Nov 27, 2019 · 4 min read
Initially published in Medium
Traders frequently ask why an Options backtest software uses “just” one price per day for its backtesting studies.
The main reason is that it is a good compromise between precision and value.
To illustrate why it is important to make this trade-off, let us consider an example. Take the SPY, at any given time the SPY Options chain has around 34 expirations (32 to 36) and offers a range of 60 to 80 strikes each. We keep 33 indicators for each record (see the list below). You get the idea of where this is going: for one symbol like SPY we need to store and process over 350 million data points. Here is the calculation:
60 Strikes *
34 Expirations *
2 Put & Call *
220 Trading days/year *
12 Years * (from 2005)
33 Data points (Bid, Ask, volume, Open Int, Open, Close, High, Low, Greeks, IV, etc) *
= 355,449,600 Total Data Points
Of course, SPY is an extraordinary symbol — we know that. Most other symbols do not offer as many contracts. But there are over 4000 symbols available at eDeltaPro Options Backtesting Software.
So increasing the pricing interval, will substantially increase what is already a massive amount of data points. The increase would inevitably translate into higher costs for the user.
A study with a larger more granular data would certainly improve the results but in a minimal way. We are really facing a typical case of marginal returns, where a large increase in the resources committed to the studies would result only in small improvements.
Let us have a look at the two cases where you can expect to see the greater impact of increasing the granularity of the data.
When using “Exit at Max Profit” there is the possibility that some winning trades would not be accounted for, so that a more granular study would have a bigger P/L. In our studies, this touches less than 2% of the trades.
Since most of the differences are to the upside, meaning that EOD data — study is more “pessimistic” and tends to underestimate the P/L. So we can live with that. A trader using intraday prices could have more opportunities to close profitable trades and thus potentially beat the results of the backtest study.
The one area where the lack of intraday data may have a more significant impact on the results is when you consider the opposite case and apply Stop Losses. With EOD prices, some losing trades could go undetected and end up as winners. In our studies, while the number of these trades tends to be small, they have a larger impact on total P/L than in the “Max Profit” case. Total profits could be overstated. That is, larger than the actual profit an intraday trader would get. A more serious issue as you do not want the studies to be more optimistic than reality.
To mitigate this, we have built into eDeltaPro, ways to add some data points already available for triggering stop losses. We can use the High and Low as well as the opening price in addition to the 3:45 PM prices. As a general rule of thumb, we do not recommend studies where the stop loss is small.
eDeltaPro Backtesting Software incorporates a warning system that will tell the traders when the strategy she is trying to test may have low reliability due to “Early Exit” conditions being too close to the “noise” of the sample. We do this by comparing the stop loss with the intraday volatility.
By comparing the stop loss with the intraday volatility, we can determine when “Early Exit” conditions are “within the noise” of the sample
The other consideration is to increase the number of occurrences. This is a general rule in any backtest. The way to accomplish this on the software is to place a new trade every day. Opening a trade every day gets you as close as possible to the empirical reality of your options trading strategy.
Please note, that the fact of increasing the occurrences by placing a trade per day on the backtest, does not mean that you should replicate exactly that system on real-life trading. It means that your results have a larger statistical significance and a higher level of confidence.
Options backtesting requires massive amounts of data. An industry-standard practice is to use the Close or the “End Of Day” price to base the calculations. This results in an excellent value proposition that largely exceeds the shortcomings
Traders should be aware of this structural nature of the data when backtesting Options and avoid simulations that go beyond the capability of the data to ensure that the results have high statistical significance and that they can be replicated in real-life trading.
We find the best option strategy to trade Microsoft [Backtest study]
Is it possible to use the insurance industry as a model to construct a low-volatility, non-correlated investment strategy using options?