I built a program once to scrape USDA reports at the instant they were released (+~50-200 ms, I think) and then "read" the report electronically, perform some calculations and issue trade instructions. It all took less than a second. The profit was statistically significant, but not practically significant given the bid/ask spread. In other words, I could prove 'academically' that my model for predicting the direction prices would go based on the contents of the report was correct, but I couldn't make (much) money on it. The sharpe was very low. I abandoned it. All I really proved to myself was that the market makers already did similar calculations and adjusted their bids and asks perfectly. As should be expected. By the way, for that report, the initial source was a specific location and link on the USDA website that I hard coded into my program in advance. They always used the same link. If you wanted to do something similar, I would suggest looking at the BLS website: https://www.bls.gov/cpi/ I doubt there's a good edge there, though, at least for retail.
So the releasing govnt depts (here BLS) generally release the data at their official time, so should be available at the same time as the earliest news outlet?
Yes, that was my experience with USDA. The releases were almost exactly to the millisecond (plus maybe 50-200 for transmission of data, time for my software to run, etc). At least in the case of the USDA, I suspect they have the report prepared and in their system, and then it is programmed to automatically release at the appropriate time. At the time I did it, I found this website handy to get a sense of transmission speeds: https://wondernetwork.com/pings/Chicago I also pinged various locations and of course determined the speed of the software I wrote. I don't actually remember the exact numbers anymore, but the bottom line was that it was rendered irrelevant due to bid/ask.
In one of my trading books the author tells the story of his days working as an Intern in the Gilt-Edged and Money Markets Division of the BOE Division when he and his colleague speculated on what the inflation number would be (in the early 90's, not now of course). He gave his guess, the colleague surmised it would be higher. The colleague asked on that basis what would the market reaction be? He said summoning all the theoretical knowledge he could muster outlining a textbook case he gave his expected implication for the Pound and Gilt. As it turned out his colleague was right about the inflation number but he was wrong about the market reaction. Perception is often different from reality. Who was this author? Jason Perl, Global Head of Fixed Income, Currencies & Commodities Technical Strategy @ UBS London.
You can try Reuters news feed via Metastock Xenith subscription for $99. But from my experience on news trading I can say that news trading is not stable for achieving consistent profits. And you cannot compete with algos, even with the fastest news feed available. There is no manual risk free trading here if you want to look for that kind.
That's very resourceful and the conclusion about the "market makers" is true as far as I have observed how markets react to news releases. The question is does an individual trader possess the resources to conduct such an analysis and use such an approach. Isn't that niche taken over by the highly knowledgeable, capable and adequately funded teams of lets say Ray Dalio, Jim Simons, Paul Tudor..... I've never even thought I could have an edge of that kind, because all the big guys have already diminished it's usability.
I agree with you mostly. However, I could be employed by one of those firms, but instead decided to go my own way. I am highly knowledgeable and capable -- but they are certainly better funded. They only reason I try some of these (crazy) strategies where I'll definitely lose to the better capitalized firms with lots of tech and brains in the room is because occasionally they do work (plus it's fun to do...perhaps that's the real secret). I had a strategy, once, that was consistently profitable until ~2015 when the expected value dropped to exactly zero. Sharply. A bar chart of historical PL on a monthly basis would have some up months, some down months, you know, a normal amount of variation in the profits and losses, with a positive expected value. Then all of a sudden, the winning months vanished, the losing months vanished and the profitability was zero (or very, very close -- gross profitability lower than transaction costs). My conclusion? Someone better capitalized found the same thing and eliminated the edge...completely. (But it was there for several years before that.) I liken the way I approach the markets to fishing. Every analysis of a new trading idea (/model) is another cast of the line. Many, many times I'll cast the line and get absolutely nothing. Sometimes I cast the line and I get a nibble that turns out to be nothing. Occasionally I get something small and, hey, that's better than nothing. And once in a while, I'll land on something sizable. When I do, I milk it for dear life until someone bigger discovers the same thing (or something adjacent to it) and eliminates the edge. But in the end, it's barely worth doing if you don't enjoy fishing, itself. If you do enjoy fishing, then even the failures aren't really a waste of time, because you're enjoying the process. It's just that you need to catch enough fish to have something to eat. Fortunately, I'm still in that category.
I'm impressed by your attitude. Wish you the best of luck, health and success with your approach. As to me, I've found that whatever fundamental or technical edge I've exploited throughout the years, it always came down to psychological edge supplementing it, mainly through the power of sitting out.
U.S. government cuts off early look at economic data for news media MAY 25, 2020 The U.S. Labor Department is ending its long-standing system of allowing news reporters to review government economic data, including high-profile reports on unemployment, retail sales and inflation, before the data is publicly released. The department for decades has given news organizations an advance look at such data to give them time to assess the numbers and prepare news reports to be distributed once the data is officially released. News organizations had long valued the 30 minutes to an hour the Labor Department had given them in what are known as "lock-up rooms" to review the data, question government officials about it and ensure that their articles were accurate and complete when the information was released. But William Beach, commissioner of the department's Bureau of Labor Statistics, said in announcing the decision last week that the past two months, when all such media lockups had been suspended over the coronavirus pandemic, had shown that the absence of the lockups did not cause much of a time delay or a decline in the accuracy of news reports. The Labor lockup room had been used to give reporters not only economic data that the department itself produced but also highly watched reports from the Commerce Department, including quarterly estimates of the nation's gross domestic product. Early this year, even before the coronavirus forced the suspension of the lockups, the department had announced its intention to remove computers from the secure room in which reporters had been given an early look at economic reports. The department said it would still allow reporters to have time to study the reports before they were released via the government's website, but they would have to do so without the benefit of electronic devices. The Associated Press and several other news organizations protested that that decision would make the economic data releases less secure and less useful to the public. The news organizations also asserted that the department's new policy would fail to meet its stated objective of leveling the playing field between high-speed investment firms, which trade off the data in mere seconds, and the general public. Since then, the two sides had been engaged in negotiations toward a potential compromise. But in his letter Tuesday, Beach said the performance of the government's website over the past few months demonstrated that the lockups were not needed at all — whether or not the coronavirus remains an ongoing threat for people who work in close proximity. "The BLS website can serve all interested users in the seconds after release time with little or no degradation in response time and a negligible error rate," Beach said.