Sanity check on sharpe ratio calculation please

tradrjoe said:
If your strategy is mean reverting, then it would bias you to often enter a buy at the bid and a sell at the ask within a backtest using last prices. In real life, this would not be realistic obviously. So it could potentially be the bulk of the outperformance shown in the backtest.
More...

I'm working on adding 2x the average spread as a cost along with the commission.would that cover this?
 
That's a good start unless you think the times you would want to enter a trade would be exactly the times when the spread would be more than average (e.g. times of higher volatility). Better to just use actual bid/ask prices if you have it.
 
tradrjoe said:
That's a good start unless you think the times you would want to enter a trade would be exactly the times when the spread would be more than average (e.g. times of higher volatility). Better to just use actual bid/ask prices if you have it.
More...
It's pretty much impossible to get a perfect backtest so there is really no good reason to make it more complicated. That's especially true given the product (FX is pretty friendly to test trading) and his desired Sharpe (I'd say anything over 4 should be tested live pretty quick).

@nooby_mcnoob - what pnl/tradevalue are you getting and how does that compare to your execution costs?
 
sle said:
It's pretty much impossible to get a perfect backtest so there is really no good reason to make it more complicated. That's especially true given the product (FX is pretty friendly to test trading) and his desired Sharpe (I'd say anything over 4 should be tested live pretty quick).

@nooby_mcnoob - what pnl/tradevalue are you getting and how does that compare to your execution costs?
More...

Something like .02% per trade whereas execution costs are .002%. Right now though I'm working on trying to figure out what I did wrong because I've been down this road before.
 
Btw, there is no possibility for lookahead bias because I literally don't use any data beyond the current tick. So it may still work, but the latest charts were incorrect because of this. The original is still correct afaict.
 
You must log in or register to reply here.
Back
Top