Twelve months of running a system that takes one window a day gives you roughly 250 trades. The backtest behind the Algomyth engine has 1,164. So a full, patient, disciplined first year hands you about a fifth of the evidence the report is built on, and costs you real money to collect.
That is the honest frame for a first year, and almost nobody sets it. The year is not where you find out whether the system is good. It is where you find out whether you can run it.
The first month is installation and rehearsal. NinjaTrader 8 has to be installed, connected to a real-time futures data feed and left running through a few sessions before anything is at stake. The strategy has to be imported, compiled and watched in simulation so you can see what it does at 9:30 AM ET rather than reading about it.
Most people compress this into a weekend and regret it. The things that go wrong in month one are dull and specific: a data feed that disconnects, a machine that sleeps at 9:20, a strategy enabled on the wrong instrument, a chart on the wrong contract month after a rollover. All of them are cheap in simulation and expensive in a funded account. A week-by-week version is in getting started in 30 days.
Your first year is not a test of the strategy. It is a test of whether you leave a tested strategy alone. Those are different questions, and only one of them is under your control.
A 50k evaluation costs roughly $100. A 100k runs about $250 and a 250k about $400. That is the cost of an attempt, not the cost of a pass, and the difference matters.
Budget for more than one attempt. Not because the system is unreliable, but because a rule set whose target sits at three times its stop will sometimes open with a losing stretch, and an evaluation has a hard drawdown limit that does not care whether the strategy would have recovered. Two or three attempts in a first quarter is normal. Treat each fee as the price of an attempt and it stops feeling like a verdict. What that looks like with a bot running it is in passing an evaluation with a bot.
| Stage | What you are actually doing | What it costs |
|---|---|---|
| Month 1 | Install, connect data, run in simulation and replay | Data subscription, your time, no trading risk |
| Months 2 to 3 | Evaluation attempts on a 50k | ≈$100 per attempt, and expect more than one |
| Months 4 to 6 | First funded account, first real drawdown | Commissions, possibly another evaluation fee |
| Months 7 to 9 | Routine, and resisting the urge to change things | Commissions, data, hosting if you use it |
| Months 10 to 12 | First payouts, first real decisions about size | Profit split, taxes, and the fees already spent |
Set the account size, the contract count and the commission rate you would actually use, then read the published report against them.
Somewhere in here the equity curve goes down and stays down for longer than feels reasonable. This is the part of the first year that decides everything else.
The backtest shows a worst peak-to-trough decline of $5,273 on a mid-size account, with the target set at three times the stop, so break-even sits at a 25% win rate and the system does not need to be right most of the time. Losing runs are arithmetic, not malfunction. Knowing that in advance helps less than you expect, because watching it happen with your own money is not the same as reading the number. That gap is the subject of the psychology of watching a bot lose.
What you actually do here is nothing, which is much harder than it sounds. The honest version of what a drawdown feels like and how to size for it is on the risk page.
Hypothetical performance. These figures come from a backtest of the Algomyth engine over historical data, not from a live account. Simulated results are prepared with hindsight, carry no financial risk, and cannot fully reflect real execution, slippage or liquidity. Past performance, real or simulated, does not guarantee future results.
If the system is working, this stretch is uneventful. A trade fires at 9:30, the stop and target are already in the market, and by 10 AM the day is over. There is nothing to watch and nothing to decide.
Boredom is when people break things. The urge is always framed as improvement: skip the days the news looks bad, widen the stop because that one was so close, add a filter you noticed last Tuesday. Each change swaps a rule set with 1,164 tested trades behind it for one with none. You are not refining the system, you are trading a new and untested one while telling yourself otherwise.
Every parameter you change mid-year resets your sample size to zero, and you keep the fees.
A successful first year does not look like the headline number. In backtest, an average year on a $50,000 account at the Medium sizing level comes to $18,814 net, a little over $1,500 a month. That is an average across a full year, not a schedule, and a first year in particular carries a month of setup, one or more failed evaluations and the fees attached to them. Where that figure comes from and what it excludes is in how much a trading bot actually makes.
A realistic definition of a good first year has four parts, and only one of them is a dollar figure:
Anything left over is profit. A year that ends roughly flat after fees with all four boxes ticked puts you in a better position than someone who made money by tripling their size in month three. One of them has a process. The other had a good quarter.
Nearly every failure fits one of these, and none of them are software problems.
Nobody can tell you that, and anyone who does is selling something. What can be said is that a first year carries costs a mature year does not: setup time, evaluation attempts that fail, and a sample too small to confirm much. The figures on this site are hypothetical results from a simulation over historical data, and past performance, real or simulated, does not guarantee future results. Plan the year so a flat outcome is survivable.
Assume more than one. A 50k evaluation runs about $100, so budgeting two or three attempts in the first quarter means setting aside $200 to $300 of pure cost. Hard drawdown limits close accounts that a strategy would otherwise have recovered from, so plan on paying that fee more than once. Fees are money you spend whether or not the account ever pays you back.
Six months is around 125 trades against a tested sample of 1,164, so a gap of that size is well within normal variation and is not by itself evidence that anything is broken. What you check instead is mechanical: is the system trading the right instrument, at the right time, at the tested size, with fills roughly where the test assumed. If the answers are yes, the difference is variance. If one of them is no, you found a configuration problem rather than a strategy problem.
Run it in simulation for a few weeks first, every time. Simulation will not tell you much about the edge, because fills are optimistic, but it will tell you whether your machine stays awake, whether your data feed holds, and whether the strategy behaves the way you expect at 9:30 AM ET. Those are the failures that ruin evaluations, and they cost nothing to find in advance.
After you have lived through a drawdown at the current size and kept the system running, not after a good month. Account rules cap the maximum anyway, and each tier has its own ceiling. Going past the sizing level whose figures you actually read means accepting a drawdown you have not experienced. Doubling contracts doubles both halves of the outcome.
Everything we write comes back to one idea: verify, then decide. The live terminal exists for exactly that.
Unlock Algomyth Live