Account size reality check: what can a $25k account actually hold?

The STS live NQ book, run at its full mini-scale sizing, draws down past a $25,000 account 99.9% of the time. That is out of 10,000 Monte Carlo reshuffles of the STS live 15-year record (3,505 real trades, 2011 to 2026, traded as a one-to-three-contract cascade that averages 1.81 contracts). That same book needs about $62,000 to hold a bad-but-normal (95th-percentile) drawdown, and about $75,000 to hold a rough (99th-percentile) one. So the honest answer to "can I run this book at full mini scale on $25k?" is no. Not this book, not at that size.

That is the whole article in three numbers. The rest is the proof, the fix for small accounts (run the same book on micros), and a sizing rule you can copy.

These buffers belong to one specific book

Every number below comes from one specific place: our live NQ book, five systematic strategies (S1, S2, S3, S5, S6) trading as one single-position account from 2011 through 2026, backtested in TradingView at one to three volatility-scaled contracts (it averages 1.81 contracts per trade). When we say "full mini scale" we mean that book as it actually traded, in NQ minis, not a hypothetical steady one-lot. These are momentum and trend-continuation entries, intraday plus one overnight sub, not mean reversion and not scalping. The drawdown distribution you are about to see is a property of that trade mix and that sizing. A book that trades a different style, a different size, or a different instrument has its own distribution and its own buffers.

So take the method, not the exact dollars. If you trade NQ discretionarily, the rule that transfers is this: size against the reshuffled drawdown distribution, not against the single realized drawdown you happened to live through. The specific figures here, like the $6,208 per-micro 95th-percentile buffer derived below, are ours. Yours will differ. The way to find them is to run the same reshuffle on your own trade record.

One sleeve drives most of the small-account drawdown

The buffer is not shared equally across the five strategies. We ran a leave-one-out test: drop each sub, re-run the 10,000-path Monte Carlo, and measure how the 95%-case drawdown of the full book changes.

Removing the Trend sleeve (S1) cuts the 95%-case drawdown by $6,543, the single largest effect of any sub. Trend is the main engine of the drawdown that drains a small account. The Overnight (S5) and Universal (S6) sleeves run the other way: dropping either one raises the drawdown, by $1,667 and $956, so they are net drawdown-dampers, not drivers.

One honest note on lenses. In the single realized 15-year path, dropping Overnight actually lowered the $28,994 drawdown, because its trades happened to land inside the one worst dip we lived through. Across 10,000 forward orderings it damps the drawdown instead. The forward Monte Carlo is the decision-relevant lens here, the same reason we size against it and not the realized 17.5%.

The survival table: size against capital

We took our real 3,505-trade portfolio, reshuffled the order of those trades 10,000 times, and measured the worst peak-to-trough drawdown of each reshuffled path. Then we asked one question for every account size and every position size: how often is that drawdown bigger than the money in the account?

That fraction is how often an ordinary reshuffled drawdown would have swallowed the whole account over the 15-year path. It is a full-path max-drawdown exceedance, not a time-to-ruin or blow-up-this-year probability. Here it is.

The bottom row is the book at full mini scale (its live one-to-three-contract cascade). Each micro row is that same book run at a fraction of mini size: since one MNQ micro is exactly one tenth of an NQ mini per point, "k micros" means the identical cascade scaled to k/10 of full size.

Read the bottom row. At full mini scale the book outdraws a $50k account better than one time in five (22.4%), and only turns green near $100k (0.1%). Read the top row. One micro is safe from $10k up. The account does not decide whether you make money. It decides whether a normal losing stretch closes you before the edge pays off.

Why the honest number is the Monte Carlo, not the 17.5%

Our book's realized maximum drawdown over 15 years was $28,994, which is 17.5% of the peak. That is the number on the tear sheet. It is real. It is also the wrong number to size against.

Here is why. That $28,994 came from one particular order of wins and losses, the order history happened to deal us. When we reshuffle those same trades 10,000 times, the realized $28,994 lands at about the 2nd percentile of the drawdowns. In plain terms, the order we lived through was luckier than 98% of the alternatives. The median reshuffle drew down about $42,021, and the 95th-percentile reshuffle about $62,078.

So if you size against the 17.5% you saw, you are betting the next 15 years deal you a top-2% ordering again. The forward drawdown is expected to be bigger. This is the same reason a live drawdown almost always runs deeper than the backtest, and why we stress-test and reshuffle every result before we trust it. Size against the distribution, not against the one path you got.

Micros are how a small account holds the same edge

If the full-scale book needs about $62k, a $10k or $25k account is not shut out. It just runs the same signals in a smaller instrument. NQ moves $20 per index point, MNQ moves $2, exactly one tenth (verified on TradingView's contract specs and cross-checked against the E-mini reference on Wikipedia). So running the identical cascade on MNQ micros instead of NQ minis divides every dollar, and every drawdown, by ten. Trading one tenth of full scale is one micro's worth of exposure.

That splits the sizing problem into small steps. Here is the same distribution, scaled per size.

One micro's bad-but-normal drawdown is about $6,208. A $10k account holds it with room, which is why the top row of the survival table is green. Step up to two or three micros and the drawdown grows in lockstep, so a $10k account starts to strain: it is outdrawn 22.4% of the time at two micros, 88.7% at three. The choice is not full-scale-or-nothing. It is how many micros your account can carry, and the answer to mini-versus-micro sizing is spelled out in NQ vs MNQ: which to trade.

The copy-paste sizing rule

Here is the rule we use, straight off the distribution above. Hold enough capital per contract to cover the 95%-case drawdown, or the 99%-case if you want more margin.

Worked examples. A $25k account holds floor(25000 / 6500) = 3 micros at the 95% rule. A $50k account holds 7. A $100k account holds 15 micros, which is one and a half times full mini scale. And a $25k account trying to run the book at full mini scale is asking a $62k-sized position to live in a $25k box, which is exactly the 99.9% cell in the table.

Two things this does not model

The Monte Carlo reshuffles the same 15 years of real trades. It measures sequence risk, the luck of the order wins and losses arrive in. It does not invent a market the book has never seen.

First, the reshuffle treats every trade as swappable in order. That is the standard trade-shuffle assumption, and it has a known blind spot: if losses in the real book cluster together, say a bad regime that stacks five losers in a row, shuffling can break up that cluster and understate the deepest drawdowns.

So we tested it directly. We re-ran the Monte Carlo with a moving-block bootstrap, which resamples the book in blocks of 20 consecutive trades and keeps every losing streak intact (our longest historical streak is 14 trades). If clustering deepened the drawdown, the streak-preserving version would run past our published buffer. It runs the other way. Keeping the streaks intact, the 95%-case mini drawdown is $47,027, well inside the $62,078 we size against, and even the streak-preserving 99%-case ($54,870) stays under it. Every block length we tried, and a second bootstrap family, landed below the iid number.

We lean conservative on purpose. The realized ordering we lived through was milder than 98% of the shuffles, and the iid buffer sits above the streak-preserving one, so the published numbers push the buffers up, not down. A book with heavier loss-clustering than ours should still trust its p99 line, not the median.

Second, a regime worse than anything between 2011 and 2026 sits outside this model entirely. The buffers here are a floor for behavior we have data on, not a promise. If NQ enters a stretch nastier than 2022 was for our overnight sub or 2015 was for our worst 14-trade streak, the real drawdown can run past the 99th percentile shown here. Treat these numbers as the minimum, and keep a margin past them.

Prop-firm accounts ask the same question

If you trade a funded account instead of your own cash, the survival math is identical. It just has a different name. A prop firm's trailing drawdown limit is an account buffer, and the question is still whether an ordinary losing stretch is bigger than the room you are given. The reason profitable traders fail evaluations so often is that their normal drawdown does not fit the buffer, which we work through in the prop-firm trailing drawdown math. Same distribution, same rule, smaller box.

What we test next

The obvious next step is regime-conditional sizing: instead of one buffer for all 15 years, compute the drawdown distribution inside high-volatility stretches versus quiet ones and size to the regime you are in. That would let a smaller account carry more micros in calm markets and fewer in storms, rather than holding one static buffer for everything. It needs a regime label we trust before we will publish a number on it, so for now the static buffers above are the honest floor.

The trades behind all of this live on the strategy page and the live-tracked tear sheet. If you want the signals that generated the 3,505-trade record we sized here, that is the one thing we sell, starting with a free trial.

How we measured this

Instrument: CME E-mini Nasdaq-100 (NQ), scaled to MNQ micros as mini dollars divided by ten. Data: our live five-strategy intraday book, the verified single-position TradingView export, 3,505 trades from 2011-06 through the 2026-06-11 export, traded in NQ minis at its one-to-three-contract volatility-scaled cascade (mean 1.81 contracts per trade, so "full mini scale" is this book as traded, not a flat one-lot), commissions and slippage included in the backtest, $1,120,402 net. Method: shuffle the 3,505 realized per-trade net P&L values 10,000 times with a deterministic seed, take the maximum peak-to-trough drawdown of each cumulative path, and report percentiles. For each account-and-size cell, count the fraction of the 10,000 paths whose drawdown, scaled by the micros held, exceeds the account. We checked the distribution four independent ways. A different-seed shuffle (median $42,140, p95 $62,283, under half a percent of drift) and a resample-with-replacement bootstrap (median $42,195, p95 $65,018) reproduced the headline cells within about two points. A streak-preserving moving-block bootstrap that keeps every losing run intact came in lower (p95 $47,027), which tells us the iid buffer we publish is the conservative one. And a fill-noise stress that perturbs every entry and exit by up to four ticks moved the 95%-case buffer by under 0.1%, so the $62k figure is not an artifact of exact fills. What this measures is sequence risk over a full 15-year path, not a first-passage time-to-ruin, and not a regime the book has not traded. What would change the numbers: a trade distribution with deeper peak-to-trough dips than ours, which would push every buffer higher.

STS Research. Educational content, not investment advice. We trade this book live and sell access to its signals; judge the data accordingly.

CFTC Rule 4.41: Hypothetical or simulated performance results have certain limitations. Unlike an actual performance record, simulated results do not represent actual trading. Also, since the trades have not been executed, the results may have under- or over-compensated for the impact, if any, of certain market factors, such as lack of liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any account will or is likely to achieve profit or losses similar to those shown.

Past performance is not indicative of future results. Futures trading involves substantial risk of loss and is not suitable for every investor.