I’ve read about making sure something in your models penalizes a stock if it’s price is too low. The only mention of an example price I’ve yet come across is the ranking back-test minimum price of $3. It certainly sounds reasonable. Does anyone have some methods they went through in filing down to such a number?
Thanks,
Flynn
I have been buying low-priced stocks for five years now and I use a cutoff of $0.50, though I very rarely buy anything with a price that low. Of the 56 stocks I currently own, ten are priced at $3.00 or less. In my experience, there’s not much difference between a stock that sells for $1.50 and a stock that sells for $50 if the bid-ask spread as a percentage of price and the daily dollar volume are the same. For example, PLPC, whose current price is $55, is a hard stock to get in and out of without paying most of the (wide) spread, while RIG, whose current price is $1.07, is extremely liquid, and your transaction cost would be tiny.
When running backtests, you should definitely put a price limit of $0.50, $1.00, or more, for reasons that Chaim elaborated on in this thread: https://www.portfolio123.com/mvnforum/viewthread_thread,12566 But it’s also important to use liquidity limits. I recommend establishing limits using MedianDailyTot(65) and LoopMedian(“Spread(Ctr)/Close(Ctr)”,35)–play around with them and see what limits are most comfortable for you, and feel free to change the 65- and 35-bar limits to whatever makes the most sense. Also keep in mind that stocks not listed on the major exchanges often have Spread = NA.
As a general rule, I have found that the $3 a share cutoff was arbitrary.
We need filters to eliminate pricing errors common among nanocaps. For example, this public screen, with a 16,055.78% annual return, demonstrates the data errors evident among very tiny stocks. To eliminate those errors, simply raise the minimums until the suspicious spikes disappear from the backtests.
These errors are almost exclusively limited to nanocaps. In my experience, a MktCap minimum is a more precise way to eliminate errors than a price filter. I usually have the following rules to minimize data errors:
MktCap > 25
Price > 0.1
I also have additional rules (depending on the position size and the turnover rate) to specify liquidity limits like Yuval mentioned. That is not because of errors but because of portfolio size limitations.
Thank you Yuval and Chipper6. Both your responses have been very interesting and helpful.
FT:
Rec you use avgdailytot(20)>cutoff such as 100000 or 1000000 depending on the type of stocks you like to trade