OK, personally I think there is some truth in EMH and that you can't predict the big financial markets. But, arbitrage seems a little bit underrated by EMH. I am not sure if Prof. Fama has stated this in his original thesis back in the 70s, but you will find many "official studies" online(and in your university) that in the EMH world arbitrage doesn't exist or when it occurs - it lasts for just a few seconds. OK....but just a "minor detail" that we have something called "high frequency trading" - Yes, you are right - "a few seconds are NOT problem for a high frequency computer". Actually, below I give you a few examples(without formulas and sophisticated code - just easy to understand arguments), as to why APT, MPT, EMH and all these abbreviations fail when it comes to arbitrage: 1. Arbitrage exists for just 2-3 seconds?: See above ^...high frequency trading. Also, see this very interesting study from Oxford which not only disproves EMH but even its "refined version" - the Adaptive Market Hypothesis: AMH (too many abbrevations already...funny). So: study 2. You cannot profit from a carry trade, due to large market risk? Fail. You can very easily: a) Go long on XAU/USD or vice versa, short it. b) Hedge the market risk with a comex gold futures gold contract. c) Receive the tom rollover on the XAU/USD. At the end - you end up with no market risk and 3%(or more) leveraged inter market rate. How much is that since both your futures and xauusd forex are margined? You are right: over 50% interest yearly without any market and default risk, the % obviously depends on your broker and how much your forex position is leveraged. The comex gold is an exchange defined initial/maintenance variance margin. You can make a similar trade with cfd/stocks. 3. Options/betting markets are very efficient: Fail. With some persistence you can easily find a broker with call/put option sell premium higher than other broker with option buy premium lower, on OTC option markets. You are obviously hedged when buying at the lower price and receiving higher premium at the other broker. Problem here is that there is market risk - but only "virtually", since you cannot lose even if one of the positions is closed. Of course you can do the same in betting markets when the odds summed on -1th power are below 1. There some other examples. Point is that when you read about arbitrage you read only about "buy microsoft stock in london and sell it in new york when prices differ" - which is complete ballooney. Not only this type of "Arbitrage" doesn't happen - but it's practically nearly impossible to make money from this even when it happens. The other type that you will read about online is the "triangular arbitrage" with 3 currencies - which occurs once in a millenium. The first examples however are much more practical, if not shocking since some of them exist for quite some time and not even 2-3 seconds. Which means that most "arbitrageurs" are actually stupid not to exploit them, despite being a public knowledge. I thought to go on details with formulas, links, references, computer code, etc. in this topic...but I am lazy and busy now.