This is a question for those that have expertise using snowie. I like to use snowie to examine matches, and from time to time use it to rollout an interesting position. I typically import matches and use the 'precise' setting for analysis. However, in reading the posts on this board over the last while, it seems that people go into great depth (deep rollouts etc.) to determine the correct plays. What are peoples opinions of the snowie 'precise' mode of evaluation? Is this good enough to get accurate results for my matches (typically 5-9 points)? Any input would be appreciated. Thanks.

Analyze your matches overnight. It took several hours. Set analysis type on custom. Set 3 ply precise, then under Miscellaneus forced rollouts for errors and cube decisions, rollouts truncated cubeful. If your computer is fast, you can go with this setting through the whole match. I f you have a Pentium 4 with 2 threads or core2duos, you can run snowie twice to analyze two matches to take full cpu power. So you have similiar results to XGRoller+, but not at speed. First care for your bigger errors. You can store it for research like this: http://www.gammonluck.com/nordicseminar.html

The 'precise' setting gives you a 3-ply evaluation on both checker plays and rollouts. I use this as a first pass, then do rollouts on positions that seem interesting. Snowie uses a variance reduction technique on 2-ply and 3-ply rollouts, so you don't need as many trials to get a given level of accuracy. Typically, a rollout setting of 1000 trials would be equivalent to somewhere between 16k to 20k rollouts done without any reduction.

As I've mentioned before, I've found a setting of 2-ply, 648 trials to yield a happy medium between speed and accuracy. For certain types of positions (mostly backgames and blitz v. prime positions), I'll crank the rollouts up to 3-ply. I do rollouts for all the 'Problems of the Week' on 3-ply as well.

Thank you both for the replies. Very helpful.