The "best" bets are bets that are on average loss-making but over the short to medium term seem to make a profit (if they weren't on average loss-making, yields would probably be too low). What you do is look until you seem to have a reliable money maker, lever up and bet big. If it all pans out - as it usually does - you pocket a big chunk of the profit. If it fails (which it will once every few years or decades), you get the government to bail you out, or at worst the company goes bust.
Heads you win (most of the time), tails you don't lose much. It's a money-making (or rather, money-diverting) ratchet.
The feedback mechanisms don't work properly, particularly in finance. Humans don't live long enough and don't defer compensation long enough for the averages to work out properly. When a large proportion of the economy is devoted to this stuff (20+% in some countries), it's almost certainly highly inefficient, but the market won't fix it because the market can't fix it: again, the feedback mechanisms aren't strong enough.
The problem is that oversight and accountability is too weak. As CEO he was authorised to take actions which should've set alarm bells ringing with others, without apparent limit, and his contract (much like many others; Fred Goodwin would be a prime UK example) was poorly drafted so as to provide him with limited personal accountability for failure. UBS 'rogue trader' Kweku Adoboli is currently in custody awaiting trial for inflicting a serious loss on his company; Wendling Weideking was paid off handsomely for far worse. Boards need to take their role of executive oversight more seriously.
Are we playing downvote-disagree?
Free markets are about giving feedback like that. They aren't magical ways of making people never make bad decisions. And neither is your preferred $ECONOMIC_SYSTEM, whatever it may. If it's claiming to be one, it's lying.