Slippage, how much in percent?

[elit]

How much slippage (in percent) should I take into consideration for stocks with daily volume of over 200,000?
Right now I'm testing stocks in Nasdaq100.

Can I use average daily volume to ultimately decide how much slippage I should use in testing?

I entered 0.10%, but I have no idea if this is valid...
Please advice me...

 

[elit]

Obsivously it depends on the size of the position. But say you're trading a 25K account and putting 2% on the line, it's 500. Or 50K, it's 1000. 100K it's 2000.

I'm just looking for something to go after. It can be too large amount too, I just want to take slippage into consideration in my tests.

People always preach about slippage not being taken into considerstion in tests, so what amount of slippage are you using?

 

[EricP]

For a Nasdaq100 stock, a slippage of one penny per share is more than enough (assuming you are trading modest size, as you suggest).

To get more fine-tuned that this requires more information on how you are doing your backtesting. For example, assuming that you are backtesting with 5 min intraday data and make the trade decision upon the completion of a 5 min bar and are assuming the execution will take place at the open of the subsequent bar. What does that next bar's open really represent? It represents the price of the NEXT trade that took place in the stock.

Note that this next trade may have taken place at the bid, or at the offer, or possibly somewhere in between. For a real (buy) order, for example, if you wanted to get into a position after the completion of 5 min bar, you would have bought at the ask price. The last trade MAY have been at the ask price, but it also MAY have actually been at the bid price. If the NEXT trade is printed at the bid, and your assume this is your fill, then during live trading you will experience a slippage equal to the spread in that stock at the time of the order (in other words, your fill at the ask minus the assumed fill at the bid - i.e. next bar open).

Assuming that roughly half of the trades in a stock take place at the bid and the other half take place at the offer, then on average your experienced slippage from this sort of backtesting will be half of the average spread in that stock (i.e. half of the time you will see zero slippage and the other half of the time, you will see the slippage equal to the entire spread).

For a Nasdaq100 stock, you will mostly see bid/ask spreads of a penny, so your overall slippage will be no better than 0.5 cents per share, and 1.0 penny per share is not a bad estimate.

Note that if your backtesting incorporates bid/ask data, and assumes all buy orders at the offer and sell orders at the bid, then your actual slippage may only be 0.1 or 0.3 cents per share. As I said, it all depends upon how the order executions are assumed in your backtesting.

Good luck,
-Eric

Quote from elit:

Obsivously it depends on the size of the position. But say you're trading a 25K account and putting 2% on the line, it's 500. Or 50K, it's 1000. 100K it's 2000.

I'm just looking for something to go after. It can be too large amount too, I just want to take slippage into consideration in my tests.

People always preach about slippage not being taken into considerstion in tests, so what amount of slippage are you using?

 

[elit]

Thanks Eric.

I'm backtesting with EOD-data (OHLC).
So for nasdaq100, there would only be average slippage of 1 to 3 penny per share, ok.

Is a correlation between volume and slippage significant or does it depend on other factors too, like volatility?

Is there a way to predict how much slippage there will be for a certain stock mathematically, or can it just be observed just by learning by doing, i.e. from experience?

 

[EricP]

Quote from elit:

Is a correlation between volume and slippage significant or does it depend on other factors too, like volatility?

Is there a way to predict how much slippage there will be for a certain stock mathematically, or can it just be observed just by learning by doing, i.e. from experience?

Typically, the higher the average daily volume, the lower the slippage. (More active stocks tend to have smaller bid/ask spreads and more size available at each level)

There is no way to predict slippage mathematically, you have to see how the stock typically trades, especially details like the average size of the bid/ask spread and the average number of shares available at each bid or ask level.

 

[ronblack]

Quote from EricP:

Typically, the higher the average daily volume, the lower the slippage. (More active stocks tend to have smaller bid/ask spreads and more size available at each level)

Yes, this is true during normal market conditions. However, slippage also occurs during fast markets or in the case of partial fills, etc and it is often more significant than that caused by the bid/ask spread.

Ron

 

[trillenium]

It all depends how liquid the stock is and whether you use limit orders or stop orders.

High Liquid stocks + stop orders = normaly hardly any slippage only more slippage around new highs or lows of the day

Low Liquid stocks + stop orders = high slippage especially at low or highs of the day


High Liquid stocks + limit orders = higher slippage than expected because you are always behind the other guys on you bid and offer and market runs away from you often

Low Liquid stocks + limit orders = Lower slippage because you can get filled sometimes at surprising levels.

These things are true for futures markets - as I am a futures trader -- for stocks i assume the same.

 

[kut2k2]

Quote from elit:

... I just want to take slippage into consideration in my tests.

People always preach about slippage not being taken into considerstion in tests, so what amount of slippage are you using?

... I'm backtesting with EOD-data (OHLC).

I use the following:

.94*O + .03*H + .03*L

The advantage of that formula is that I make no assumptions about whether the slippage is favorable or unfavorable, it simply is accounted for.

HTH