"Minimum deposit: $100" appears on a lot of EA product pages. What that number tells you is that the EA can open a trade. It doesn't tell you whether the account can survive the strategy. Those are different questions, and the second one comes down to arithmetic.
This post works through that arithmetic: position sizing, the 0.01-lot floor, losing streaks, drawdown, risk of ruin and margin. Every number below is an illustrative example computed from stated assumptions. None of it is a backtest or a performance claim. Plug in your own EA's statistics.
Assumptions used throughout (illustrative): EURUSD-style symbol, pip value $10 per 1.00 lot ($0.10 per 0.01 lot), broker minimum lot 0.01, lot step 0.01, lots rounded down.
1. Position sizing: the core formula
For a fixed-fractional EA risking a percentage of equity per trade:
Lots = (Equity × Risk%) / (StopLoss_pips × PipValue_per_lot)
Then the broker constraint applies:
Lots_actual = max(MinLot, floor(Lots / LotStep) × LotStep)
ActualRisk% = Lots_actual × StopLoss_pips × PipValue_per_lot / Equity
That max(MinLot, …) is where small accounts get into trouble.
Worked table: $500 / $2,000 / $10,000
| Equity | Target risk | SL (pips) | Raw lots | Actual lots | Actual risk |
|---|---|---|---|---|---|
| $500 | 1% | 20 | 0.025 | 0.02 | 0.80% |
| $500 | 1% | 50 | 0.010 | 0.01 | 1.00% |
| $500 | 2% | 20 | 0.050 | 0.05 | 2.00% |
| $500 | 2% | 50 | 0.020 | 0.02 | 2.00% |
| $2,000 | 1% | 20 | 0.10 | 0.10 | 1.00% |
| $2,000 | 1% | 50 | 0.04 | 0.04 | 1.00% |
| $2,000 | 2% | 50 | 0.08 | 0.08 | 2.00% |
| $10,000 | 1% | 20 | 0.50 | 0.50 | 1.00% |
| $10,000 | 1% | 50 | 0.20 | 0.20 | 1.00% |
| $10,000 | 2% | 20 | 1.00 | 1.00 | 2.00% |
At $500, rounding already pulls the 20-pip case down to 0.8%. More important is that you have almost no granularity. One lot step (0.01) is worth 0.2–0.5% of the account per trade, so the sizing can't adapt smoothly as equity moves.
2. The 0.01-lot floor: the real minimum deposit
Rearrange the formula to get the smallest account at which 0.01 lot still matches your target risk:
MinEquity = (MinLot × StopLoss_pips × PipValue_per_lot) / Risk%
| SL (pips) | 0.5% risk | 1% risk | 2% risk |
|---|---|---|---|
| 20 | $400 | $200 | $100 |
| 30 | $600 | $300 | $150 |
| 50 | $1,000 | $500 | $250 |
| 100 | $2,000 | $1,000 | $500 |
Below these values the EA cannot risk your target percentage. It has to trade 0.01 lot, so the real risk is higher than the setting says. Take an EA with a 100-pip stop on a $200 account: one 0.01-lot loss is $10, or 5% of equity, whatever "Risk = 1%" says. If the EA stacks several positions (grid, multi-pair, multiple magic numbers), multiply accordingly.
Rule of thumb from the table: the real minimum deposit is set by your widest stop and your maximum concurrent positions, not by the EA vendor's headline number.
3. Losing streaks: compounding drawdown
With fixed-fractional risk r, equity after k consecutive losses is (1 − r)^k , so:
Drawdown after k losses = 1 − (1 − r)^k
| Risk/trade | 5 losses | 10 losses | 15 losses | 20 losses |
|---|---|---|---|---|
| 1% | 4.9% | 9.6% | 14.0% | 18.2% |
| 2% | 9.6% | 18.3% | 26.1% | 33.2% |
| 5% | 22.6% | 40.1% | 53.7% | 64.2% |
And the recovery needed afterwards is 1 / (1 − DD) − 1 :
| Drawdown | Gain needed to recover |
|---|---|
| 10% | 11.1% |
| 20% | 25.0% |
| 30% | 42.9% |
| 50% | 100.0% |
How long a streak should you expect?
People routinely underestimate this. Below is the probability of seeing at least one losing streak of length ≥ k within 500 trades (computed exactly with a Markov-chain recursion, not simulated):
| Streak length | Win rate 45% | Win rate 60% |
|---|---|---|
| 6 | 99.9% | 71.0% |
| 8 | 85.4% | 17.7% |
| 10 | 43.3% | 3.0% |
| 12 | 15.6% | 0.5% |
| 14 | 5.0% | — |
A 45%-win-rate system (common for trend-following EAs with a reward:risk ratio above 1) will almost certainly hit 8 losses in a row over 500 trades, and has a roughly 43% chance of 10. At 2% risk, those 10 losses alone are an 18.3% drawdown. That isn't a sign the strategy is broken. It's what the strategy normally looks like.
4. Max drawdown and risk of ruin
Streaks are only part of it. Real drawdowns are choppy sequences of losses mixed with small wins. To estimate them I ran a Monte Carlo:
Illustrative model: win rate 45%, reward:risk 1.5 (positive expectancy of +0.125R per trade), 1,000 trades, fixed-fractional compounding, 20,000 simulated paths (seed fixed). "Ruin" here means a 50% peak-to-trough drawdown, the point where most traders switch the EA off.
| Risk/trade | P(≥50% DD) | Median max DD | 95th pct max DD |
|---|---|---|---|
| 1% | 0.01% | 18.8% | 30.0% |
| 2% | 7.6% | 35.0% | 52.6% |
| 5% | 97.5% | 69.6% | 88.1% |
| 10% | 100.0% | 94.7% | 99.6% |
Two things stand out:
- This strategy has a positive edge, and at 5% risk it still almost always hits a 50% drawdown. A real edge doesn't save you from oversized risk.
- Doubling the risk from 1% to 2% does far more than double the danger. Ruin probability goes from about 0.01% to about 7.6%.
Turning drawdown into a capital number
Turn it around. Decide the largest drawdown in dollars you will tolerate before you intervene, then size the account from the expected drawdown, not the median one:
RequiredCapital = MaxTolerableLoss_$ / Expected_MaxDD%
Using the 95th-percentile column: if you'd stop the EA after losing $1,500, at 1% risk (95th pct ≈ 30%) you need about $5,000. At 2% risk (≈ 52.6%) you need about $2,850, but you've also accepted a roughly 1-in-13 chance of a 50% drawdown. Neither answer is "$100".
If your EA's backtest shows a max DD of X%, treat it as a lower bound. A single historical path is one draw from a distribution like the one above, and the live future isn't obliged to be kinder than the sample.
5. Margin and leverage: a separate constraint
Risk-per-trade is about the stop loss. Margin is about whether the broker lets the position exist at all, and whether a stop-out happens before your stop loss.
Margin = Lots × ContractSize × Price / Leverage
For EURUSD at 1.0800, 0.01 lot = 1,000 units = $1,080 notional:
| Leverage | Margin per 0.01 lot | Margin per 0.10 lot |
|---|---|---|
| 1:30 | $36.00 | $360.00 |
| 1:100 | $10.80 | $108.00 |
| 1:500 | $2.16 | $21.60 |
Worked example: a grid EA holding 8 positions of 0.05 lot (0.40 lot total) on a $500 account at 1:30 ties up $1,440 of margin. It can't open the grid at all. At 1:500 it needs only $86.40, so the grid opens fine, and now the account is exposed to 0.40 lot × $10 = $4/pip. A 125-pip adverse move wipes out the whole $500. High leverage doesn't reduce risk. It just takes away the margin limit that would have stopped the trade.
Keep an eye on margin level ( Equity / Margin × 100 ). Many brokers start stop-out somewhere around 50% or lower. Check your broker's exact level, because it's the line where positions get closed for you at the worst available price.
6. A practical checklist
Before choosing a deposit for any EA:
- Find the widest stop and the maximum concurrent exposure (all positions, all symbols) the EA can hold.
- Check the 0.01-lot floor with MinEquity = MinLot × SL × PipValue / Risk% . If your account is below it, your real risk is higher than the setting.
- Estimate realistic losing streaks from the EA's win rate. Don't use the backtest's longest streak.
- Size from a tail drawdown (95th percentile or worse), not the median or the single backtest figure.
- Check margin at full exposure on your actual leverage, and how far that is from your broker's stop-out.
- Decide your intervention line in dollars now, before the drawdown, while you're still thinking clearly.
Monitoring the numbers live
All of this is planning maths. Once the EA is live, what matters is whether current floating drawdown, margin pressure and exposure are still inside the limits you planned for. That's the problem I'm working on with Virtual Risk Desk: a read-only MT5 agent and on-chart panel that reports account drawdown, margin pressure and exposure to a risk dashboard. It doesn't trade, and it doesn't replace doing the sums above first.
All figures are illustrative examples computed from the stated assumptions (EURUSD-style $10/pip per lot, 0.01 minimum lot, fixed-fractional sizing, simplified win/loss model with no spread, slippage, swap or correlation). They are not backtests, forecasts or performance claims. Run the same formulas with your own EA's statistics and your broker's contract specifications.
