Comparison: 94 X-Clubs vs 943 National Results

The full-field results for 943 players taking part in the New Zealand-Wide Pairs on 1st Novemebr was compared with the event scored over 94 X-Clubs players from six Compass clubs to ascertain the degree of accuracy which X-Clubs is producing.

We have already seen that wild swings can occur when a club session with six tables is scored across a field of 50 tables. The conclusion is that the bigger the field, the more objective the result in cross-field scoring. Here we are given the opportunity to compare a sample of one tenth the population with the actual population itself - which we may assume to be 100% accurate.

We started with fishing out the scores of the 94 Compass players and looking at the differences with their X-Clubs scores. Adding up all these differences we got a total of just 0.45 ie - the average of the X-Clubs scoring was virtually identical to the average score population.

53 scores moved up, 34 of them being E/W scores.
41 scores moved down, 12 of them being E/W scores.

There were 7 scores that moved down by more than 1% - one by -1.53% and another by -1.81%

There were 10 scores that moved up by more than 1% - one by +1.4% and another by +2.06%

82% of X-Club scores were within +/-1.00% of the national scores

72% of X-Club scores within +/-0.75% of national scores
64% of X-Club scores within +/-0.50% of national scores
20% of X-Club scores within +/-0.25% of national scores

As a predictor of placings there is a marked dichotomy - the bottom half of the X-Clubs listing being unduly pessimistic by an
average of 35 places with the top half overestimating by an average of 41 places. Overall an average over-prediction of 3 places. Dividing these figures by 10 makes them easier to digest.

These results seem reasonable enough, but the hope is that one day when our groups get "large enough" to be able to identify club sessions that could be thought of as advanced and to ringfence them off from other club sessions so that each division can play in its own class. This exercise was valuable in determining what that "large enough" value should be.

It seems that a field of 100 pairs is not quite enough to produce a stable result when compared to a field of 1000. Certainly 150 would be closer to an acceptable degree of predictive realism. That means that we would need about 250 pairs playing in a given session before we could split off 100 pairs to play amongst themselves. Some sessions currently are reaching up to 100 pairs.