i met with a HFT guy and he told me in FX HFT world, Sharpe of 40 to 60 is possible even with transaction costs. how could this be possible....and what might be the strategies? could anybody shed some lights?

Hi,This seems really high to me as well, but you need to know what sharpe ratio you are talking about... In hft since you are doing many trades per day, if your market making strategy has a statistical edge it canreach a definite target for your daily performance quite easily, thus obtaining a really low variability in your daily returns. In the end I guess the sharpe ratio based on daily returns is not the main statistics to look at when you want to evaluate a HFT strategy. In HFT I found that people are looking at how many bp you make for your turnover, and the sharpe per trade (getting in and out of a position) rather than a sharpe based on daily returns. You need also to think that most HFT strats have very limited capacity.

Max DD = 100%, always bet on black.

Be really careful when discussing and comparing Sharpe Ratios as while everyone says they know exactly what it means there are actually various different flavours. You'll see lots of variations in trading programs too.For example the excess return and standard deviation may be measured month-on-month or year-on-year. Both of these will have different values. Similarly people often talk about Sharpe Ratio in a per-trade context, for example the average of trade trade returns divided by the standard deviation of trade returns.Furthermore when calculating the 'excess return' different people may use different base for which the excess return is above. Generally this is the risk-free rate however it may in some people's view be the rate foregone by having funds committed to the trade rather than an alternative investment. Or some just use zero as the idea of a risk-free rate may be too complicated to consider Also you need to make sure that a sufficient time frame and/or number of sample points have been taken. For example consider a trading strategy of writing heavily out of the money options, especially engineered to produce a very consistent profit (while not being exercised). You may go on for months or years producing a very stable income and hence ridiculously high Sharpe Ratio. However, we all know that every now and then things go wrong, those out outside events will foul things up.

So if you trade intraday you could say your risk free rate is zero, you don't roll over any positions so no interest rate charge.You trade on a very small spread which amounts to your total transaction cost on spot fx. You can enter a tp between the spread too.Many small successful trades for a specific amount smooths the pnl and the stdev approaches 0.

That's really amazing, just always wonder what kind of strategy hft guys employ. have been tests lots of different models, pairs/triplet trading, other technical based strategies like momentum and mean reveresion, I have never found anything can work consistently. So really curious what other people do.

If every month you beat cash by 1bp. By the end of the year you have made 12bps and your sharpe ratio is infinity. All well and good, but lots of work (and lots of flow) for little profit."Lever up!" you say.Next year you leverage 100 times and hope make 12%, but your credit dries up because of credit squeeze, you have to delever in a thin market. Annual sharpe ratio of -100.Google LTCM. Conclusion: Find me a strategy with sharpe of 40 with less than 5x leverage and returns more than 10% pa, then I'll sit up an listen.

The high leverage is just a decision based on the high SR of the strategy. If someone tells me they have a strategy with a SR of 10, my question would be "what is the potential (not backtested) drawdown of this strategy?"For a non-robust strategy, millions of transactions in a year = significant chance of a ruinous number of snake eyes coming up in a row.

Your friend is most likely working for "the house" (a.k.a. the bucket shop)and the strategy makes money by unnecessarily intermediatingcustomer flow within the matching engine.If that is the case, the strategy is not replicable and the edgeis based on a combination of two things.First, having a priviledged position at the center of captive customer order flow.Second, a loosely regulated market structure where noone cares ifthe customer gets ripped off "a little bit".Great money while it lasts though.

Sharpes of 40-60, after transaction costs and after funding costs are quite abundant in HFT space. Unlevered per annum regularly exceed 100%. Some strategies make money from client flow (e.g. GETCO), but most have no client flow. Some of these strategies are market making, but many are much more directional. Well 40-60 is really what you would get if you were intentionally running the strategy well below capacity. As you get closer to capacity Sharpes start to rapidly decay. Most strategies run closer to 10-15 Sharpes, because wouldn't you rather be making more money?All of this is irrelevant though. When thinking about HFT its better to think about returns to labor rather than returns to capital. Building those strategies and the infrastructure to support them takes thousands of highly skilled man hours. Most disappointingly strategy capacity is always lower than what you think it will be. Is it that unbelievable that a team of 20 highly intelligent people can come up with a risk-free way to print $5 million a year?Keep in mind the opportunity costs for these same people would be to go work for some credit fund that can collect billions of dollars AUM and make more on the management fees alone.