There was a free sample from ORATS at some point but I lost it, so I gotta ask. Looking on ORATS site I cannot find any free sample, only this: So my question is: The stock price (stkPx) provided with the samples is the actual price of the underlier at the moment of taking the snapshot, or it's the close price of the stock, which I can take from other providers, like Yahoo Finance? Because with the data downloaded from https://historicaloptiondata.com/ I'm getting the close price of the stock. Which is most of the times, obviously not the price at the moment of taking the sample, which greatly limits the usefulness of the data. It's very approximate. What sort of precise backtests can I do if the price of the stock is ambiguous in a range of +/- 1% from the close price? I try to deduce the actual value of the stock by using put-call parity and dividend estimates but it's not a very precise process. So is ORATS data precisely accurate or has the same limitation?
The stock price is snapped at the same time as the options are snapped 14 minutes before the close. Markets are much more representative at that time. At the close data if very suspect. We have now over 65 million backtests pre-canned and ready for people to review, and have been running backtests for over 10 years using this data. The historical quotes go back to 2007. We also have minute by minute data going back to August 2020. You can either download the data or utilize our storage and databases using our API. https://orats.com/intraday-data-api Take advantage of the cost savings for this data by using the link in my signature. Let me know if you have any questions.
Hi Matt, thanks for the answer, it's very good if the stock price is sampled at the same time with the options. Is there some free sample of the historical data available for sale?
@Matt_ORATS : I'm not sure if I can ask such question (please disregard if you don't feel like answering it), but does ORATS collect the options data themselves or it buys it from someone else (like Refinitiv) and resells it? Because the data I've seen is sampled haphazardly. Not only the provided stock price isn't in sync with the option prices, but the options of the same term are sometimes sampled at different times (often enough to be annoying). In real data you never see violations of put-call parity, but in historical data it's there and the only explanation is that subsets of the options are stored at different times. When one saves options data, they should at least employ this simple algorithm: 1) All data of an options TERM on a certain underlier has to be sampled at the same moment. Ideally all terms should. 2) Only sample the data WHEN THE UNDERLIER STOPPED MOVING FOR A CERTAIN TIME. If it moves like crazy, the ultra low latency market makers can barely keep up with it, I don't expect some cheap "realtime" marketdata subscription to do so. Therefore the option prices will be all over the place making data qvasi-unusable. Just wait some darn time, don't just blindly sample "now", which I bet it's what almost every vendor does based on how their data looks. So wait a couple of seconds and measure how long the underlier didn't move since the last time. If it's at least a second, then options prices have stabilized and can be stored along with the underlier price.
An example of what I'm talking about. Here's the SPY data I get on the day of 2007-Aug-17 at the option expiration 2007-Sep-28. Put prices should be increasing with strike, there's obvious inserts that have no place there.
Here's a visual representation of the crap I'm dealing with. Volatility is a mess and I'm getting parity violations all over the place. I'm subtracting the intrinsic value from in-the-money options, so I can compare left and right, that is calls and puts. By put-call parity, the fair price should be about the same and not go out of the bid-ask spread.
Here's the original option prices, I'm trying to run some backtests but they are only good as the data I'm using and right now it stinks. I'm bending over backwards trying to figure out rules to filter out the crap, because I can't go manually over every single data point out the millions I got. And everytime I think I finally have something usable, I'm looking over the results and the data that generated them and get this:
I mean for fuck's sake, I paid for data that I thought was usable and not total crap. I was supposed to concentrate on designing and testing trading strategies, not working a full time job as a data janitor
That made me chuckle... I am SURE you know this but if you paid for ANY service or product with a credit card... and you are unhappy with it as far as its honesty. Open a dispute with the credit card fraud dep't disputing a payment to any vendor who does shady selling of a product of ANY kind... I have never lost a legitimate dispute that i have opened with my CCard dept... Including against Microsoft and other large firms.
It's not so bad. With a little bit of processing the data can be corrected and afterwards it's fairly usable. I'll put up the code for data correction on GitHub at some point. At least the price is accessible for mere mortals born out of a woman. Compare this to buying from such as Bloomberg, Refinitiv & Co. At some point I found some data provider that wasn't one of these two and found the courage to ask for historical options data. Only $40,000 discounted price special for me. I also had access to $100,000 / year data subscription from Refinitiv on a contract that I worked. Not even options, just pure stocks, future and commodities. If you think the data was impeccable think again. Gaps in the data, prices that went to zero, etc.