One of the best book I have ever read on quantitative trading is Inside the Black Box by Rishi Narang. It doesn't give secret recipes to win money on the markets but it goes really deep inside the components of quantitatives strategies (alpha model, risk model, transaction cost model, portfolio construction model and execution model), the decisions quants have to think about when designing such strategies and good advices to evaluate the already running strategies. All that well written with clear schemes to help the reader follow the ideas. You may have to read more advanced books on each component but it gives a good framework to think about quant strategies. And back to the original post there is a good discussion on the pros and cons of trend following vs mean reverting strategies.
I liked that one a lot too. I think if anyone's looking for a book that they set down on the desk and flip through the pages while writing some code, close the book, run the program and make money, they're out of their minds. But the author did a terrific job, as you said, of laying out the basic logical architecture of an application that attempts do what a human trader would do. I actually refactored a few things to meet his model ideas, and then end up un-doing them halfway, but his basic ideas are still solid. Great book, very approachable.
Another Big issue is different between buy price and sell price. A short term trade to win after overcome two products buy and sell price different is much harder than backtest with just the trading price.