You found an edge in the data. Positive expectancy. Statistically significant. You're ready to trade it. You're about to lose money anyway.

An edge that cannot survive its own costs is not an edge. It's a rounding error with a story attached.

The fantasy #

Most traders imagine the path goes like this:

1. Find a pattern that works.
2. Backtest it. See profits.
3. Fund an account.
4. Collect money.

What actually happens for most people who get past step 1: they discover that the distance between "statistically significant" and "actually profitable after costs, slippage, sizing, and drawdowns" is enormous. Most never cross it.

Finding a signal is not the hard part. Keeping it alive through everything the market does to erode it — that's the hard part.

Costs are not a line item — they're a filter #

Most traders think of commissions and fees as a fixed overhead. A few dollars per trade. The cost of doing business.

That framing is wrong, and it can be fatal.

What matters is cost relative to edge. If your strategy makes an average of $5 per trade gross and costs $4 per trade all-in, your net edge is $1. That's an 80% cost ratio. One tick of extra slippage — a slightly wider spread, a partial fill, a microsecond of latency — and your edge is gone.

Think about what "all-in" actually means for a futures trade: exchange fees (per side), clearing fees, broker commission, NFA regulatory fee, slippage, and spread cost if you're crossing. Most traders see one line on their statement and think that's the number. It isn't.

For a single E-mini S&P 500 round trip, a realistic all-in cost is roughly $25–30. For a 10-Year Treasury Note, it's closer to $34. These aren't controversial numbers. They're what the fee schedules and observed fills actually produce.

A strategy that makes $130 per trade gross on ES has real margin. A strategy that makes $35 per trade gross on ZN is already underwater. The signal might be "real" in both cases — statistically distinguishable from noise. But one survives costs and the other doesn't.

Most edges traders find are closer to the ZN case than the ES case. The gross per-trade profit is thin enough that costs eat most or all of it. And because most people never calculate their true all-in costs, they don't know it's happening. They just see a slow bleed and assume the strategy stopped working.

Statistical significance is not the same as tradeable #

In academic finance, the t-statistic measures how confident you can be that an observed pattern is real rather than noise. Higher is better. A t-stat of 2 is the traditional bar for "probably real."

The problem: if you test 50 strategies and pick the best one, the expected maximum t-stat from pure chance alone is about 2.3. Your "best" strategy can clear the bar for significance without any real edge at all. It's just the winner of a random tournament.

Harvey, Liu, and Zhu argued in the Review of Financial Studies (2016) that for a genuinely new trading factor, you need a t-statistic above 3 — specifically because you tried many things and picked the winner. Most published strategies don't come close once you account for all the variants the researcher tested and threw away.

This is multiple testing bias, and it's everywhere. Optimizing indicator parameters over historical data is a test. Trying different entry rules until one "works" is a test. Switching instruments to find the one that fits is a test. Every time you look at a result and decide whether to keep it, you've run a test. The more tests you run, the higher the bar for the winner to mean anything.

Most traders run hundreds of these and count none of them.

The regime problem #

Say you clear all those hurdles. You have a genuine edge. It survives costs. It clears a reasonable statistical bar.

It probably doesn't work all the time.

Strategies that profit from mean reversion — fading moves, buying dips, selling rallies — tend to make their money in high-volatility environments. The same strategy that prints during a crisis can bleed for years in a calm market.

That creates a brutal psychological trap. Your edge might be real but dormant for two or three years at a time. During those years, you're losing money — or barely breaking even — on a strategy that you correctly identified as having positive expectancy. The math is right. Your account is still going down.

If your strategy made 62% of its total profit in a single year out of eleven, that's not a diversified income stream. That's a lottery ticket with better odds than most. You'll need to survive ten lean years to capture the one good one. Can your account handle that? Can your psychology?

The sizing trap #

This is where it gets dangerous.

The Kelly Criterion gives you the theoretically optimal fraction of your account to risk per trade, given your edge and your win/loss ratio. Full Kelly maximizes long-run growth. It also produces drawdowns of 50–75%. Nobody survives that psychologically. Most accounts don't survive it financially either.

So people use fractional Kelly — half, quarter, or less. But fractional Kelly depends on an accurate estimate of your edge. And your estimate is noisy. It's always noisy — you're working from finite historical data.

A strategy with a holdout t-stat of 1.7 has a standard error of roughly 37% on its mean. You don't know if the true edge is $130/trade or $50/trade or $10/trade. Quarter-Kelly based on $130 puts you in a completely different risk regime than quarter-Kelly based on $50. Same formula, wildly different outcomes, and you have no way to know which input is right.

This is why professional risk managers don't use Kelly for thin edges. They use fixed fractional risk — 0.1% to 0.25% of the account per trade, regardless of the estimated edge. Less "optimal" in theory. Far more survivable in practice.

The forced liquidation spiral #

This is the scenario that actually kills accounts.

Your strategy has a disaster stop — a hard exit no matter what. On ES, a 3-ATR stop at current volatility means roughly $4,500 of risk per contract. Sounds manageable on a $100K account.

Then the market gaps through your stop overnight. Your planned $4,500 loss becomes $7,000 or $9,000. Your broker liquidates at the worst available price because you've hit maintenance margin.

Now you're undercapitalized. You can't take the next signal at full size. You trade smaller, or you skip it entirely. The recovery trade that would have made back the loss? You weren't in it.

And then it compounds. The gap loss exceeds planned loss. The account drops below sizing threshold. The next trade is undersized or skipped. The recovery passes you by. Each loss makes the next one more likely to trigger another forced liquidation.

The math of ruin is asymmetric. A 50% drawdown requires a 100% gain to recover. A 75% drawdown requires a 300% gain. The deeper the hole, the more improbable the escape.

At a $25K account trading one ES contract, stress testing shows roughly an 82% probability of ruin. That's not a bad strategy — that's an honest strategy at an insufficient account size.

What to actually do with this #

If this sounds discouraging, good. The alternative is discovering it with real money.

Most traders believe "I just need a good strategy and the money will follow." It's a comforting idea and it's wrong. Strategy selection is necessary but nowhere near sufficient. Costs, sizing, and survival math are the actual gatekeepers, and most people skip all three.

So do the math first. Before you put a dollar into a live account, calculate your actual all-in cost per trade — not the broker's advertised commission, but exchange fees, clearing, slippage, and spread. Compare that number to your realistic per-trade edge. If costs eat more than half your gross edge, you don't have a tradeable strategy. You have an interesting observation.

If the math works:

• Paper trade long enough to confirm your fills and slippage match your assumptions. They won't. Adjust.
• Start live on micro contracts (MES, MNQ, MYM). Don't assume micros cost 1/10th of full-size — the exchange fees don't scale linearly. Check the actual fee schedules.
• Size your account so that your worst realistic single-trade loss — a gap through your stop, not your planned stop — represents no more than 1–2% of capital.
• Accept that "works but not consistently enough to live on" is an honest outcome for a strategy. Most edges belong in a diversified book, not as standalone income.

Related reading #

• Your Backtest Doesn't Know You Exist — the simulation gap that inflates every edge estimate.
• And You're Failing the Test — why structure beats willpower, especially during drawdowns.

Open questions #

• What's your actual all-in cost per trade? Have you ever calculated it down to the penny?
• If your strategy's best year contributed more than 40% of its total profit, how would you survive the other years?
• How do you distinguish between "this strategy stopped working" and "this strategy is in a drawdown within its normal behavior"?

This post is educational content about trading concepts. It is not investment advice. The numbers used are illustrative of general principles, not recommendations for specific trades or account sizes. See our terms for more.