[thepossum]

Hi

After more than 9 months playing and observing GiB in General IMP play and IMP and MP tournaments Im getting a little curious about something.

Everyone involved in statistical modelling knows that fake variance or perterbations are added to many simulations in order to ensure that simulated means and variances are not biased.

It crosses my mind observing many groups of tables that artifical variance is being added in order to give more variance in scores

While over time and many tournaments, hands and simulations these things average out and give unbiased results it cn be somewhat unfair to those on the receiving end and somewhat lucky to those who are the recipients

Finally everyone involved in statistics knows that in order to get unbiased results in simulations require so many simulations or replicationsthat it is essentially impossible to get statistically unbiased results. The estimators may be unbiased but there is almost certainly bias in the actual mean and variances obtained

So Im asking if these methods are used in simulations, IMP general hands and IMP/MP tournaments

Other users have observed that the variance between groups of tables, within groups of tables and across tournaments with very few tables is far too large

I appreciate this has a desirable feature of making results on sites like this something of a lottery which some would regard as fair to all users. However given that the site is now being used for real ACBL (not BBO) masterpoints is this not a more serious matter worthy of attention to ensure high quality results in tournaments that actually give some benefit to the better players and reduce the randomness

Also, an additional question. I havent studied the maths of bridge tournaments enough but given the variance between table groups in massive tournaments has anyone ever considered if any form of standardisation could be used to ensure more accurate reflection of results, based upon a variance adjustment. Im sure its very complex, since to get 100s of sets of hands to behaviour similar behaviour must be highly complex I would be interested in any references

Just asking as an observation. Obviously these simulation and competition things have all been worked out. Im fairly new to computer tournaments

regards P

[hrothgar]

I can't follow your discussion about bias and variance.

With respect to some of your other questions / comments.

1. From my perspective, the folks running organized bridge aren't interested in these topics. They certainly aren't willing to invest in the type of record keeping that would be necessary to study these questions in a more systematic manner. My attitude has always been if/when the organizations are willing to invest in record keeping we can worry about this stuff.

2. You might find the following topic of interest. It obliquely references some of the issued that you are raising

https://www.bridgeba...__1#entry957304