Compounding and Volatility Drag: When a Positive Average Still Loses

Written by chwoo · Crypto trader, 7 years. Running an automated futures bot since spring 2026. · 2026-07-07 · updated 2026-09-24

Compounding is usually sold as the friendly half of finance: leave money alone, let returns earn returns, retire. In trading it has a second face. The same mechanism that turns small gains into large ones turns a healthy-looking average trade into a losing account, and it does so quietly enough that most people never find the leak.

So rather than describe it, we measured it on this site's eight presets. The question was simple: if you take the average return of every trade a strategy made, does that average tell you what the account did?

Two strategies with a positive average trade that still lose

BTC, each preset on its default timeframe, 90% position size, realistic costs, data through 21 September 2026. "Promised" is what you would get by earning exactly the average trade, every trade, for the whole run.

Open the strategy this guide measures in the backtester →

Start with the Donchian breakout. Its average trade is positive, +0.010%, and over 480 trades that average promises +4.3%. The account finished at -43.1%. The trend-filtered RSI does the same thing on a smaller scale: +0.035% per trade on average, and -0.2% at the end.

Nothing was subtracted that the average did not already include. Costs are in those trade returns. The gap is created purely by the order of operations: gains and losses multiply rather than add, and a loss removes capital that every later gain would have been applied to.

Supertrend momentum shows how large that distance gets on a strategy that does work. Its average trade of +0.616% promises +717.7% across 380 trades. It delivered +193.6%. The average overstated the outcome by more than 500 percentage points, and the strategy is genuinely profitable.

Why bigger positions stop paying proportionally

Supertrend momentum again, changing only the share of the account committed to each trade.

Read the third and fourth columns together. At 25% and 50% the return multiple is larger than the size multiple (2.62 against 2.5, and 5.28 against 5.0), which is compounding working for you. From 75% onward it falls behind, and at full size a tenfold position buys only 8.04 times the return while the drawdown grows from 4.9% to 40.9%.

That crossover is drag becoming visible. Bigger positions make each individual loss remove a bigger share of the capital that the next win has to work with. There is a size past which you are buying drawdown and giving back return, and on this strategy it sits somewhere between half and three quarters of the account.

Where we were wrong about order

We expected trade order to matter to the final balance, on the reasoning that an early loss compounds differently from a late one. To check, we shuffled the 380 trades 200 times and recompounded each shuffle.

Every one of the 200 shuffles finished at exactly +193.6%.That is not a rounding artifact. Final equity is the product of every trade's multiplier, and multiplication is commutative, so the order genuinely cannot change where you end up. Our expectation was simply wrong, and the table is the correction.

The drawdown column is the part that does move, and it moves a lot: from 37.0% to 90.1% depending only on the sequence. Note where the real history sits. The actual order produced a 37.5% drawdown against a median shuffle of 63.2%. The historical path was close to the luckiest arrangement of the same trades. A strategy with this trade distribution can easily deliver a 63% drawdown in the future while returning exactly the same +193.6%.

Symmetric moves are not symmetric

The cleanest demonstration needs no strategy at all. Take an account that gains x% and loses x%, alternately, one hundred times.

Fifty wins and fifty losses of the same size is a strategy with a 50% win rate and a payoff ratio of exactly 1. It loses in every row, because a 30% gain does not undo a 30% loss: 1.3 × 0.7 is 0.91. At ±30% the account keeps 0.9% of what it started with.

This is why volatility itself costs money, and why a smoother strategy with a lower average trade can beat a wilder one with a higher average trade. It is also the arithmetic behind the leverage results in our position sizing guide: leverage multiplies the swing size, and this table shows what swing size does.

The patient side of the same mechanism

Compounding pays as reliably as it charges. How many trades a given edge needs to double an account:

A half-percent per trade sounds like nothing and doubles the account in 139 trades. The catch is the first table: that has to be the geometric mean, not the average trade. Supertrend momentum has an average trade of 0.616% and a geometric mean of 0.284%, so the honest row to read for it is the 0.25% one, not the 0.5% one.

What to do with this

• Judge a strategy by its total return, not its average trade. Two of the eight presets have a positive average and a losing account.
• Treat the gap between the two as a cost. It is volatility drag, it grows with position size, and no amount of trading skill removes it.
• Look for the size where return stops keeping up with size. On the strategy above it is between 50% and 75%, past which you buy drawdown and give back return.
• Do not read the backtest's drawdown as typical. The same 380 trades in a different order gave a median of 63.2% against the 37.5% history actually produced.
• Prefer smoother to higher. The ±30% table loses 99.1% at a 50% win rate and a payoff of 1.