New strategy - trading "trash" stocks, intra-day trading and technical analysis


I’ve been experimenting with a new strategy lately that is quite different from my current quant strategies (and assuming most P123 users). I’d like to explain the strategy, and get any feedback from others who have had similar experiences. I touched on it in the thread below, but I’ll elaborate here.,12533#!#81884

After building up a few models and strategies over time, you realize that most models do not perform well all the time (except a few, I do have a couple that I remain invested in through thick and thin that have performed well out of sample). Other strategies do exceptionally well under some market conditions, while tanking in others. I spend some of my investing time trying to find those strategies in their “prime time”. One way I do this is simply by seeing their recent performance, particularly during a market shake-up. This is typically when previously poorly performing factors do well, and vice versa; in other words, when a strategy is experiencing a positive reversal.

After performing this exercise for a relatively simple deep value strategy, I found some nice outperformance recently. Looking at the stocks the strategy is screening, these stocks are very deep value. In fact, you could argue that some of these have little to no value at all, and are “trash”. Little to no sales (and profit), late reporting, very poor recent headlines, chronic diluters. Yet, when run through the sim, they have been doing well.

Like any deep value strategy though, the idea is to find stocks at such a low price that they have nowhere to go except up (in theory at least).
Another interesting development - backtesting using close prices only (standard for P123), you miss much of the performance in these stocks. Many of these stocks experience some wild price action intra-day, I would even hazard to say extreme, which isn’t reflected in the close price.

So, instead of your weekly or monthly rebalance, the opportunity lies in selling during one of these violent price surges. Easier said than done, hence the reason for this post.
Now we are getting into “day” or “swing trading” territory. This is a different world compared to investing, even if you rebalance weekly.

The strategy currently is to buy a number of these deep value/trash names, and hold. I keep them on my watchlist intraday, and when they start increasing in price (+20% or more) then I watch them very closely. A move of +10% or even 15% is typically just noise, but I find after 20% (especially with high volume) the stock will likely get legs.

In more high quality names, price moves like this are usually preceded by a positive news release, earnings beat, etc. For these “trash” socks, in some cases there is a rare news release (FDA approval or positive development for biotech names), but often times there is nothing in the mainstream media. Once you peruse Twitter for these moving stocks, you can see a great deal of activity. Some of these stocks are even subject of highly driven promoters, trying to get the price as high as possible.

In terms of deciding when to sell, when you start looking at true day trading strategies, there are countless technical indicators to help (whether they work or not is something I’m still trying to determine; any by they way, for something called “technical analysis” the measures are awfully subjective). Right now, I’m selling these extreme price movers “by gut”. It’s only been a month in the strategy, but so far so good. 4 stocks have had returns 200-400% within 2 weeks. On the flipside, as many of these are trash, many of these stocks slowly burn though, losing 20-40% in a few weeks if you just hold them.

Most of these stocks have low float, so it doesn’t take much to move them. For example, the median float of the S&P500 stocks is 281 million shares, for this “trash” strategy it is in the neighbourhod of 15M shares.

Very interestingly, for these stocks with sudden moves in a day, they often do not hold the gains for very long. Sometime the next day they sell-off, typically in 1 to 2 days, up to a week.
So the trick is to find or time more of these extreme winners, and to minimize the number of losers.

I’ve looked at day trading techniques, but I’m reluctant to start learning a whole new way of trading/investing when things work well already. That said, I’m very interested in complementing my current investing with trading techniques to enhance returns.

All in all, this strategy and these stocks take me right out of my comfort zone for several reasons:
• Very low quality
• Fundamentals count for almost nothing
• Highly volatile
• Attract speculation and “promotion”
• Require regular monitoring (hourly or even shorter)
• Very difficult to backtest (using P123)

Reading the above without experiencing it, I would have been immediately dismissive. One month is by no means long enough to decide if this performance so far is just luck, or if there is some signal in the strategy, but based on what I’ve seen I’m curious.

My questions out to the community – do you incorporate day trading into your P123 strategies? Do you find any technical indicators helpful? Do you “trade volatility”? What are your “day” or “swing trading” experiences?



Just a thought based on something I am looking at with something entirely different.

You mention 20% as separating noise from a real signal. Probably stick with that or at least keep looking at it.

But you might also look at “Effect Size” which is calculated like a z-score but has a rich meaning. Effect size is used a lot in Bayesian statistics which is a really rich subject and probably does not need to be discussed in detail here. As a simplified, practical example, a drug company might have something that shows statistically significant results but squash its development if it really does not have that large of an effect (as measured by the effect size) in practical terms. But also something with a greater effect size is more likely to be found to be statistically significant.

So this one statistic COULD give insight as to whether the equity will give excess returns AND how big the return might be. That is the theory anyway. Assuming it looks like this has any usefulness at all should you check it out.

So I guess you might be using bars during the day and you would have to figure out what time-period is best for your bars. Maybe continue to filter the equities with a 20% change.

Then just see if—starting at the period where the change began (i.e., the 'inflection point")—whether the (average return of the bars)/(standard deviation of the bar’s return) helps you separate-out the better stocks.

Ultimately, this will either work or it won’t and a lot of hand-waving on my part will not change the result. But there is a lot of good statistical reasoning suggesting that it could be helpful and I think it can help with ETFs.

Okay a little hand-waving: maybe you don’t want the stocks with just one (or a few) bars with large changes. Maybe you want to identify the stocks with consistent interest (e.g., a whale that keeps buying or a story that is consistently bringing in new buyers).

And actually the statistical justification is easier. You are measuring the “Effect Size” of the story in a formal way. How effective it has been up until now (which could change of course). Again, this is not to say I have tried effect size as a factor for your situation or that I have any data suggesting that it will work for your situation.

BTW, for selling–which seems to be a big concern–an effect-size-window could help. Meaning if the effect size has not been so good and/or is declining over X number of bars it might be time to sell. Assuming it looks like this any benefit for what you are doing.



Thanks Jim. I’ve looked into the “effect size”, and wrapping my head around how I could use it in the strategy. Still simmering on it, will keep you posted.