Quantpedia contacted me a few months ago and asked if I’d be interested in reviewing their site on my blog. I’m always looking for new ideas for trading systems, so I said “sure!” (Disclosure: they provided me with free account access during the review period.)
Quantpedia.com is an aggregator and interpreter of academic papers on trading and financial research. An encyclopedia of quant-based trading research, if you will. Their sources appear mainly to be online academic journals and articles, and they provide links to the original sources for in-depth reading if you so choose.
If you’ve ever read a financial research paper, you know that it’s often difficult to get to the heart of the matter. That’s where Quantpedia comes in: they’ve done the hard work of evaluating the papers for actionable ideas, and have categorized them for easy searchability.
Each trading idea includes a brief introduction and a description of the underlying principle at work. Below that, you’ll find a section for theoretical performance statistics, such as the markets the strategy trades, the period of rebalancing, the complexity of the strategy, theoretical returns, the “confidence” level of the strategy, etc. You can filter by these statistics as well.
Below that, you have the actual trading system instructions. This could be a couple of sentences or a few paragraphs, depending on the complexity of the system. Quantpedia also provides references to the main paper they used as a source, plus related writing on the subject. Finally, they provide a chart called “Hypothetical Future Performance”, which is a graph showing the range of returns over a 60 month period.
The site is quite simple to navigate. The real action is on the “screener” page. You are presented with (as of this writing) 394 systems, of which 66 are free. The screener consists of a series of popup menus, allowing you to narrow down the selection process.
While I also invest for the long term, my personal interest is in shorter-term swing trades. My first thought was, “Where is a filter for the average trade duration?” The closest the site comes to providing that is the “period” filter, which selects the rebalancing period. Since I don’t rebalance my trades, this wasn’t quite ideal for my personal style.
I inquired with Quantpedia, and was told that their main clients are mostly fund managers, for whom rebalancing was a necessary concern. Fair enough, but I had to use other ways to find ideas that might be useful. I found, for example, that filtering for a Sharpe ratio greater than 1.0 gave me shorter-length trades.
A side note: the service is priced with the institutional investor in mind, and not the casual or hobbyist trader. You can access 60 strategies for free, but the full service is $499 per year. There are other options depending on your budget as well.
The trading instructions are sometimes a little vague, and anyone wanting to use these systems “as is” might be left wondering, “Do I trade at the open? The close? What about position size?” Those sorts of things are not always clear.
SInce I’m going to test systems for myself anyway, I didn’t find that to be a problem per se. Quantpedia is a source for trading ideas, rather than a recipe for your day-to-day trading. You still have to do some of the intellectual heavy lifting.
Financial researchers often ignore little things like commissions and slippage when doing their research. “Commissions” of course are the fees you pay to your broker to make a trade. “Slippage” is the deviation from the expected result that you experience in real-world trading.
Slippage can appear in lots of places in your trades, and it’s tough to estimate accurately when backtesting. Maybe your buy-stop price is always entering at market a few cents above your backtested stop price. Maybe you’re using a trigger price instead of entering a hard limit, and you’re losing a few cents here and there. Heck, you even buy or sell at the market open, and still have a price that’s a few cents off from what your trading data says.
All of this acts as a drag on your trading. Because it’s hard to quantify, and because it’s different for each trader, researchers tend to ignore this. For traders though, this is a Very Bad Thing.
Take for example a system that is described on Quantpedia. It aims to capture the supposedly higher returns that are found with overnight trading. I decided this sounded interesting, and went about implementing the system using my backtest software.
Looks pretty good, right? Now let’s throw in a little slippage. I used this parameter:
Slippage: Max(0.05, 0.001 * FillPrice)
Which means, the greater of:
$0.05 per share
0.001 x the buy price (per share)
How does it look now?
Not so good. Did I use slippage that is similar to real-world conditions? I don’t know, but it’s a guess. I’d have to actually run the system and then record the slippage results, which could get expensive.
Is this Quantpedia’s fault? No, of course not. However, it is this sort of real-world error that you need to take into account when using academic research to implement actual trading systems. This means that Quantpedia is best used as a stepping-off point for your own research. It is not a service that will spoon-feed you ready-made systems to trade.
That is where Quantpedia is most useful. If you’re the backtesting type, you’ll likely discover enough interesting systems to send you scurrying off to your computer for hours (or days). You will likely implement them in ways that differ from what you read, but that’s ok.
Interestingly, near the end of my review process, Quantpedia announced a partnership with QuantConnect. QuantConnect is implementing backtests of selected systems in Quantpedia and providing real-world, out-of-sample results. As of this writing, there are about 20 systems that have been tested.That ‘overnight system’ I tested above showed similar results when QuantConnect tested it. The edge for that particular system doesn’t seem to exist in the real world.
I like that Quantpedia is implementing this backtesting and out-of-sample testing, because it will certainly weed out the “fantasy” systems. This will increase the usefulness of the site immensely. Even if a given system doesn’t look realistic after QuantConnect tests it, you might still find a useful idea or two to test yourself.
An example: I came across a system involving short-term trades that capture the dividend payments from equities. The problem was, I didn’t have access to alternative data that the strategy requires (beyond price/volume/splits/dividends). I couldn’t implement the system as described, but it sent me off into a new research direction. I was able to come up with a system that looks pretty good…at least in backtest mode. I’m currently testing this system with real trading (in small lots) to see what slippage effects there are. I’m quite sure I would never have explored using dividends as a signal if it weren’t for Quantpedia.
If you’re the sort of trader who takes confidence in hard numbers and cold statistics, and you’re moving enough money around to justify the subscription price, Quantpedia might be a very useful idea generator. Even if it doesn’t fit your budget, it’s worth checking out the free systems. You don’t even have to sign up with an account. You never know where they might lead you!