The exit is always the most important trade…Can a subscriber idea for an exit modification improve results?
PDS issues an exit signal in the evening to be acted on at the next close. One of the subscribers left a comment on the site wondering whether selling a stock when it gaps on the open is better than waiting to sell it at the close. It is an interesting idea, and I decided to backtest it, although caveats must be added to the results (found after the summary).
I am using the 10% Equity Per Trade model as the baseline model as it will make the most trades. When conditions are right, it will have 10 open trades at once. This should give plenty of samples over the short term.
Rules:
- Apply standard exit BUT close at open (instead of close) IF the open gaps more than X% above previous close.
- Commissions of .01/share included.
- De-listed data used (survivor free).
Results:
Summary:
- The exit modification of selling the open when the stock will gap up MAY improve results (see caveats) as over the long-term tests, it has beat both the baseline and baseline with selling the open.
- Just selling at the open instead of the close has done well since 1.1.2011. However, over the long-term, selling at the open has slightly under-performed the baseline.
I think this is an interesting exit modification, and I’m grateful to Rick for introducing it. If one is an experienced trader and understands how to properly place orders, monitoring liquidity and the bid-ask spread, I believe this modification could improve results. Of course it would have to be applied diligently and consistently or the trader would likely introduce biases which would probably negate any positive benefit.
Caveats:
Trying to backtest selling the open because of a gap-up introduces a “look-ahead” bias. In real life, there is absolutely no way to know exactly where a stock will open. Yet, when backtesting, we know exactly where it opened and can then place our sells according to the open AND get the open price. In real life, if a stock gaps +4%, the likelihood of getting the exact open price is very, very slim. There is no sure-fire way to know that your sell order will fill for more than 2%, or 1% above the previous close. Thus, one must take these results as a guide, and not as the gospel. Furthermore, the open is notoriously illiquid, which can exacerbate the problems highlighted above.
Very nice modification woodshedder. Thanks go to Rick for suggesting it. It looks like it may really improve results. There are a lot of caveats, though. With getting the opening price, my understanding is that if you do a Market on Open or Limit on Open order you would be guaranteed to get whatever the opening cross is. Nonetheless, I suppose you would still have the consolidated/proper open business where the market data does not show the actual open price. You’d have to find some way to get around that issue… At any rate, nice work.
Data, the biggest issue, in my opinion, is not being able to know with any assurance whether the stock will truly gap open or not. Often, at 9:29, the bid/ask spread is very wide. The ask may be suggesting a large gap up and the bid may be suggesting a large gap down. For less liquid stocks, this could be a real problem. For more liquid stocks, if one is fast with order entry, he could wait until the stock opens, confirm the gap, and then place the order. Still, this is fraught with problems because the first few minutes after the open are illiquid, even on a liquid stock. Market orders may crush the price and limit orders may not get filled before another trader’s market order crushes the price. So while we are guaranteed, when using a MOO order to get the opening cross, we have no good way to know exactly where the cross will happen in relationship to the previous close.
Ah, I see what you mean – sometimes the Limit on Open (LOO)/LOC orders don’t fill because there’s not enough liquidity at the official opening price (so you get a partial fill). This hardly happens with the LOC and the PDS, though, so maybe it wouldn’t be a big issue with LOO? I guess it would depend on the liquidity of the open v. the close plus the liquidity of the gapping picks.