We spent two months grinding through the data because every day, we see buyers with born-to-lose tattoos on their foreheads making lowball offers doomed to instant failure. And then sometimes they win.

Nuts! Figuring out fast when a deal will fly and when it won’t is the key to making our margins work.

We’ve tried to sort out when a seller will give ground, a posting that immediately set off a discussion at Newsweek and U.S. News & World Report and across the blogs. We’ve talked about it on TV.

Meanwhile, pricing discussions have been all the rage in our discussion forums, with visitors asking how negotiable new construction is, or even what to offer on an individual property.

Enter the real estate scientist, who shreds MLS data, website logs, academic research and tax records to provide reliable answers to practical questions: what really works when marketing a listing, what type of listings sell in less than seven days, whether high commissions lead to faster sales (they don’t).

Today’s report is about how to find the sweetest deals. Redfin’s Rob McGarty and Mose Andre crunched the numbers for all 9,053 sales of single-family homes in Fairfax (Virginia ‘burbs), King (Seattle area) and Los Angeles counties, between April 15 and June 15 this year, to look at homes that sold for a lot less than the asking price.

What surprised us right off the bat was how small the average discount was: after all the bad news about the housing market, it was still less than 2.5%. We were also surprised to see that a small fraction of listings got a huge discount, of more than 10% off list price.

This means that 90% of the time, lowballers are wasting their time. But that 10% of the time, they’re saving $50,000. The trick is finding the listings ripe for a discount.

So for each market we split the sales into two buckets: those that got the biggest discount (top ten percentile, an average of 11.4% off the final asking price), and all the rest (the 90th percentile and below, which got an average discount of 1.5%).

Then we compared the two groups to see what the sales involving a big discount had in common:

1.    83% more likely to have been on the market for 90+ days.
2.    73% more likely to be marketed as fixer-uppers.
3.    20% less likely to feature a noteworthy remodel.
4.    28% more likely to have already been price-reduced (though there were some confounding factors).
5.    52% more likely to have been seller-owned for 20 years or more.
6.    9% more likely to have been seller-owned for less than five years, a slight but surprising correlation.
7.    Only 9% more likely to be a short sale or bank-owned. Most consumers expect this correlation to be much stronger.

The portrait that emerged is  of a bimodal market (much like the VC market right now) — a few homes in the most sought-after neighborhoods sell for well above asking price, most sell near the asking price and anything distressed or run-down (more homes are run-down now than before so the drop in prices is in part a reflection of the drop in the quality of the homes being sold) just keeps dropping and dropping.

The bottom-line is that the condition of the home and its length of time on the market — surprise, surprise — are the main drivers for discounting — but how long the seller has owned the home is a big factor too. And foreclosures aren’t as negotiable as people think.

The report has already gotten picked up by the LA Times and now the Seattle Times too (FRONT PAGE!!!).

Get the full paper (PDF), which breaks out all the data by county, and goes into the methodology too.