I'm surprised they're aggregating at the zip code level. The neighborhood differences within a zip code seem too great for an average measurement to have much meaning to an individual homebuyer/seller. Case Shiller breaks values down into three tiers of price, which provides a little more insight.
The arithmetic average for home prices is not too useful either. I clicked on a random county in Wyoming that had an average of more than $600K. Turns out that it's like 20 cheap houses (most going for $150K to $200K) and one huge farm for $29.5 million.
Personally, I'd want it at the neighborhood level. Though there may not be enough data within a reasonable timeframe at that level, it can at least be reasonably compared to a broader index for the city.
I don't think you get much at the zip code level that you wouldn't get at the Metro level.
The probability of a second reduction is also too simple. It makes no account for the market turning positive. That's precisely the error in modelling that failed us on the upswing: the assumption that present trends will continue.
A pricing heatmap in 10% bands would be pretty good I think. Houses next to each other tend to be more or less the same value with a few edge case scenarios -- the farm as noted in the other reply to this.
Realtors can and do handle price reductions by delisting and relisting the property. This hides the ask price drop, and zeroes the days on market, to the advantage of sellers and realtors. A brief look at Trulia's "methodology" description suggests that they are just mapping these easily gamed MLS numbers.