Many AB users have good histories of EOD data but limited real time data
histories for back testing. This can cause some problems such as overly
optimistic results when the trading logic uses entry and exit stops.
We all know, or should know, that if you are using the closing price then for
realistic results in a back test (BT), the trading logic should use the open
(or later) of the following day as the entry price. But, what if you are using
stops to enter and exit, donât you use the stop price as the entry or exit
price in the back test?
The answer is yes (+- slippage) if you are using real time (RT) data for your
BT. However, if all you have available is EOD data for longer term BTs then
you need to be wary of calculating results using stop prices.
The problem is the close price. What is the close? In RT terms the close is
the current tick . With EOD data the close is the last tick of the entire
dayâ¦big difference there, so you cannot use EOD close in your calculations
for stop entries. Again if you use close, the trade is an EOD only trade and
has to take place the next day whether you specified a stop level or not.
So what about the stop price? To try to simulate the results properly you
need to look for highs to be greater than your long entry stop price and for
lows to be less than your short entry price. Reverse that logic for stop exits.
I had a trading system that had the best walk forward results I have ever
seen, but the nagging thought was that there was a flaw in the logic somewhere
as the results were too good, even though the code passed the simple look
forward test.
After watching the system trade for several weeks (its an infrequent trader)
I could see the problem. Running the BT frequently throughout the day with an
RT data feed from IB the RT close (current tick) would rise above the long
entry stop level and I would be long at the ~stop price. However, a couple
hours later the RT close would be below the long entry stop price and the trade
would disappear and âit never happenedâ.
This trade turns out to be a loser when evaluated in RT, but, it does not
show up in the BT evaluation with EOD data because of the use of close.
In this case using high for the BT logic will trigger the trade and it will
show up in the BT results as a loser. You will have the entry at the stop
price and the exit will have to be treated in a similar fashion for more
realistic results.
Bottom line: if you use stops and you only have EOD data available for
extended back tests you need to use high and low rather than close in your
trading logic.
Cheers
Sid
