Algorithmic trading software
What does most algorithmic trading software actually do?
Almost every product sold as algorithmic trading software does one job: it executes rules a person wrote. You define the conditions, it watches for them, and when they occur it acts. The judgment happened once, at the moment you wrote the rule. Everything after that is execution.
A fixed-rule arrangement has a known weakness. Markets change and rules do not. A rule that worked in one regime keeps firing in the next, and the software has no way of noticing, because noticing was never part of its job.
What counts as algo trading software?
Algo trading software is the common shorthand for algorithmic trading software, and it covers four quite different things. Knowing which one you are looking at saves most of the evaluation.
Execution algorithms
Software that takes an order someone has already decided on and works it into the market in pieces, to reduce cost or impact. It decides how to trade, never whether to.
Rule engines
You write the conditions, it watches for them and acts. Most retail algo trading software is this. It is reliable, predictable and exactly as good as the rules.
Research platforms
Tools for writing and backtesting strategies, sometimes with live trading attached, often sold as an algorithmic trading platform. Powerful if you code, and mostly a workshop rather than a finished product.
Self-grading systems
Software that decides for itself and scores its own decisions against what the market did. The smallest category, and the only one that notices when it has stopped working.
What did we build instead?
It grades every decision against what happened next
TradeAgentic records what it decided, then finds out what the market actually did, and scores itself. That record is the input to later decisions. A system that cannot tell whether it was right cannot improve, and most algorithmic trading software cannot.
It records refusals, not just trades
Most systems log fills. Ours logs the candidates it declined and why, which is the half of the record that tells you whether it is missing things it should have taken. You can read it.
It moves capital on measured results
Multiple strategies compete for one pool of capital, and allocation follows measured outcomes rather than conviction. You set which strategies and asset classes are in scope; nothing asks you to pick between them trade by trade.
It runs entirely on your machine
No RLG server decides anything. Your broker keys stay in the macOS Keychain on your own hardware, and no market or account data is sent to RLG. Our outage cannot become your outage.
What should you check before believing any of it?
Rerun the research yourself
The backtester has a real cost model and options repricing, and validation designed to reject findings that do not hold out of sample. You can reproduce it from one command on your own machine. A backtest you cannot rerun is marketing, ours included until you have run it.
Read the decision record before you fund anything
Run it on a paper account and read what it did and declined. If the record does not persuade you, no sales page should.
Check where the protection sits
Protective orders rest at the brokerage, not inside our software. That means a crash, a reboot or us going out of business does not leave a position unprotected. Ask every vendor this and accept only a specific answer.
Read what it will never do
The governing charter sets out the things the system is not permitted to do, and it ships with every license tier. Ask any vendor for their equivalent before you buy, not after.
What is the life of an algorithmic strategy?
Every strategy, whoever or whatever designs it, should pass through the same stages, in this order. Skipping one is how a good backtest becomes a bad account.
- An idea, stated precisely enough that two people would code it the same way.
- Research on historical data, with costs modeled honestly.
- Validation on data held back from the research, used once.
- Paper trading, long enough to see ordinary and unusual days.
- Live trading at a size smaller than feels necessary.
- Ongoing comparison of live behavior against what the research promised.
- Retirement, when the evidence says it has stopped working.
The last stage is the one most algorithmic trading software has no concept of. Rules keep running until a person notices. Deciding when a strategy has stopped working is itself a job, and it is usually left to you.
A checklist before you buy algorithmic trading software
- Where it runs, and who holds your broker credentials.
- Whether protective orders rest at the broker or only inside the software.
- Whether you can rerun the backtests yourself with realistic costs.
- Whether it runs on a paper account for as long as you want.
- Whether it records what it declined as well as what it traded.
- Whether its figures reconcile to your broker's statement.
- What it will never do, written down.
- What you keep if you stop paying.
Where it fits and where it does not
It suits operators who intend to run it
TradeAgentic is licensed to people and firms who will run it themselves, on their own account, with their own keys. It is not a managed service and there is no version where we trade on your behalf.
It is not for someone wanting a signal feed
If you want alerts to act on manually, this is more system than you need. The value here is that judgment is continuous and does not wait for you.
It runs US equities, crypto and multi-leg options
Equities on the consolidated tape, crypto around the clock, and defined-risk options spreads. If your market is outside that, it is the wrong tool today.
What you actually take delivery of
A native application, not a login
The application installs on your own Mac and runs there. There is no account to log into and no dashboard that stops working when a subscription lapses.
The research, not just the result
Every license tier includes the reproducible research and the owner's guide, because a system you cannot audit is one you should not run. That is a deliberate commercial choice: it makes us easier to check and harder to take on faith.
Books that reconcile to your broker
Realized profit and fees are rebuilt from the brokerage's own record rather than from what the software intended, and positions and protection are reconciled at startup before any new order. This is the unglamorous part that decides whether the numbers on your screen are real.
Where algorithmic trading software usually fails
The rules keep firing after the regime changes
Rules firing after the regime changed is the single most common way these systems lose money, and it is invisible from inside a product that has no concept of being wrong. Ask any vendor how their software detects that a rule has stopped working, and treat vagueness as the answer.
The backtest assumed fills nobody gets
Frictionless fills, mid-price on spreads and no market impact turn mediocre systems into impressive charts. A cost model that reflects what you actually pay produces less flattering and far more useful numbers.
Nobody reconciles until something breaks
Software that reports what it intended rather than what the broker recorded drifts from reality quietly, and the gap is usually discovered at the worst moment.
Common questions
- What is the difference between algorithmic trading software and this?
- Conventional algorithmic trading software executes rules a person wrote and has no mechanism for noticing when those rules stop working. TradeAgentic grades every decision against what the market subsequently did, records refusals as well as trades, and moves capital between strategies on measured results rather than on the rule you set.
- Do I need to write code to use algorithmic trading software?
- No code, and no picking trades. What you set is the envelope: the risk limits, which asset classes are in scope, and which strategies are available to the desk. Inside that envelope it decides continuously on measured results rather than asking you which one to run today. Changing the system itself rather than operating it is the Institutional tier.
- Can I verify the backtests before buying?
- Yes, and you should. The research is reproducible from one command on your own machine, with a real cost model and options repricing rather than frictionless fills. Treat any backtest you cannot rerun yourself as marketing, whoever is showing it to you.
- Is algorithmic trading software worth it?
- It depends entirely on whether the system can tell you why it did something and whether it improves. Software that executes rules you wrote is worth roughly what those rules are worth, and rules decay. Judge it on the decision record and reproducible research rather than on a returns curve, and run it in paper for months before funding it.
- What is algo trading software?
- Algo trading software is shorthand for algorithmic trading software: programs that trade by following an algorithm rather than a person placing each order. It covers execution algorithms, rule engines you configure, research platforms for building strategies, and a small number of systems that decide for themselves and grade the results.
- Is algorithmic trading legal for individuals?
- Using software to place orders in your own brokerage account is generally permitted in the United States, subject to your broker’s account agreement and API terms, securities laws, margin rules and state law, including state rules on digital-asset trading. This is general information, not legal advice; confirm your position with your broker and a lawyer. Individuals can use algorithmic trading software on their own brokerage accounts. The same rules apply as for manual trading, and your broker applies its margin and order checks. What matters more is that the software cannot exceed the limits you set.
- Can algorithmic trading software run without me watching?
- Most can place orders unattended. Fewer can be trusted to. That needs protection resting at the broker, a hard daily loss limit, reconciliation with your broker, and a refusal to trade when its data or its own checks fail.
TradeAgentic is an autonomous trading desk for macOS, licensed to operators and firms who intend to run it themselves.
Request licensing terms