The problem set has been udpated to resolve some annoyances with non-mating tactics.
The major changes are:
- Ambiguity threshold has been made more aggressive.
- More positions which have winning second best moves have been removed.
- Pruning made more aggressive in some situations.
- Pruning relaxed where opponent has counterplay.
Hopefully these changes result in a noticeable improvement in quality. They resolve almost all of the issues raised on the forum related to non-mating tactics (which isn't to say that annoyances in this area don't still remain!). For example issues with #9352, #6453, #3421 and #1679 have been resolved. Overall about 1500 problems have been removed, around 1500 new problems added, and just under 300 existing problems have been replaced with improved move sequences.
The last change is probably the most debatable in terms of quality improvement. Previously, tactics were pruned when the material taken by both sides for the rest of the tactic was equal, however this led to some situations where the point of the tactic was non-obvious, for example in #3761. The replacement position #11112 requires the user to prove they understand the tactical point, rather than just find the first move. This improvement is gained by only pruning due to equal material exchanges when the opponent doesn't have a piece take on the next move. Whilst this allows the above example to play out to a more natural conclusion, it also creates situations such as position #101 where the replacement tactic requires the player to take back the opponent queen after the opponent takes the users queen at the end of the tactic. The tactical point was really made on the previous move but the user is required to show they can answer the queen take. Currently I've decided that the lack of clarity in position #3761 is worse than than requiring the user to prove they can answer counterplay, however I'm very much open to feedback on this. I can easily switch back to the old behaviour in the next problem set update, however it will mean losing some of the clarity/precision improvements made to positions like #3761. Overall these counterplay changes only impact less than 3% of the problem set, so hopefully its not a major issue either way.
As this update only targets improvements to non-mating tactics, some known issues related to mating tactics are unresolved. Specifically, positions which have a long mating opportunity, with a shorter non-mate but material winning advantage will still expect the user to find the mate.
The next step in quality improvement will be to generate a larger problem set with the current problem generator. This will allow the long mating problems that even the most skilled user has issues with to be seen less often as the problem's ratings further increase. It will also allow the use of an improved problem selection algorithm which whilst still providing a range of problems, should provide problems with more appropriate difficulties across the skill spectrum.
After a larger problem set has been produced, the issue of long mates with enticing material winning opportunities will be revisited, although the natural pressure towards extreme ratings for these problems should lessen their impact as a larger number of positions enter the problem set.
This update has required a large number of changes to the problem generator, so its possible there are quality decreases I haven't noticed yet. Please let me know if you find any new (or old) issues in the current problem set.
Regards,
Richard.
