This article is a statement of philosophy, how and why we run a rules-based system, not investment advice. It is educational and general, not a recommendation to buy, sell, or size any position, and nothing here tells you what to do with your own money.
“I have a strong conviction on this one.” It is the most expensive sentence in investing. Conviction feels like insight. Often it is fear, greed, and overconfidence wearing the costume of insight. A rule cannot feel any of those things, and that is precisely the point.
The short version
Over a long run, a validated rule tends to beat human conviction because conviction is where our biases do their damage: we sell winners too early, marry losers, and abandon a plan at the exact moment it is hardest to hold. A rule does the same thing every time and can be tested in advance. The honest catch: a rule can be wrong too, and a curve-fit rule is just overconfidence made permanent, so the rule has to earn trust before it earns capital.
Conviction is where the biases live
The case against discretion is that humans are predictable in the wrong ways, and markets are very good at finding those ways. Decades of behavioural research keep landing on the same short list, and every item on it lives inside the feeling we call conviction.
The disposition effect, documented by Hersh Shefrin and Meir Statman, is the tendency to sell winners too soon to lock in a gain and to hold losers too long to avoid booking a loss. It is the exact opposite of what a trend-following edge needs, and it feels responsible while you do it. Overconfidence is the well-measured gap between how sure we are and how often we are right, and it is worst precisely on the positions we are most sure about. Then there is fear and greed on a timer: the pull to add risk after a run of green (when the odds are quietly worsening) and to capitulate at the bottom of a drawdown (when the plan most needs to be followed). We catalogue the full list in the behavioural biases piece; the through-line is that none of them announce themselves as bias. They all arrive dressed as a strong, reasonable conviction about this particular name, right now.
A rule is consistent, and consistency is the edge
Set the biases aside for a moment and you are still left with the plain mechanical argument: a rule does the same thing in the same conditions every single time. It does not sleep badly, read the news, or care that the last three trades lost. That boring consistency pays. In a domain where a real edge is small and only shows up over many repetitions, actually taking every signal the method generates, including the uncomfortable ones, is most of what turns an edge on paper into an edge in an account.
Consistency also buys something conviction can never offer: it is testable. Because a rule is written down before the trade, you can run it across years of history and across regimes and ask whether it actually paid, then measure how it behaved in the bad stretches, not just the good ones. A conviction cannot be backtested, because it did not exist until the moment it was felt, and by then the outcome is already unfolding. A rule can be examined in the cold, in advance, by someone other than the person who wants it to be true. That is the difference between a hypothesis and a hunch.
The honest counter: a rule can be wrong too
Here is the part that the mechanical-is-always-better crowd skips, and we will not. A rule is not virtuous simply because it is a rule. A bad rule executed with perfect discipline just loses money reliably instead of erratically. Two failure modes matter most.
The first is that the rule is simply wrong about the world: the relationship it leans on was never real, or it was real once and the market has since adapted around it. The second is subtler and more seductive. A rule tuned until it fits the past beautifully is overconfidence crystallised into code: a curve fit that has memorised the noise of one history and will meet the next one defenceless. The discretionary trader at least knows they are guessing. The over-fit rule dresses the same guess up in an equity curve and a confidence it has not earned. We take that trap seriously enough to have written a whole piece on it, why backtests lie, because the failure that looks most like proof is the one that fools you hardest.
So the honest claim is narrower than “rules win.” It is this: a rule that has been validated, out of sample, across regimes, and judged on the drawdowns as well as the returns, beats conviction over time, mostly by removing the panic and the self-flattery. An unvalidated rule earns none of that. The discipline is faith in mechanics that have survived a genuine attempt to break them.
Where Shishin stands
This is the whole reason we are built the way we are. Shishin is a systematic, non-advisory service: a daily ranked board produced by a four-engine, regime-routed model, run to written rules. The macro switch decides which of the four guardians, Genbu, Suzaku, Byakko, or Seiryu, is appropriate for the conditions, and the same logic runs whether the last week was calm or ugly. There is no seat at the desk for a person to override the process because this trade “feels different.” That discretionary override, the moment a human overrules the system on conviction, is exactly where the behavioural research says most of the damage gets done, so we designed the seat out.
Notice the deliberate limit on what that gives you. A ranked board is research, not a set of instructions: it never tells anyone what to buy, sell, size, or when to act. Removing conviction from our process is not a licence to hand you ours in its place. And because “trust the rules” is a claim, not a courtesy, we hold ourselves to the counter above. The five-year backtest is a hypothesis with its caveats stated, and the live results are paper-traded in the open, wins and losses together, on the public record and independently attested at the verification page. The point of the rules is that you do not have to take our conviction on faith. You can check the record they produced.
One more honest note on conviction, because the word is doing two jobs. There is the discretionary conviction we are arguing against, a feeling that overrides the plan, and there is the structured kind, sizing a position by how strongly the method itself rates it. Those are not the same thing, and we keep them apart in position sizing by conviction. The rule can express a graded, rules-based conviction without ever handing the wheel back to a gut feeling.
The bottom line
Unexamined conviction is the enemy of good investing. The value of a rule is that it makes the conviction explicit, testable, and consistent, then holds it there when fear and greed would pull it apart. That only works if the rule has actually been validated, because a rule can be wrong, and a curve-fit one is confidently wrong. Get that part right and the mechanical version wins the way the tortoise wins: not by being brilliant on any single trade, but by refusing to be stupid on the ones that matter.
Sources & further reading
- Shefrin, H. & Statman, M. (1985). “The Disposition to Sell Winners Too Early and Ride Losers Too Long.” The Journal of Finance, 40(3), 777 to 790.
- Barber, B. M. & Odean, T. (2000). “Trading Is Hazardous to Your Wealth.” The Journal of Finance, 55(2), 773 to 806.
- Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux.
- Shishin research: behavioural biases, position sizing by conviction, and why backtests lie.