Test results
Testing method:
the auction is carried out by two competing Teams with 4 bots each;
Team 1 plays a match against Team 2 using random hands (at least 50 000);
each Team plays an independently defined bidding system;
bots know and perfectly understand the systems and conventions used by other bots;
each bot only knows its own cards;
the results are calculated on the basis of all contracts achieved in the first and second room;
the management program determines the number of tricks using the following methods:
Single Dummy - the number of tricks based on the visible cards of partners and information about opponents obtained from the auction;
Double Dummy - the number of tricks based on the visible cards of all players;
the lead and a play are not analyzed. It is assumed that the number of tricks taken will correspond to the estimated number of tricks.
The adopted method of analysis is devoid of the most important disadvantages of other methods:
different skill levels of players;
too few compared hands;
failure to take into account the impact of the convention on the entire system due to the lack of analysis of hands in which the convention did not appear;
disregarding possible or no intervention;
underestimating or overestimating the impact of scores for rare sequences on the overall system performance as the outcome is per hand.
It seems that for a large number of analyzed hands the results obtained in both comparisons should not differ too much.
The approximate accuracy of the calculations is:
for 20,000 hands: 0,4% and 0,04 imp/d;
for 50,000 hands: 0,06 % and 0,006 imp/d.
The Single Dummy method is approximately 100 times faster than the Double Dummy method. If necessary, Double Dummy can be disabled by clicking on the "Double dummy" label.
For individual systems