There have been a bunch of books on the wisdom of mobs. The basic assumption is usually that the average behaviour of a group is for a lot of purposes better than most of that of the individuals. A popular example is usually the stock market. Individual investors quite often think they can beat the market, but obviously on average they can't. Mix in the transaction costs and the fact that some people do stupid things and you end up with a situation that on average it is actually better to follow the index than to think that you are smarter than the market. And indeed, index following funds have been rather popular and successful. Of course following the index means you're always going to do a little worse.
As long as you just follow the index, you are of course doing exactly as the market in general. However, every so often the index has to be re-weighted. Stocks that have done worse are taken out, or are getting a lower weight, stocks that are doing better are increased in weight, or new stocks even come in. If you follow this behaviour by buying more stocks in the better doing stocks and selling some from the worse performing, than you will be trailing the market, since your basically buying high and selling low (well, a new stock in the index might obviously still go up, but a stock that gets dropped presumably is on a lower price than when it came into the index).
Anyway, I was thinking, if you knew about all the trades that ever went on, you could do better. But how much? You could exactly follow the market by just executing all the trades your self too to some extend. Could you do better? I would think you could probably pick the very worst and maybe the best ones too of the investors you are tracking. If you just trade a little bit more like the best guys and a little less like the guys that are not doing so great, you're there.
What if you would track only a random sample of like ten percent of all people that trade on the stock market? It would probably still work. So how many people would you need? Probably any decent investment bank would have enough to go on. They can just look at the trades their customers make, disregard the customers that seem not so smart from what they're doing, overemphasise the trades of the guys that seem to excel and there they go.
Of course you could argue that this would never work, since there are these banker types that have the inside information and that the best thing you can expect is to nearly do as well as the market. So where do we think these bankers get the inside information from? Ah...