I define algorithmic trading as a trading method in which securities firms and other such companies refer to users’ order book information and generate direct profits.
For example, it is a trading method in which an order is placed just before a large order hits the order book.
System trading, on the other hand, differs from algorithmic trading in that it uses logic to carry out reproducible trades based on past stock price data for individual stocks.
Get historical stock price data for individual stocks
This site also sells stocks, but you can obtain historical stock price data for individual stocks.
By obtaining push data using the au Kabucom Securities API, you can obtain highly accurate data.
The amount of data for each stock per day is nearly 100MB for some stocks, with some stocks exceeding 150,000 records.
The data starts at 9:00 on a business day and ends at 15:30, so it takes 5.5 hours.
150,000 / 5.5 / 60 / 60 = a massive amount of data, or about 7.5 times per second.
Inventing the logic
I think the ideas are endless, including not only market depth information and trading volume, but also tick momentum, such as jumping, position and bet control such as difficult flats and pyramiding, scraping references to the Nikkei Financial Yearbook, and gap strategies.
We create calculation formulas using programs that will actually trade these.
Past stock price data for individual stocks is read in order and a simulation is performed, and the ordering methods for this logic include market orders, limit orders, stop-loss orders, and reverse orders.
I have experienced this many times myself, and for example, I have calculated the buy price and sell price in reverse for market orders, which resulted in a huge win in the calculations, and there are also cases where the limit price “if it hits” logic doesn’t actually hit at the planned price, or the order doesn’t get executed at all…
Even if you win in a simulation with this logic, it is generally only important for about 50% of the total.
This is the first hurdle, because if you can’t win with logic, you will never be able to win.
Creating a stock trading program
Just because it’s a program doesn’t mean that everyone will get the same results.
It’s not enough to just have programming knowledge; for example, even if the logic says to place a market order, there may be cases where limit orders will produce better results.
Programmers who value error handling may slow down the program’s behavior, and unless you can read the logic, it will be difficult to write optimal, factorized code.
The au Kabucom Securities API specifications allow you to monitor and trade tick data for up to 50 stocks with one account, which is one of the advantages of system trading.
Even in this case, speeds may slow down and position management may become difficult, so unless you know the rules of stock trading to a certain extent, it will be difficult to write optimized code.
There is little information available online about the au Kabucom Securities API, and the API specifications seem a bit quirky, but in reality, you will be testing in a live environment, which is quite difficult.
This section is roughly 20% important.
I think that if the person who comes up with the logic asks an outside programmer to create the logic, even if it is a winning logic, there may be cases where the logic is not communicated and the team actually loses.
Actual Transactions
Once the logic and program are more or less complete, it’s time to make actual trades, but here too, it doesn’t work the way we expected at first.
Is it because of a miscalculation in the logic, an inadequate program optimization, a problem with the trading environment of the computer or line, or a problem with the order or settlement method of the trade…
We make a final decision through trial and error.
This is also done in a live environment, so there is a certain amount of cost involved, and it is difficult to overcome, so I think it accounts for about 30% of the overall importance.
The logic program checks the discrepancy between the theoretical value and the actual value of actual transactions, and compares and verifies the results based on the tick data the following day.
Capital Efficiency Calculation
Once you have finally completed a winning system, you then optimize the efficiency of your funds.
Unlike FX, stocks have the concept of an order book, so if you order too many units, the results will be negative in spot trading.
You calculate the expected value in daily and monthly interest rates, and also calculate drawdown and profit factors to ensure optimal fund management.
If you start with 100 coins as a benefit, you will not be able to operate with 200 coins unless your funds double, but if you start with 1,000 coins as a benefit, you will be able to operate with 1,100 coins if your funds increase by 10%, so the capital efficiency is completely different.
Maintenance
I think you could win forever if you used a universal logic that was not dependent on the characteristics of the stock or other things, but I don’t think this is possible with stock trading.
To take an extreme example, you might be winning until yesterday and then start losing from today, but the look of the chart will still change depending on the movements of other traders participating in the stock.
You need to test it daily and prepare stocks that are compatible with your logic.
Certainly, when you write out the steps like this, you can see that there are many difficulties.
Perhaps it’s inevitable that there is not much information available online.
However, I think system trading has already become the standard around the world.
As Japanese people, I would like to create a culture of system trading that can compete on the world stage, before we are taken over by black ships like the current IT industry.