Distribution figures below describe how the Shishin backtest’s 404 positions clustered across the five-year window. They are not a projection of any forward trade outcome. Live execution will produce a different distribution for reasons that include slippage, fills, and the normal divergence of sample from model.
The narrative around systematic trading is the ten-bagger. The experience is the boring middle: most trades are neither triumphs nor blow-ups, they settle a few percent either side of zero, and close. Both are true at once, and the interesting part is how. The middle is where the system spends its time. The tail is where it makes its money. The middle’s job is to make waiting for the tail cheap.
The shape of the distribution
Across the full five-year backtest, the four-engine stack took 404 positions. Fewer than half made money (46%), and the median trade lost about 2%. Two trades in three finished within ten percent of where they started. If you watched the book trade by trade, that is what you would see: a long run of small, forgettable outcomes.
It is not a normal distribution. It is sharply right-skewed. Stops compress the loss side into a tight cluster: the worst tenth of trades lost between 9.5% and 18.5%, and 39 of those 40 were stop-outs. The win side has no such ceiling. The best tenth gained between 26% and 355%.
Where the money actually comes from
Split the 404 positions into deciles by outcome and add up the dollars:
- The worst 10% (40 trades) cost about 18% of the final net profit.
- The middle 80% (324 trades, from -9.5% to +24.6%) earned about 15% of it.
- The best 10% (40 trades) earned about 103% of it, more than the whole net, because they also paid for the losers.
So the curve is not built in the middle. It is built in the tail, by roughly forty trades over five years, about eight a year. We say this plainly because the opposite claim is the comfortable one, and our own ledger contradicts it. Leave out the winners tests how much of the edge survives without them.
So what is the middle for?
If the middle earns only a sliver, why take those trades at all? Because nobody can tell in advance which setups will become the tail. The ten-baggers looked like every other entry on the day they were bought. The only way to own the forty is to take the four hundred. The middle is the cost of being in the room when the tail arrives, and the system’s real achievement is that this cost is slightly negative: the middle 80% paid for itself and a little more.
That is the standard to hold a momentum or breakout system to. Not “does it find great trades”, it will, occasionally, by participating, but “how little does it cost to keep participating”. A system whose middle bleeds will either run out of capital or run out of the operator’s patience before the next tail trade shows up.
What changes in practice
Three operational habits fall out of taking the distribution seriously.
- You stop chasing. The trade that looks most obvious in real time is no more likely to be the outlier than any other. Sizing up because a setup feels strong is a guess about which trade lands in the tail, and the data does not support the guess.
- You do not get selective. Skipping setups to avoid the boring ones mostly skips tail trades too, because you cannot tell them apart in advance. A missed tail trade costs far more than the middle saves.
- You design for the floor. Stops, position caps, and re-entry rules exist so the worst trades stay small and the middle stays cheap. If a single loser can swallow a month of ordinary trading, the system cannot afford to wait for its tail.
What the tail trades look like
The asymmetry shows up in time as well as size. The best tenth was held a median of 18 trading days, and 31 of the 40 exited on the moving-average trail, the rule that lets a winner run until the trend breaks. The worst tenth was held a median of 3 days: the stop hit early, the position closed, the system moved on. The middle sits between, at about nine days.
Long winners, short losers: that is the structural shape of any expectancy-positive trend-following strategy, not something unique to Shishin. What matters is the discipline to leave it alone: to let the trail, not a feeling, decide when a winner is done, and to accept the stop on the losers without second-guessing it. Every trade is in the public ledger if you want to check the shape yourself.
The institutional point
Quant marketing rewards content about the spectacular trade, and the ledger says the spectacular trades really are where the money is. The trap is drawing the wrong lesson from that. The operator who tries to trade only the tail ends up over-fitting to it, filtering out ordinary-looking setups that the next tail trade was going to come from. The boring middle is not where the money is made. It is what lets you be there when it is.