Petko AleksandrovRisk warning. Index CFDs and futures are leveraged trading products. Losses can exceed deposits with some account types, and automated execution does not reduce market exposure. Everything below is educational material rather than personalized guidance, and the results described are historical.
The Direct Answer
Algorithmic trading on the US30 means software opens, manages, and closes positions on a DJIA-linked instrument according to rules written in advance. No discretion at the moment of execution. The code watches price, applies its conditions, and sends orders through the trading platform whether you are at the screen or asleep.
This page covers a specific workflow rather than theory alone: how several trading strategies get tested on a demo account first, screened using trade count, profit factor, net points, and win rate, then promoted to live capital only after passing those filters. Screening criteria for each algorithmic trading candidate, performance data, statistical weaknesses of the method, and the costs involved all appear below.
One caution before anything else. The filters described here are my operational habits, refined across years of running algorithmic trading accounts. They are not validated standards, and I will point out where the evidence behind them is thinner than I would like.
What Algorithmic Trading Actually Means
Strip away the marketing and you are left with code that tells a computer when to buy and sell securities. Conditions get defined, the machine checks them continuously against incoming price data, orders go out automatically.
Institutional desks have operated this way since the 1970s, when program trading first moved large baskets of stock through exchange systems. Retail access to algorithmic trading arrived much later, through platforms like MetaTrader that let individuals attach an Expert Advisor to a chart and let it work. More recently, AI-driven algorithmic tools have entered the conversation, though most retail algorithms including mine remain rule-based rather than adaptive.
Three layers exist in any algo trading setup:
| Layer | Function | Who controls it |
| Signal logic | Decides direction and timing | Strategy developer |
| Algorithmic execution | Converts signals into live orders | Platform and broker |
| Risk framework | Sets exposure, sizing, and shutdown rules | Account holder |
Most traders new to algorithmic trading obsess over the first layer. My experience across dozens of trading accounts suggests the third one determines survival.
Which Instrument Gets Traded
Instrument definition matters more than almost anything else on this page, and the original 2020 article skipped it entirely, which I regret.
The index itself cannot be bought. What you actually hold is a derivative referencing it, and choosing between products changes pricing, trading costs, session hours, and margin requirements substantially.
Before running anything, open your broker’s contract specification and write down all of the following. Every one of them changes your results:
| Field | Why it matters |
| Exact broker symbol | US30, DJ30 and WS30 are not always the same product |
| Product type | CFD, futures, ETF or synthetic, each priced differently |
| Contract size | Two brokers can differ by a factor of ten |
| Minimum volume | Determines the smallest risk you can take |
| Account and quote currency | Conversion affects every result you record |
| Trading session hours | Index products close, unlike currency markets |
| Typical spread in points | The main cost on short holding periods |
| Commission per lot | Sometimes zero, sometimes significant |
| Overnight financing | Charged separately for long and short positions |
Two brokers offering apparently identical symbols can differ by a factor of ten in contract size. Copying settings from somebody else’s trading account without checking the specification window ranks among the more expensive mistakes available to a new algorithmic trader.
What “Points” Means on an Index
FX Blue and similar reporting tools display a pips column regardless of asset class, which creates confusion the moment you move from currency markets into equity indexes.
On a major currency pair, a pip is a standardized fractional movement. On the US30, that same column usually reports index points, where one unit of movement equals one point of the underlying. Some brokers subdivide further. Others quote to a single decimal place and count that decimal as the reporting unit.
Consider figures from my own demo record: 694.2 net points recorded alongside $60.84 in net profit. Those two numbers look wildly inconsistent until you work out what they imply together, which is roughly nine cents of profit per point of favorable movement. That ratio tells you the position sizing was extremely small, and demonstrates why any single trade count or point total in isolation mean very little.
Before reading any algorithmic trading report on an index:
- Find the tick size in the contract specification.
- Find the tick value in your account currency.
- Multiply movement by value to get money, always.
- Treat point counts as a measure of accuracy rather than profitability.
I still like watching the points column, honestly. It shows whether an algorithm is catching real moves or scraping noise. But money is what leaves the account.
How the Algorithms Place Orders
Strategies discussed here were generated in EA Studio and run as Expert Advisors in MetaTrader. Specifics differ between algorithms, though any one of them should be able to state:
- Entry conditions: Which indicators, which thresholds, which timeframe.
- Exit conditions: Target, protective level, indicator reversal, or time-based closure.
- Position sizing: Fixed volume or a percentage calculation tied to balance.
- Concurrency limits: How many positions any single algorithm may hold at once.
- Session filters: Whether entries are permitted outside main trading hours.
- News handling: Whether high-impact events pause the algorithm.
Publishing those specifications separately from performance figures is something I’d encourage for anyone sharing results, and something the earlier version of this page failed to do. A balance curve without the trading rules behind it is decoration.
The Testing Sequence
Three stages, in order, and skipping any of them tends to end badly.
- Historical simulation: Useful for eliminating obvious failures and understanding how an algorithm behaves through different market regimes. Backtest data describes the past only. Every equity curve I have generated looked wonderful before it met live spreads.
- Demo forward testing: Screening happens at this stage. Trading algorithms run on a demo account under current conditions, generating fresh trades that nobody optimized against, and I watch which ones perform right now rather than which performed beautifully back in 2019.
- Live deployment with small size: Promoted strategies start with minimal exposure. Slippage, requotes, and financing charges appear at this point and nowhere earlier.
Why bother with the middle stage? Because every algorithm cycles through profitable phases and stagnant ones. Balance charts from the generation stage show it clearly: growth, then a drawdown, then flat, then growth again. Perfectly normal, and largely unpredictable. My aim is holding positions in algorithms that appear constructive at this moment, using current market behavior as the evidence rather than a simulation from years back.
Whether that aim is achievable is a separate question, addressed below.
Portfolio Performance Data
A demo snapshot from the original 2020 material recorded $60.84 net profit across ten deployed strategies, with 694.2 net points, tracked through FX Blue. Those two figures are all that survives from that account.
Which is the point of this section, really. Two numbers cannot support any conclusion, and I’d rather show you what a usable record looks like than dress up an incomplete one:
| Metric to record | What it tells you |
| Result type | Demo, backtest and live results are not comparable |
| Test period and calendar days | Short windows measure luck, not edge |
| Initial balance | Turns currency figures into percentages |
| Total trades | The single best indicator of whether the sample means anything |
| Winners and losers, with percentages | Foundation for win rate |
| Portfolio profit factor | Gross gains against gross losses |
| Maximum equity drawdown | The worst moment, which averages hide |
| Average result per trade | Reveals whether costs are eating the edge |
| Largest gain and largest loss | Exposes reliance on outliers |
| Longest losing run | What you would have needed to sit through |
| Average holding time | Determines exposure to overnight financing |
| Total costs paid | Frequently larger than beginners expect |
Keep that trading record from your first demo account onward. Reconstructing it later is impossible, as I found out.
Screening Method One: Profit Factor
Wait for a minimum of five closed trades from an algorithm, then check whether its profit factor sits above 1.2.
Profit factor divides gross gains by gross losses. Anything above 1.0 means the algorithm made more than it lost across the sample. A reading of 1.2 means twenty percent more.
Simple enough. Five trades is a screening trigger, not proof of an edge, and I want to be blunt about that because the earlier version of this page implied more confidence than the data supports.
Screening Method Two: Net Points
Same minimum count, different measure. Trading strategies clearing more than 100 net points get considered.
From one session’s data: the leading algorithms showed 358, 204, 194, 180, and 178 points respectively. Six passed the threshold that day. Combined with the profit factor requirement, only three cleared both.
Applying the point filter requires knowing your broker’s convention, per the earlier section. A hundred points on one specification might represent ten times the money it represents on another.
Screening Method Three: Win Rate
Ten closed trades minimum, with at least seven of them profitable.
This filter has an obvious weakness, and naming it beats letting a reader find out expensively. Win rate ignores magnitude entirely. An algorithm winning seven times out of ten while losing catastrophically on the remaining three loses money overall, and it passes this screen comfortably. Pair the measure with average win against average loss, or with profit factor, before acting on it.
| Screening approach | The rule | What limits it |
| Profit factor | Five closed trades, above 1.2 | Small samples produce unstable readings |
| Net movement | Five closed trades, above 100 points | Meaningless without contract specifications |
| Win rate | Seven winners from ten | Ignores reward against risk entirely |
You are welcome to build your own trading filter. Mine developed through repetition rather than research, which is worth knowing before adopting it.
Why Five Trades Prove Almost Nothing
Statistical honesty section, and probably the part most algorithmic trading content skips, though the most important part of this page.
Five closed trades is a tiny sample. A coin-flip algorithm with zero genuine edge produces a profit factor above 1.2 across five outcomes surprisingly often, purely through variance. Ten trades barely improves matters. Establishing a stable win rate typically requires hundreds of observations, not dozens.
Several other problems deserve naming:
- Chasing recent performance: Selecting whatever performed well last week may simply mean selecting whatever got lucky last week.
- Switching after the damage: Removing an algorithm once it falls below threshold means the losses already happened. You capture the decline and potentially miss the recovery.
- Multiple comparisons: Screening twenty candidates and promoting three guarantees that some passers cleared the bar by chance alone.
- Regime change: Market conditions suiting an algorithm in one quarter can vanish in the next without warning.
- Deployment costs: Rotating strategies incurs spread and commission on every cycle.
Does the filter improve outcomes out of sample? Testing that properly means running the screening rule itself as a strategy across a long historical record and comparing against simply holding everything. I have not run that test. Nobody I know in this space has published one either. Until somebody does, treat the algo trading workflow here as a plausible heuristic rather than a demonstrated improvement.
Diversification Needs Measuring, Not Assuming
My longstanding rule has been running at least three algorithms live rather than one, on the reasoning that when several are losing, others compensate. Trading a single strategy leaves you fully exposed the moment it enters a stagnant phase.
Sensible instinct. Incomplete reasoning, though.
Several strategies trading the same index during identical hours using similar indicator families are not really independent. They can all buy the same breakout and all lose together. Genuine risk spreading requires evidence, which means reporting:
| Measurement | Why it matters |
| Return correlation between algorithms | Values near 1.0 mean no real spreading of exposure |
| Overlapping entry timestamps | Simultaneous trades concentrate risk |
| Shared indicator families | Similar logic produces similar failures |
| Maximum combined exposure | Total position size when everything fires at once |
| Combined portfolio drawdown | Deeper than any individual figure suggests |
| Directional bias | An all-long portfolio is a leveraged long position |
I have never published a correlation matrix for these strategies, and I’m not going to pretend otherwise. Until those numbers exist, “many strategies” describes a count rather than a risk property.
Sizing, Exposure, and Shutdown Rules
Position sizing deserves more attention than strategy selection, though it receives roughly a tenth of the discussion in most algorithmic trading material I read.
- Decide maximum acceptable portfolio drawdown before deploying anything.
- Size individual trades so worst-case simultaneous exposure stays inside that limit.
- Account for the possibility that every algorithm opens at once, because occasionally they will.
- Set a hard equity floor at which everything stops, defined in advance.
- Keep leverage modest on index products, where a hundred-point gap costs real money quickly.
Overnight financing applies to positions held past the daily rollover, and some strategies hold through the night. Those charges accumulate quietly and appear in your statement rather than your platform’s profit column.
Adding, Pausing, and Removing
Monitoring continues after promotion to a live trading account. Live performance gets checked against the same criterion used for selection, so an algorithm whose profit factor drops under 1.2 comes off the trading account.
Consistency between selection and removal rules matters. Promoting on one basis and holding through decline on another is how portfolios quietly fill up with underperformers.
A useful practice for anyone starting out: open a second demo account and treat it exactly as though it held real money. Place the three passing algorithms there, run both accounts in parallel, and rotate between them according to whichever method you have chosen. Practicing the process costs nothing except patience. Rushing into live trading before you trust your own workflow costs considerably more, as plenty of traders learn the expensive way.
Demo Against Live Trading
Results diverge between the two environments, sometimes substantially:
- Demo servers often fill at the quoted price with no slippage whatsoever.
- Live trading introduces latency between your terminal and the broker.
- Spreads widen around economic events and at session transitions.
- Requotes and partial fills occur on live accounts and rarely on demo.
- Financing charges may be simulated inaccurately or omitted entirely in demo conditions.
- Order queue position matters at scale, though less so for small retail volume.
Expect live trading figures to come in below demo figures. Budgeting for that gap in advance beats being surprised by it.
Frequently Asked Questions
Does automated trading need a VPS?
For any algorithm running outside your waking hours, effectively yes. A virtual private server keeps the platform connected when your computer sleeps, updates, or loses internet access, which matters because an unmanaged open position on a leveraged index product can move against you overnight. Hosting near your broker’s servers also reduces latency slightly. Many brokers provide free hosting once account balance passes a threshold, so ask support before paying separately for a provider.
How does this compare to futures rather than CFDs?
Exchange-traded contracts of this kind clear centrally with centralized pricing, standardized specifications, and transparent volume data, while CFDs are bilateral agreements with your broker. Contract sizes on the exchange-traded version are considerably larger, which raises the minimum capital required. Financing works differently too: futures embed carrying costs in the contract price rather than charging daily. Neither structure is universally better, though the exchange route suits larger trading accounts and CFDs suit smaller ones with fractional sizing.
Can these algorithms trade around major economic announcements?
Mine are configured to avoid them, and I would suggest the same for anyone starting out. Releases such as US employment figures, inflation prints, and Federal Reserve decisions produce sudden gaps where orders fill well away from expected levels. Automated logic cannot assess whether a spike reflects genuine repricing or momentary illiquidity. Filtering by economic calendar, or simply pausing during scheduled events, removes an entire category of unpredictable trading outcomes.
What profit factor should I actually target?
Higher readings look attractive, though a value calculated across few observations means little regardless of size. Across several hundred closed trades, sustained figures between 1.3 and 1.6 would represent solid performance for an index algorithm after trading costs. Be suspicious of anything above 2.0 from retail sources, since such numbers usually come from short samples, curve-fitted parameters, or backtest data excluding realistic spread assumptions. Sample size deserves more attention than the headline value.
Is an API connection better than platform-based execution?
A direct exchange or broker API offers faster order routing and finer control, which benefits high-frequency trading where milliseconds affect outcomes. Retail traders holding positions for hours gain almost nothing from that speed. Platform execution through MetaTrader handles the same job with far less development work, no server maintenance, and a much shorter path to running something live. Complexity should follow need, and most people never reach the point where that kind of connection solves a real problem.
How much capital does index algo trading require?
Enough that minimum position sizes represent a sensible fraction of the balance, which depends entirely on your broker’s contract specification. If the smallest available volume risks several percent of your account on a routine adverse move, that account is too small for the instrument regardless of algorithm quality. Calculate backward from worst-case exposure across everything deployed simultaneously, then add a buffer, rather than starting from a number that feels affordable.
Do these methods work on other markets?
The screening framework transfers reasonably well, since profit factor, sample count, and win rate apply to any instrument. Threshold values need adjusting, though. A hundred points means something entirely different across currency markets, commodities, or a stock market index, and holding periods vary with volatility characteristics. Testing thresholds separately for each market beats assuming settings developed on one asset carry across to another. Options and futures introduce further differences worth studying independently.
What happens if every algorithm loses simultaneously?
It happens, and planning for it separates a bad month from a closed account. Correlated failure typically occurs during regime changes: a trending market turns choppy, and strategies built for continuation break at once. Your protection is the predetermined equity floor mentioned earlier, plus position sizing that assumes simultaneous exposure rather than average exposure. Nobody avoids these periods entirely, so surviving them with capital intact becomes the realistic goal.
Disclosure: Material here is educational and does not constitute a recommendation to buy or sell any financial instrument. Automated trading reduces emotional decision-making during execution, which is genuinely useful, though it introduces operational risks of its own: connectivity failures, coding errors, broker-side changes, and unattended exposure among them. Consider guidance from a regulated financial professional before committing capital to any trading system.
