is programmed for all the NS hands to hold 16 counts for a set, a player could endeavor to play as few boards as possible. Making all of the NS hands greater than 20 incentivizes a player to play as fast as possible.

 

What about a MBT that has every single hand contain exactly 20 points in the EW and NS hands? While you would know instantly the GIB's point count across from you, the tourney would still be random, and the cheating element wouldn't exist. More importantly, all of the players would be on the same level playing field as far as potential.

 

I think you'd have a lot of low level part score skirmishes, and this is good test of skill for the participants.

[TrialBid]

It is easy to understand the frustration behind this line of inquiry. Is there some kind of strategy that would stay cheat-proof without creating a radically different game? I'm not entirely sure. 20HCP to a side is cheat-proof but is a very strange game. There've been some all 10pt hands goulash tourneys on BBO, for example, that really weird me out.

 

Here's a strategy that might be good enough:

 

- Everyone's first deal is random.

- From the 2nd deal forward, over a moving window of, say, up to five deals, your side's point expectation would be boosted or diminished by the previous hands. If it is a log type of weighting, hands of all possible values could occur on any deal, only the relative likelihoods based on HCP expectation would change.

- If you've done badly with good cards, you'd be worse off than you are now. But that sounds OK to me.

- If you have had a run of bad cards, you'd have greater hope of a couple of good hands to offset your losses.

 

There are two things, though, that would help the MBT experience a lot that might be easier to implement:

 

1- Have longer MBTs. More hands, the luck evens out. Has to be trivial to implement.

 

2- When you and your GIB win the declaration, the human will always declare. Why would that be hard?

 

I've corresponded with Fred a few times about GIB in MBTs and I've repeatedly been assured that the declaration improvement will eventually come.

 

That is so obvious a need for a true test of skill. Moreover, it will take some of the CPU load off BBO servers--a resource I hear frequently cited as crucial for delivering the GIB-based services.

 

When Fred comes back victorious from the Team Trials :), perhaps MBTs will get another look.

[hrothgar]

Hi Phil

 

As Bergen famously said... "Points Schmoints". Lets assume that you constrain the number of points. Next thing you know, folks are going to start raising all sorts of objections about distribution... "N/S kept getting hands with Spade fits. We had Club fits so there was nothing that we could do"

 

Furthermore, I think that it would be a big mistake... The more constraints that you place on the system, the easier the system is to crack. I think that the MBT should be a test of bridge skill, not a contest to see who can develop the best expert system to determine the expected strength bias on deal number 6...

[pclayton]

Hi Phil

 

As Bergen famously said...  "Points Schmoints".  Lets assume that you constrain the number of points.  Next thing you know, folks are going to start raising all sorts of objections about distribution...  "N/S kept getting hands with Spade fits.  We had Club fits so there was nothing that we could do"

 

Furthermore, I think that it would be a big mistake...  The more constraints that you place on the system, the easier the system is to crack.  I think that the MBT should be a test of bridge skill, not a contest to see who can develop the best expert system to determine the expected strength bias on deal number 6...

Hi Richard:

 

Yes I've considered the distribution aspect of 20 points only sets. But just because you restrict one parameter (points), that its required to restrict other parameters.

 

I've just been on the losing side of too many game / game / slam outcomes and have zero chance of coming out ahead.

 

These MBTs need to be tests of skill, not, er, lotteries. Isn't that how BBO stays in fair territory as far as not becoming a 'gambling' site?

 

Tell me how you'd 'crack' something like this?

[mike777]

I have no idea if money bridge should be:

1) random

2) test of skill

3) lottery

 

I just think these are three seperate outcomes.

[Guest Jlall]

To put it bluntly, if the bad players never won they would stop playing. I'm sure BBO wants a high element of chance to be involved so anyone can win if they get lucky enough, but in the long term the better players are still winning (assuming MBTs are even beatable, I haven't played them in a while but I remember thinking the rake was too high). I think MBTs are the perfect gambling game... super fast paced with high volatility.

[hrothgar]

Tell me how you'd 'crack' something like this?

Lets assume that I had reason to suspect that Fred and Uday were playing games with the random number generators in order to deliberately even out the number of HCPs allocated to North / South and East / West. In theory, I might suspect this because BBO decided to announce that they were going to make the MBT more fair. Alternatively, its pretty easy to design a test to determine whether the distribution of HCP's is statistically different from those created by a fair random number generator. (If you are seriously interested in the subject, take a look at the tests that the manufacturers of deal generators use to validate that their product isn't biased).

 

Once you've determined that the system is biased, the next step is to try to determine what algorithm is being used to smooth the point total. To start with, you need to play a significant number of boards at the MBT and save the point totals. After that, you need to try to determine what kind of lag is built into the system and what kind of smoothing function is being used. Given the MBT tournaments are pretty short, there aren't all that many combinations to test.

 

There are a lot of different programs that you can use for this. MatLab, Octave, or SAS are all obvious choices...

[jtfanclub]

These MBTs need to be tests of skill, not, er, lotteries. Isn't that how BBO stays in fair territory as far as not becoming a 'gambling' site?

Think of Blackjack. Even if you count perfectly, you only win something like 51% of the time. And yet, it's considered a game of skill, and you can make a living at it.

 

If you play MBTs enough, it stops being luck (or rather, the luck cancels out) and it becomes skill. Skill at guessing what GIB's going to do next, IMHO, but what can I say?

[pclayton]

These MBTs need to be tests of skill, not, er, lotteries.  Isn't that how BBO stays in fair territory as far as not becoming a 'gambling' site?

Think of Blackjack. Even if you count perfectly, you only win something like 51% of the time. And yet, it's considered a game of skill, and you can make a living at it.

 

If you play MBTs enough, it stops being luck (or rather, the luck cancels out) and it becomes skill. Skill at guessing what GIB's going to do next, IMHO, but what can I say?

If thats the case, the 20% rake is guranteed to be a killer.

 

It would be nice if enough randomness was taken out so that a good player can have a positive EV.

 

If poor players want a crap shoot so that they will win occasionally, thats fine too. I think the distributional aspect might add this in.